Authors: Zhenxing Dou, Yijiao Wang, Tao Zou, Zhiwei Chen, Fei Liu, Peng Wang, Weisheng Zhao
Abstract: In recent years, machine learning has been extensively applied to data prediction during process ramp-up, with a particular focus on transistor characteristics for circuit design and manufacture. However, capturing the nonlinear current response across multiple operating regions remains a challenge for neural networks. To address such challenge, a novel machine learning framework, PRIME (Physics-Related Intelligent Mixture of Experts), is proposed to capture and integrate complex regional characteristics. In essence, our framework incorporates physics-based knowledge with data-driven intelligence. By leveraging a dynamic weighting mechanism in its gating network, PRIME adaptively activates the suitable expert model based on distinct input data features. Extensive evaluations are conducted on various gate-all-around (GAA) structures to examine the effectiveness of PRIME and considerable improvements (60\%-84\%) in prediction accuracy are shown over state-of-the-art models.
Authors: Tianyu Sun
Abstract: We develop Policy Gradient with Second-Order Momentum (PG-SOM), a lightweight second-order optimisation scheme for reinforcement-learning policies. PG-SOM augments the classical REINFORCE update with two exponentially weighted statistics: a first-order gradient average and a diagonal approximation of the Hessian. By preconditioning the gradient with this curvature estimate, the method adaptively rescales each parameter, yielding faster and more stable ascent of the expected return. We provide a concise derivation, establish that the diagonal Hessian estimator is unbiased and positive-definite under mild regularity assumptions, and prove that the resulting update is a descent direction in expectation. Numerical experiments on standard control benchmarks show up to a 2.1x increase in sample efficiency and a substantial reduction in variance compared to first-order and Fisher-matrix baselines. These results indicate that even coarse second-order information can deliver significant practical gains while incurring only D memory overhead for a D-parameter policy. All code and reproducibility scripts will be made publicly available.
Authors: Diego Granziol
Abstract: Whilst there have been major advancements in the field of first order optimisation of deep learning models, where state of the art open source mixture of expert models go into the hundreds of billions of parameters, methods that rely on Hessian vector products, are still limited to run on a single GPU and thus cannot even work for models in the billion parameter range. We release a software package \textbf{HessFormer}, which integrates nicely with the well known Transformers package and allows for distributed hessian vector computation across a single node with multiple GPUs. Underpinning our implementation is a distributed stochastic lanczos quadrature algorithm, which we release for public consumption. Using this package we investigate the Hessian spectral density of the recent Deepseek $70$bn parameter model.
Authors: Tianyi Shi, Zhu Meng, Yue Chen, Siyang Zheng, Fei Su, Jin Huang, Changrui Ren, Zhicheng Zhao
Abstract: Time series forecasting plays a crucial role in various fields, and the methods based on frequency domain analysis have become an important branch. However, most existing studies focus on the design of elaborate model architectures and are often tailored for limited datasets, still lacking universality. Besides, the assumption of independent and identically distributed (IID) data also contradicts the strong correlation of the time domain labels. To address these issues, abandoning time domain supervision, we propose a purely frequency domain supervision approach named cross-dimensional frequency (X-Freq) loss. Specifically, based on a statistical phenomenon, we first prove that the information entropy of the time series is higher than its spectral entropy, which implies higher certainty in frequency domain and thus can provide better supervision. Secondly, the Fourier Transform and the Wavelet Transform are applied to the time dimension and the channel dimension of the time series respectively, to capture the long-term and short-term frequency variations as well as the spatial configuration features. Thirdly, the loss between predictions and targets is uniformly computed in the frequency domain. Moreover, we plug-and-play incorporate X-Freq into multiple advanced forecasting models and compare on 14 real-world datasets. The experimental results demonstrate that, without making any modification to the original architectures or hyperparameters, X-Freq can improve the forecasting performance by an average of 3.3% on long-term forecasting datasets and 27.7% on short-term ones, showcasing superior generality and practicality. The code will be released publicly.
Authors: Pooja Mangal, Sudaksh Kalra, Dolly Sapra
Abstract: Deploying deep convolutional neural networks (CNNs) on resource-constrained devices presents significant challenges due to their high computational demands and rigid, static architectures. To overcome these limitations, this thesis explores methods for enabling CNNs to dynamically adjust their computational complexity based on available hardware resources. We introduce adaptive CNN architectures capable of scaling their capacity at runtime, thus efficiently balancing performance and resource utilization. To achieve this adaptability, we propose a structured pruning and dynamic re-construction approach that creates nested subnetworks within a single CNN model. This approach allows the network to dynamically switch between compact and full-sized configurations without retraining, making it suitable for deployment across varying hardware platforms. Experiments conducted across multiple CNN architectures including VGG-16, AlexNet, ResNet-20, and ResNet-56 on CIFAR-10 and Imagenette datasets demonstrate that adaptive models effectively maintain or even enhance performance under varying computational constraints. Our results highlight that embedding adaptability directly into CNN architectures significantly improves their robustness and flexibility, paving the way for efficient real-world deployment in diverse computational environments.
Authors: Chongyang Tan, Ruoqi Wen, Rongpeng Li, Zhifeng Zhao, Ekram Hossain, Honggang Zhang
Abstract: Federated Learning (FL) enables distributed model training across edge devices in a privacy-friendly manner. However, its efficiency heavily depends on effective device selection and high-dimensional resource allocation in dynamic and heterogeneous wireless environments. Conventional methods demand a confluence of domain-specific expertise, extensive hyperparameter tuning, and/or heavy interaction cost. This paper proposes a Tool-aided Evolutionary Large Language Model (T-ELLM) framework to generate a qualified policy for device selection in a wireless FL environment. Unlike conventional optimization methods, T-ELLM leverages natural language-based scenario prompts to enhance generalization across varying network conditions. The framework decouples the joint optimization problem mathematically, enabling tractable learning of device selection policies while delegating resource allocation to convex optimization tools. To improve adaptability, T-ELLM integrates a sample-efficient, model-based virtual learning environment that captures the relationship between device selection and learning performance, facilitating subsequent group relative policy optimization. This concerted approach reduces reliance on real-world interactions, minimizing communication overhead while maintaining high-fidelity decision-making. Theoretical analysis proves that the discrepancy between virtual and real environments is bounded, ensuring the advantage function learned in the virtual environment maintains a provably small deviation from real-world conditions. Experimental results demonstrate that T-ELLM outperforms benchmark methods in energy efficiency and exhibits robust adaptability to environmental changes.
Authors: Zhen Li, Yupeng Su, Songmiao Wang, Runming Yang, Congkai Xie, Aofan Liu, Ming Li, Jiannong Cao, Yuan Xie, Ngai Wong, Hongxia Yang
Abstract: Large Language Models (LLMs) have demonstrated impressive performance on complex reasoning benchmarks such as GSM8K, MATH, and AIME. However, the substantial computational demands of these tasks pose significant challenges for real-world deployment. Model quantization has emerged as a promising approach to reduce memory footprint and inference latency by representing weights and activations with lower bit-widths. In this work, we conduct a comprehensive study of mainstream quantization methods(e.g., AWQ, GPTQ, SmoothQuant) on the most popular open-sourced models (e.g., Qwen2.5, LLaMA3 series), and reveal that quantization can degrade mathematical reasoning accuracy by up to 69.81%. To better understand this degradation, we develop an automated assignment and judgment pipeline that qualitatively categorizes failures into four error types and quantitatively identifies the most impacted reasoning capabilities. Building on these findings, we employ an automated data-curation pipeline to construct a compact "Silver Bullet" datasets. Training a quantized model on as few as 332 carefully selected examples for just 3-5 minutes on a single GPU is enough to restore its reasoning accuracy to match that of the full-precision baseline.
Authors: Shuchen Wu, Stephan Alaniz, Shyamgopal Karthik, Peter Dayan, Eric Schulz, Zeynep Akata
Abstract: Neural networks are often black boxes, reflecting the significant challenge of understanding their internal workings. We propose a different perspective that challenges the prevailing view: rather than being inscrutable, neural networks exhibit patterns in their raw population activity that mirror regularities in the training data. We refer to this as the Reflection Hypothesis and provide evidence for this phenomenon in both simple recurrent neural networks (RNNs) and complex large language models (LLMs). Building on this insight, we propose to leverage cognitively-inspired methods of chunking to segment high-dimensional neural population dynamics into interpretable units that reflect underlying concepts. We propose three methods to extract these emerging entities, complementing each other based on label availability and dimensionality. Discrete sequence chunking (DSC) creates a dictionary of entities; population averaging (PA) extracts recurring entities that correspond to known labels; and unsupervised chunk discovery (UCD) can be used when labels are absent. We demonstrate the effectiveness of these methods in extracting entities across varying model sizes, ranging from inducing compositionality in RNNs to uncovering recurring neural population states in large models with diverse architectures, and illustrate their advantage over other methods. Throughout, we observe a robust correspondence between the extracted entities and concrete or abstract concepts. Artificially inducing the extracted entities in neural populations effectively alters the network's generation of associated concepts. Our work points to a new direction for interpretability, one that harnesses both cognitive principles and the structure of naturalistic data to reveal the hidden computations of complex learning systems, gradually transforming them from black boxes into systems we can begin to understand.
Authors: Peimian Du, Jiabin Liu, Xiaowei Jin, Mengwang Zuo, Hui Li
Abstract: This research confronts the challenge of substantial physical equation discrepancies encountered in the generation of spatiotemporal physical fields through data-driven trained models. A spatiotemporal physical field generation model, named HMT-PF, is developed based on the hybrid Mamba-Transformer architecture, incorporating unstructured grid information as input. A fine-tuning block, enhanced with physical information, is introduced to effectively reduce the physical equation discrepancies. The physical equation residuals are computed through a point query mechanism for efficient gradient evaluation, then encoded into latent space for refinement. The fine-tuning process employs a self-supervised learning approach to achieve physical consistency while maintaining essential field characteristics. Results show that the hybrid Mamba-Transformer model achieves good performance in generating spatiotemporal fields, while the physics-informed fine-tuning mechanism further reduces significant physical errors effectively. A MSE-R evaluation method is developed to assess the accuracy and realism of physical field generation.
Authors: Andrew Liu, Axel Elaldi, Nicholas T Franklin, Nathan Russell, Gurinder S Atwal, Yih-En A Ban, Olivia Viessmann
Abstract: Invariant Point Attention (IPA) is a key algorithm for geometry-aware modeling in structural biology, central to many protein and RNA models. However, its quadratic complexity limits the input sequence length. We introduce FlashIPA, a factorized reformulation of IPA that leverages hardware-efficient FlashAttention to achieve linear scaling in GPU memory and wall-clock time with sequence length. FlashIPA matches or exceeds standard IPA performance while substantially reducing computational costs. FlashIPA extends training to previously unattainable lengths, and we demonstrate this by re-training generative models without length restrictions and generating structures of thousands of residues. FlashIPA is available at https://github.com/flagshippioneering/flash_ipa.
Authors: Forsad Al Hossain, Tauhidur Rahman
Abstract: By replacing standard non-linearities with polynomial activations, Polynomial Neural Networks (PNNs) are pivotal for applications such as privacy-preserving inference via Homomorphic Encryption (HE). However, training PNNs effectively presents a significant challenge: low-degree polynomials can limit model expressivity, while higher-degree polynomials, crucial for capturing complex functions, often suffer from numerical instability and gradient explosion. We introduce a robust and versatile training framework featuring two synergistic innovations: 1) a novel Boundary Loss that exponentially penalizes activation inputs outside a predefined stable range, and 2) Selective Gradient Clipping that effectively tames gradient magnitudes while preserving essential Batch Normalization statistics. We demonstrate our framework's broad efficacy by training PNNs within deep architectures composed of HE-compatible layers (e.g., linear layers, average pooling, batch normalization, as used in ResNet variants) across diverse image, audio, and human activity recognition datasets. These models consistently achieve high accuracy with low-degree polynomial activations (such as degree 2) and, critically, exhibit stable training and strong performance with polynomial degrees up to 22, where standard methods typically fail or suffer severe degradation. Furthermore, the performance of these PNNs achieves a remarkable parity, closely approaching that of their original ReLU-based counterparts. Extensive ablation studies validate the contributions of our techniques and guide hyperparameter selection. We confirm the HE-compatibility of the trained models, advancing the practical deployment of accurate, stable, and secure deep learning inference.
Authors: Jintao Zhang, Jia Wei, Pengle Zhang, Xiaoming Xu, Haofeng Huang, Haoxu Wang, Kai Jiang, Jun Zhu, Jianfei Chen
Abstract: The efficiency of attention is important due to its quadratic time complexity. We enhance the efficiency of attention through two key contributions: First, we leverage the new FP4 Tensor Cores in Blackwell GPUs to accelerate attention computation. Our implementation achieves 1038 TOPS on RTX5090, which is a 5x speedup over the fastest FlashAttention on RTX5090. Experiments show that our FP4 attention can accelerate inference of various models in a plug-and-play way. Second, we pioneer low-bit attention to training tasks. Existing low-bit attention works like FlashAttention3 and SageAttention focus only on inference. However, the efficiency of training large models is also important. To explore whether low-bit attention can be effectively applied to training tasks, we design an accurate and efficient 8-bit attention for both forward and backward propagation. Experiments indicate that 8-bit attention achieves lossless performance in fine-tuning tasks but exhibits slower convergence in pretraining tasks. The code will be available at https://github.com/thu-ml/SageAttention.
Authors: Peter Chen, Xiaopeng Li, Ziniu Li, Xi Chen, Tianyi Lin
Abstract: Reinforcement learning (RL) has demonstrated significant success in enhancing reasoning capabilities in large language models (LLMs). One of the most widely used RL methods is Group Relative Policy Optimization (GRPO)~\cite{Shao-2024-Deepseekmath}, known for its memory efficiency and success in training DeepSeek-R1~\cite{Guo-2025-Deepseek}. However, GRPO stalls when all sampled responses in a group are incorrect -- referred to as an \emph{all-negative-sample} group -- as it fails to update the policy, hindering learning progress. The contributions of this paper are two-fold. First, we propose a simple yet effective framework that introduces response diversity within all-negative-sample groups in GRPO using AI feedback. We also provide a theoretical analysis, via a stylized model, showing how this diversification improves learning dynamics. Second, we empirically validate our approach, showing the improved performance across various model sizes (7B, 14B, 32B) in both offline and online learning settings with 10 benchmarks, including base and distilled variants. Our findings highlight that learning from all-negative-sample groups is not only feasible but beneficial, advancing recent insights from \citet{Xiong-2025-Minimalist}.
Authors: Rui Liu, Rui Xie, Zijun Yao, Yanjie Fu, Dongjie Wang
Abstract: Feature selection removes redundant features to enhanc performance and computational efficiency in downstream tasks. Existing works often struggle to capture complex feature interactions and adapt to diverse scenarios. Recent advances in this domain have incorporated generative intelligence to address these drawbacks by uncovering intricate relationships between features. However, two key limitations remain: 1) embedding feature subsets in a continuous space is challenging due to permutation sensitivity, as changes in feature order can introduce biases and weaken the embedding learning process; 2) gradient-based search in the embedding space assumes convexity, which is rarely guaranteed, leading to reduced search effectiveness and suboptimal subsets. To address these limitations, we propose a new framework that can: 1) preserve feature subset knowledge in a continuous embedding space while ensuring permutation invariance; 2) effectively explore the embedding space without relying on strong convex assumptions. For the first objective, we develop an encoder-decoder paradigm to preserve feature selection knowledge into a continuous embedding space. This paradigm captures feature interactions through pairwise relationships within the subset, removing the influence of feature order on the embedding. Moreover, an inducing point mechanism is introduced to accelerate pairwise relationship computations. For the second objective, we employ a policy-based reinforcement learning (RL) approach to guide the exploration of the embedding space. The RL agent effectively navigates the space by balancing multiple objectives. By prioritizing high-potential regions adaptively and eliminating the reliance on convexity assumptions, the RL agent effectively reduces the risk of converging to local optima. Extensive experiments demonstrate the effectiveness, efficiency, robustness and explicitness of our model.
Authors: Nikola Zubi\'c, Davide Scaramuzza
Abstract: Deep Selective State-Space Models (SSMs), characterized by input-dependent, time-varying parameters, offer significant expressive power but pose challenges for stability analysis, especially with discontinuous gating signals. In this paper, we investigate the stability and regularity properties of continuous-time selective SSMs through the lens of passivity and Input-to-State Stability (ISS). We establish that intrinsic energy dissipation guarantees exponential forgetting of past states. Crucially, we prove that the unforced system dynamics possess an underlying minimal quadratic energy function whose defining matrix exhibits robust $\text{AUC}_{\text{loc}}$ regularity, accommodating discontinuous gating. Furthermore, assuming a universal quadratic storage function ensures passivity across all inputs, we derive parametric LMI conditions and kernel constraints that limit gating mechanisms, formalizing "irreversible forgetting" of recurrent models. Finally, we provide sufficient conditions for global ISS, linking uniform local dissipativity to overall system robustness. Our findings offer a rigorous framework for understanding and designing stable and reliable deep selective SSMs.
Authors: Junhyung Park, Patrick Bloebaum, Shiva Prasad Kasiviswanathan
Abstract: Benign overfitting is a phenomenon in machine learning where a model perfectly fits (interpolates) the training data, including noisy examples, yet still generalizes well to unseen data. Understanding this phenomenon has attracted considerable attention in recent years. In this work, we introduce a conceptual shift, by focusing on almost benign overfitting, where models simultaneously achieve both arbitrarily small training and test errors. This behavior is characteristic of neural networks, which often achieve low (but non-zero) training error while still generalizing well. We hypothesize that this almost benign overfitting can emerge even in classical regimes, by analyzing how the interaction between sample size and model complexity enables larger models to achieve both good training fit but still approach Bayes-optimal generalization. We substantiate this hypothesis with theoretical evidence from two case studies: (i) kernel ridge regression, and (ii) least-squares regression using a two-layer fully connected ReLU neural network trained via gradient flow. In both cases, we overcome the strong assumptions often required in prior work on benign overfitting. Our results on neural networks also provide the first generalization result in this setting that does not rely on any assumptions about the underlying regression function or noise, beyond boundedness. Our analysis introduces a novel proof technique based on decomposing the excess risk into estimation and approximation errors, interpreting gradient flow as an implicit regularizer, that helps avoid uniform convergence traps. This analysis idea could be of independent interest.
Authors: Huiliang Zhang, Ping Nie, Lijun Sun, Benoit Boulet
Abstract: Multivariate time series (MTS) forecasting has a wide range of applications in both industry and academia. Recently, spatial-temporal graph neural networks (STGNNs) have gained popularity as MTS forecasting methods. However, current STGNNs can only use the finite length of MTS input data due to the computational complexity. Moreover, they lack the ability to identify similar patterns throughout the entire dataset and struggle with data that exhibit sparsely and discontinuously distributed correlations among variables over an extensive historical period, resulting in only marginal improvements. In this article, we introduce a simple yet effective k-nearest neighbor MTS forecasting ( kNN-MTS) framework, which forecasts with a nearest neighbor retrieval mechanism over a large datastore of cached series, using representations from the MTS model for similarity search. This approach requires no additional training and scales to give the MTS model direct access to the whole dataset at test time, resulting in a highly expressive model that consistently improves performance, and has the ability to extract sparse distributed but similar patterns spanning over multivariables from the entire dataset. Furthermore, a hybrid spatial-temporal encoder (HSTEncoder) is designed for kNN-MTS which can capture both long-term temporal and short-term spatial-temporal dependencies and is shown to provide accurate representation for kNN-MTSfor better forecasting. Experimental results on several real-world datasets show a significant improvement in the forecasting performance of kNN-MTS. The quantitative analysis also illustrates the interpretability and efficiency of kNN-MTS, showing better application prospects and opening up a new path for efficiently using the large dataset in MTS models.
Authors: Shuyi Chen, Shixiang Zhu, Ramteen Sioshansi
Abstract: Extreme weather events are placing growing strain on electric power systems, exposing the limitations of purely reactive responses and prompting the need for proactive resilience planning. However, existing approaches often rely on simplified uncertainty models and decouple proactive and reactive decisions, overlooking their critical interdependence. This paper proposes a novel tri-level optimization framework that integrates proactive infrastructure investment, adversarial modeling of spatio-temporal disruptions, and adaptive reactive response. We construct high-probability, distribution-free uncertainty sets using conformal prediction to capture complex and data-scarce outage patterns. To solve the resulting nested decision problem, we derive a bi-level reformulation via strong duality and develop a scalable Benders decomposition algorithm. Experiments on both real and synthetic data demonstrate that our approach consistently outperforms conventional robust and two-stage methods, achieving lower worst-case losses and more efficient resource allocation, especially under tight operational constraints and large-scale uncertainty.
Authors: Wajdi Hammami, Soumaya Cherkaoui, Shengrui Wang
Abstract: Quantum computing may offer new approaches for advancing machine learning, including in complex tasks such as anomaly detection in network traffic. In this paper, we introduce a quantum generative adversarial network (QGAN) architecture for multivariate time-series anomaly detection that leverages variational quantum circuits (VQCs) in combination with a time-window shifting technique, data re-uploading, and successive data injection (SuDaI). The method encodes multivariate time series data as rotation angles. By integrating both data re-uploading and SuDaI, the approach maps classical data into quantum states efficiently, helping to address hardware limitations such as the restricted number of available qubits. In addition, the approach employs an anomaly scoring technique that utilizes both the generator and the discriminator output to enhance the accuracy of anomaly detection. The QGAN was trained using the parameter shift rule and benchmarked against a classical GAN. Experimental results indicate that the quantum model achieves a accuracy high along with high recall and F1-scores in anomaly detection, and attains a lower MSE compared to the classical model. Notably, the QGAN accomplishes this performance with only 80 parameters, demonstrating competitive results with a compact architecture. Tests using a noisy simulator suggest that the approach remains effective under realistic noise-prone conditions.
Authors: Nikhil Kapasi, William Whitehead, Luke Theogarajan
Abstract: Many real-world tasks, from associative memory to symbolic reasoning, demand discrete, structured representations that standard continuous latent models struggle to express naturally. We introduce the Gaussian-Multinoulli Restricted Boltzmann Machine (GM-RBM), a generative energy-based model that extends the Gaussian-Bernoulli RBM (GB-RBM) by replacing binary hidden units with $q$-state Potts variables. This modification enables a combinatorially richer latent space and supports learning over multivalued, interpretable latent concepts. We formally derive GM-RBM's energy function, learning dynamics, and conditional distributions, showing that it preserves tractable inference and training through contrastive divergence. Empirically, we demonstrate that GM-RBMs model complex multimodal distributions more effectively than binary RBMs, outperforming them on tasks involving analogical recall and structured memory. Our results highlight GM-RBMs as a scalable framework for discrete latent inference with enhanced expressiveness and interoperability.
Authors: Hongyu Cheng, Amitabh Basu
Abstract: Mixed-integer programming (MIP) provides a powerful framework for optimization problems, with Branch-and-Cut (B&C) being the predominant algorithm in state-of-the-art solvers. The efficiency of B&C critically depends on heuristic policies for making sequential decisions, including node selection, cut selection, and branching variable selection. While traditional solvers often employ heuristics with manually tuned parameters, recent approaches increasingly leverage machine learning, especially neural networks, to learn these policies directly from data. A key challenge is to understand the theoretical underpinnings of these learned policies, particularly their generalization performance from finite data. This paper establishes rigorous sample complexity bounds for learning B&C policies where the scoring functions guiding each decision step (node, cut, branch) have a certain piecewise polynomial structure. This structure generalizes the linear models that form the most commonly deployed policies in practice and investigated recently in a foundational series of theoretical works by Balcan et al. Such piecewise polynomial policies also cover the neural network architectures (e.g., using ReLU activations) that have been the focal point of contemporary practical studies. Consequently, our theoretical framework closely reflects the models utilized by practitioners investigating machine learning within B&C, offering a unifying perspective relevant to both established theory and modern empirical research in this area. Furthermore, our theory applies to quite general sequential decision making problems beyond B&C.
Authors: Jinzhou Cao, Xiangxu Wang, Jiashi Chen, Wei Tu, Zhenhui Li, Xindong Yang, Tianhong Zhao, Qingquan Li
Abstract: Fine-grained economic mapping through urban representation learning has emerged as a crucial tool for evidence-based economic decisions. While existing methods primarily rely on supervised or unsupervised approaches, they often overlook semi-supervised learning in data-scarce scenarios and lack unified multi-task frameworks for comprehensive sectoral economic analysis. To address these gaps, we propose SemiGTX, an explainable semi-supervised graph learning framework for sectoral economic mapping. The framework is designed with dedicated fusion encoding modules for various geospatial data modalities, seamlessly integrating them into a cohesive graph structure. It introduces a semi-information loss function that combines spatial self-supervision with locally masked supervised regression, enabling more informative and effective region representations. Through multi-task learning, SemiGTX concurrently maps GDP across primary, secondary, and tertiary sectors within a unified model. Extensive experiments conducted in the Pearl River Delta region of China demonstrate the model's superior performance compared to existing methods, achieving R2 scores of 0.93, 0.96, and 0.94 for the primary, secondary and tertiary sectors, respectively. Cross-regional experiments in Beijing and Chengdu further illustrate its generality. Systematic analysis reveals how different data modalities influence model predictions, enhancing explainability while providing valuable insights for regional development planning. This representation learning framework advances regional economic monitoring through diverse urban data integration, providing a robust foundation for precise economic forecasting.
Authors: Tsutahiro Fukuhara, Junya Hara, Hiroshi Higashi, Yuichi Tanaka
Abstract: We propose a robust aggregation method for model parameters in federated learning (FL) under noisy communications. FL is a distributed machine learning paradigm in which a central server aggregates local model parameters from multiple clients. These parameters are often noisy and/or have missing values during data collection, training, and communication between the clients and server. This may cause a considerable drop in model accuracy. To address this issue, we learn a graph that represents pairwise relationships between model parameters of the clients during aggregation. We realize it with a joint problem of graph learning and signal (i.e., model parameters) restoration. The problem is formulated as a difference-of-convex (DC) optimization, which is efficiently solved via a proximal DC algorithm. Experimental results on MNIST and CIFAR-10 datasets show that the proposed method outperforms existing approaches by up to $2$--$5\%$ in classification accuracy under biased data distributions and noisy conditions.
Authors: Yuhang Liu, Yingxue Zhang, Xin Zhang, Ling Tian, Xu Zheng, Yanhua Li, Jun Luo
Abstract: Understanding and predicting urban dynamics is crucial for managing transportation systems, optimizing urban planning, and enhancing public services. While neural network-based approaches have achieved success, they often rely on task-specific architectures and large volumes of data, limiting their ability to generalize across diverse urban scenarios. Meanwhile, Large Language Models (LLMs) offer strong reasoning and generalization capabilities, yet their application to spatial-temporal urban dynamics remains underexplored. Existing LLM-based methods struggle to effectively integrate multifaceted spatial-temporal data and fail to address distributional shifts between training and testing data, limiting their predictive reliability in real-world applications. To bridge this gap, we propose UrbanMind, a novel spatial-temporal LLM framework for multifaceted urban dynamics prediction that ensures both accurate forecasting and robust generalization. At its core, UrbanMind introduces Muffin-MAE, a multifaceted fusion masked autoencoder with specialized masking strategies that capture intricate spatial-temporal dependencies and intercorrelations among multifaceted urban dynamics. Additionally, we design a semantic-aware prompting and fine-tuning strategy that encodes spatial-temporal contextual details into prompts, enhancing LLMs' ability to reason over spatial-temporal patterns. To further improve generalization, we introduce a test time adaptation mechanism with a test data reconstructor, enabling UrbanMind to dynamically adjust to unseen test data by reconstructing LLM-generated embeddings. Extensive experiments on real-world urban datasets across multiple cities demonstrate that UrbanMind consistently outperforms state-of-the-art baselines, achieving high accuracy and robust generalization, even in zero-shot settings.
Authors: Ziqing Xu, Hancheng Min, Salma Tarmoun, Enrique Mallada, Rene Vidal
Abstract: Most prior work on the convergence of gradient descent (GD) for overparameterized neural networks relies on strong assumptions on the step size (infinitesimal), the hidden-layer width (infinite), or the initialization (large, spectral, balanced). Recent efforts to relax these assumptions focus on two-layer linear networks trained with the squared loss. In this work, we derive a linear convergence rate for training two-layer linear neural networks with GD for general losses and under relaxed assumptions on the step size, width, and initialization. A key challenge in deriving this result is that classical ingredients for deriving convergence rates for nonconvex problems, such as the Polyak-{\L}ojasiewicz (PL) condition and Descent Lemma, do not hold globally for overparameterized neural networks. Here, we prove that these two conditions hold locally with local constants that depend on the weights. Then, we provide bounds on these local constants, which depend on the initialization of the weights, the current loss, and the global PL and smoothness constants of the non-overparameterized model. Based on these bounds, we derive a linear convergence rate for GD. Our convergence analysis not only improves upon prior results but also suggests a better choice for the step size, as verified through our numerical experiments.
Authors: Yiming Zhang, Sitong Liu, Alex Cloninger
Abstract: We address the computational and theoretical limitations of existing distributional alignment methods for unsupervised domain adaptation (UDA), particularly regarding the estimation of classification performance and confidence without target labels. Current theoretical frameworks for these methods often yield computationally intractable quantities and fail to adequately reflect the properties of the alignment algorithms employed. To overcome these challenges, we introduce the Optimal Transport (OT) score, a confidence metric derived from a novel theoretical analysis that exploits the flexibility of decision boundaries induced by Semi-Discrete Optimal Transport alignment. The proposed OT score is intuitively interpretable, theoretically rigorous, and computationally efficient. It provides principled uncertainty estimates for any given set of target pseudo-labels without requiring model retraining, and can flexibly adapt to varying degrees of available source information. Experimental results on standard UDA benchmarks demonstrate that classification accuracy consistently improves by identifying and removing low-confidence predictions, and that OT score significantly outperforms existing confidence metrics across diverse adaptation scenarios.
Authors: Ananyae Kumar Bhartari, Vinayak Vinayak, Vivek B Shenoy
Abstract: Parameter estimation in inverse problems involving partial differential equations (PDEs) underpins modeling across scientific disciplines, especially when parameters vary in space or time. Physics-informed Machine Learning (PhiML) integrates PDE constraints into deep learning, but prevailing approaches depend on recursive automatic differentiation (autodiff), which produces inaccurate high-order derivatives, inflates memory usage, and underperforms in noisy settings. We propose Mollifier Layers, a lightweight, architecture-agnostic module that replaces autodiff with convolutional operations using analytically defined mollifiers. This reframing of derivative computation as smoothing integration enables efficient, noise-robust estimation of high-order derivatives directly from network outputs. Mollifier Layers attach at the output layer and require no architectural modifications. We compare them with three distinct architectures and benchmark performance across first-, second-, and fourth-order PDEs -- including Langevin dynamics, heat diffusion, and reaction-diffusion systems -- observing significant improvements in memory efficiency, training time and accuracy for parameter recovery across tasks. To demonstrate practical relevance, we apply Mollifier Layers to infer spatially varying epigenetic reaction rates from super-resolution chromatin imaging data -- a real-world inverse problem with biomedical significance. Our results establish Mollifier Layers as an efficient and scalable tool for physics-constrained learning.
Authors: Sahil Rajesh Dhayalkar
Abstract: We develop a novel theoretical framework for analyzing ReLU neural networks through the lens of a combinatorial object we term the ReLU Transition Graph (RTG). In this graph, each node corresponds to a linear region induced by the network's activation patterns, and edges connect regions that differ by a single neuron flip. Building on this structure, we derive a suite of new theoretical results connecting RTG geometry to expressivity, generalization, and robustness. Our contributions include tight combinatorial bounds on RTG size and diameter, a proof of RTG connectivity, and graph-theoretic interpretations of VC-dimension. We also relate entropy and average degree of the RTG to generalization error. Each theoretical result is rigorously validated via carefully controlled experiments across varied network depths, widths, and data regimes. This work provides the first unified treatment of ReLU network structure via graph theory and opens new avenues for compression, regularization, and complexity control rooted in RTG analysis.
Authors: Sahil Rajesh Dhayalkar
Abstract: We present a complete theoretical and empirical framework establishing feedforward neural networks as universal finite-state machines (N-FSMs). Our results prove that finite-depth ReLU and threshold networks can exactly simulate deterministic finite automata (DFAs) by unrolling state transitions into depth-wise neural layers, with formal characterizations of required depth, width, and state compression. We demonstrate that DFA transitions are linearly separable, binary threshold activations allow exponential compression, and Myhill-Nerode equivalence classes can be embedded into continuous latent spaces while preserving separability. We also formalize the expressivity boundary: fixed-depth feedforward networks cannot recognize non-regular languages requiring unbounded memory. Unlike prior heuristic or probing-based studies, we provide constructive proofs and design explicit DFA-unrolled neural architectures that empirically validate every claim. Our results bridge deep learning, automata theory, and neural-symbolic computation, offering a rigorous blueprint for how discrete symbolic processes can be realized in continuous neural systems.
Authors: Shihao Zhang, Haoyu Zhang, Ian Colbert, Rayan Saab
Abstract: We introduce Qronos -- a new state-of-the-art post-training quantization algorithm that sequentially rounds and updates neural network weights. Qronos not only explicitly corrects errors due to both weight and activation quantization, but also errors resulting from quantizing previous layers. Our iterative algorithm is based on an interpretable and disciplined optimization framework that subsumes and surpasses existing data-driven approaches. At each step, Qronos alternates between error correction and diffusion via optimal update rules. Importantly, we prove that Qronos admits an efficient implementation that uses the Cholesky decomposition for solving least-squares problems. We also demonstrate that Qronos is compatible with existing transformation techniques such as Hadamard-based incoherence processing and weight-activation scaling equalization, among others. We evaluate Qronos using recent autoregressive language generation models in the Llama3 family; Qronos consistently outperforms previous state-of-the-art adaptive rounding methods when quantizing the weights, activations, and/or KV caches.
Authors: Keenan Eikenberry, Lizuo Liu, Yoonsang Lee
Abstract: This work investigates the use of Wasserstein correlation -- a normalized measure of statistical dependence based on the Wasserstein distance between a joint distribution and the product of its marginals -- for unsupervised representation learning. Unlike, for example, contrastive methods, which naturally cluster classes in the latent space, we find that an (auto)encoder trained to maximize Wasserstein correlation between the input and encoded distributions instead acts as a compressor, reducing dimensionality while approximately preserving the topological and geometric properties of the input distribution. More strikingly, we show that Wasserstein correlation maximization can be used to arrive at an (auto)encoder -- either trained from scratch, or else one that extends a frozen, pretrained model -- that is approximately invariant to a chosen augmentation, or collection of augmentations, and that still approximately preserves the structural properties of the non-augmented input distribution. To do this, we first define the notion of an augmented encoder using the machinery of Markov-Wasserstein kernels. When the maximization objective is then applied to the augmented encoder, as opposed to the underlying, deterministic encoder, the resulting model exhibits the desired invariance properties. Finally, besides our experimental results, which show that even simple feedforward networks can be imbued with invariants or can, alternatively, be used to impart invariants to pretrained models under this training process, we additionally establish various theoretical results for optimal transport-based dependence measures. Code is available at https://github.com/keenan-eikenberry/wasserstein_correlation_maximization .
URLs: https://github.com/keenan-eikenberry/wasserstein_correlation_maximization
Authors: Sagnik Mukherjee, Lifan Yuan, Dilek Hakkani-Tur, Hao Peng
Abstract: Reinforcement learning (RL) yields substantial improvements in large language models (LLMs) downstream task performance and alignment with human values. Surprisingly, such large gains result from updating only a small subnetwork comprising just 5 percent to 30 percent of the parameters, with the rest effectively unchanged. We refer to this phenomenon as parameter update sparsity induced by RL. It is observed across all 7 widely used RL algorithms (e.g., PPO, GRPO, DPO) and all 10 LLMs from different families in our experiments. This sparsity is intrinsic and occurs without any explicit sparsity promoting regularizations or architectural constraints. Finetuning the subnetwork alone recovers the test accuracy, and, remarkably, produces a model nearly identical to the one obtained via full finetuning. The subnetworks from different random seeds, training data, and even RL algorithms show substantially greater overlap than expected by chance. Our analysis suggests that this sparsity is not due to updating only a subset of layers, instead, nearly all parameter matrices receive similarly sparse updates. Moreover, the updates to almost all parameter matrices are nearly full-rank, suggesting RL updates a small subset of parameters that nevertheless span almost the full subspaces that the parameter matrices can represent. We conjecture that the this update sparsity can be primarily attributed to training on data that is near the policy distribution, techniques that encourage the policy to remain close to the pretrained model, such as the KL regularization and gradient clipping, have limited impact.
Authors: Arjun Prakash, Naicheng He, Denizalp Goktas, Amy Greenwald
Abstract: The dependency of the actor on the critic in actor-critic (AC) reinforcement learning means that AC can be characterized as a bilevel optimization (BLO) problem, also called a Stackelberg game. This characterization motivates two modifications to vanilla AC algorithms. First, the critic's update should be nested to learn a best response to the actor's policy. Second, the actor should update according to a hypergradient that takes changes in the critic's behavior into account. Computing this hypergradient involves finding an inverse Hessian vector product, a process that can be numerically unstable. We thus propose a new algorithm, Bilevel Policy Optimization with Nystr\"om Hypergradients (BLPO), which uses nesting to account for the nested structure of BLO, and leverages the Nystr\"om method to compute the hypergradient. Theoretically, we prove BLPO converges to (a point that satisfies the necessary conditions for) a local strong Stackelberg equilibrium in polynomial time with high probability, assuming a linear parametrization of the critic's objective. Empirically, we demonstrate that BLPO performs on par with or better than PPO on a variety of discrete and continuous control tasks.
Authors: Xilong Wang, John Bloch, Zedian Shao, Yuepeng Hu, Shuyan Zhou, Neil Zhenqiang Gong
Abstract: Multi-modal large language model (MLLM)-based web agents interact with webpage environments by generating actions based on screenshots of the webpages. Environmental prompt injection attacks manipulate the environment to induce the web agent to perform a specific, attacker-chosen action--referred to as the target action. However, existing attacks suffer from limited effectiveness or stealthiness, or are impractical in real-world settings. In this work, we propose EnvInjection, a new attack that addresses these limitations. Our attack adds a perturbation to the raw pixel values of the rendered webpage, which can be implemented by modifying the webpage's source code. After these perturbed pixels are mapped into a screenshot, the perturbation induces the web agent to perform the target action. We formulate the task of finding the perturbation as an optimization problem. A key challenge in solving this problem is that the mapping between raw pixel values and screenshot is non-differentiable, making it difficult to backpropagate gradients to the perturbation. To overcome this, we train a neural network to approximate the mapping and apply projected gradient descent to solve the reformulated optimization problem. Extensive evaluation on multiple webpage datasets shows that EnvInjection is highly effective and significantly outperforms existing baselines.
Authors: Imon Banerjee, Sayak Chakrabarty
Abstract: The m-out-of-n bootstrap, originally proposed by Bickel, Gotze, and Zwet (1992), approximates the distribution of a statistic by repeatedly drawing m subsamples (with m much smaller than n) without replacement from an original sample of size n. It is now routinely used for robust inference with heavy-tailed data, bandwidth selection, and other large-sample applications. Despite its broad applicability across econometrics, biostatistics, and machine learning, rigorous parameter-free guarantees for the soundness of the m-out-of-n bootstrap when estimating sample quantiles have remained elusive. This paper establishes such guarantees by analyzing the estimator of sample quantiles obtained from m-out-of-n resampling of a dataset of size n. We first prove a central limit theorem for a fully data-driven version of the estimator that holds under a mild moment condition and involves no unknown nuisance parameters. We then show that the moment assumption is essentially tight by constructing a counter-example in which the CLT fails. Strengthening the assumptions slightly, we derive an Edgeworth expansion that provides exact convergence rates and, as a corollary, a Berry Esseen bound on the bootstrap approximation error. Finally, we illustrate the scope of our results by deriving parameter-free asymptotic distributions for practical statistics, including the quantiles for random walk Metropolis-Hastings and the rewards of ergodic Markov decision processes, thereby demonstrating the usefulness of our theory in modern estimation and learning tasks.
Authors: Harshil Vejendla, Haizhou Shi, Yibin Wang, Tunyu Zhang, Huan Zhang, Hao Wang
Abstract: Recent advances in uncertainty estimation for Large Language Models (LLMs) during downstream adaptation have addressed key challenges of reliability and simplicity. However, existing Bayesian methods typically require multiple sampling iterations during inference, creating significant efficiency issues that limit practical deployment. In this paper, we investigate the possibility of eliminating the need for test-time sampling for LLM uncertainty estimation. Specifically, when given an off-the-shelf Bayesian LLM, we distill its aligned confidence into a non-Bayesian student LLM by minimizing the divergence between their predictive distributions. Unlike typical calibration methods, our distillation is carried out solely on the training dataset without the need of an additional validation dataset. This simple yet effective approach achieves N-times more efficient uncertainty estimation during testing, where N is the number of samples traditionally required by Bayesian LLMs. Our extensive experiments demonstrate that uncertainty estimation capabilities on training data can successfully generalize to unseen test data through our distillation technique, consistently producing results comparable to (or even better than) state-of-the-art Bayesian LLMs.
Authors: Tunyu Zhang, Haizhou Shi, Yibin Wang, Hengyi Wang, Xiaoxiao He, Zhuowei Li, Haoxian Chen, Ligong Han, Kai Xu, Huan Zhang, Dimitris Metaxas, Hao Wang
Abstract: While Large Language Models (LLMs) have demonstrated impressive capabilities, their output quality remains inconsistent across various application scenarios, making it difficult to identify trustworthy responses, especially in complex tasks requiring multi-step reasoning. In this paper, we propose a token-level uncertainty estimation framework to enable LLMs to self-assess and self-improve their generation quality in mathematical reasoning. Specifically, we introduce low-rank random weight perturbation to LLM decoding, generating predictive distributions that we use to estimate token-level uncertainties. We then aggregate these uncertainties to reflect semantic uncertainty of the generated sequences. Experiments on mathematical reasoning datasets of varying difficulty demonstrate that our token-level uncertainty metrics strongly correlate with answer correctness and model robustness. Additionally, we explore using uncertainty to directly enhance the model's reasoning performance through multiple generations and the particle filtering algorithm. Our approach consistently outperforms existing uncertainty estimation methods, establishing effective uncertainty estimation as a valuable tool for both evaluating and improving reasoning generation in LLMs.
Authors: Alberto Sinigaglia, Davide Sartor, Marina Ceccon, Gian Antonio Susto
Abstract: Fair classification is a critical challenge that has gained increasing importance due to international regulations and its growing use in high-stakes decision-making settings. Existing methods often rely on adversarial learning or distribution matching across sensitive groups; however, adversarial learning can be unstable, and distribution matching can be computationally intensive. To address these limitations, we propose a novel approach based on the characteristic function distance. Our method ensures that the learned representation contains minimal sensitive information while maintaining high effectiveness for downstream tasks. By utilizing characteristic functions, we achieve a more stable and efficient solution compared to traditional methods. Additionally, we introduce a simple relaxation of the objective function that guarantees fairness in common classification models with no performance degradation. Experimental results on benchmark datasets demonstrate that our approach consistently matches or achieves better fairness and predictive accuracy than existing methods. Moreover, our method maintains robustness and computational efficiency, making it a practical solution for real-world applications.
Authors: Joshua Inman, Tanmay Khandait, Lalitha Sankar, Giulia Pedrielli
Abstract: In this paper we propose for the first time the hyperparameter optimization (HPO) algorithm POCAII. POCAII differs from the Hyperband and Successive Halving literature by explicitly separating the search and evaluation phases and utilizing principled approaches to exploration and exploitation principles during both phases. Such distinction results in a highly flexible scheme for managing a hyperparameter optimization budget by focusing on search (i.e., generating competing configurations) towards the start of the HPO process while increasing the evaluation effort as the HPO comes to an end. POCAII was compared to state of the art approaches SMAC, BOHB and DEHB. Our algorithm shows superior performance in low-budget hyperparameter optimization regimes. Since many practitioners do not have exhaustive resources to assign to HPO, it has wide applications to real-world problems. Moreover, the empirical evidence showed how POCAII demonstrates higher robustness and lower variance in the results. This is again very important when considering realistic scenarios with extremely expensive models to train.
Authors: Wei Zhang, Arif Hassan Zidan, Afrar Jahin, Yu Bao, Tianming Liu
Abstract: Momentum-based gradients are essential for optimizing advanced machine learning models, as they not only accelerate convergence but also advance optimizers to escape stationary points. While most state-of-the-art momentum techniques utilize lower-order gradients, such as the squared first-order gradient, there has been limited exploration of higher-order gradients, particularly those raised to powers greater than two. In this work, we introduce the concept of high-order momentum, where momentum is constructed using higher-power gradients, with a focus on the third-power of the first-order gradient as a representative case. Our research offers both theoretical and empirical support for this approach. Theoretically, we demonstrate that incorporating third-power gradients can improve the convergence bounds of gradient-based optimizers for both convex and smooth nonconvex problems. Empirically, we validate these findings through extensive experiments across convex, smooth nonconvex, and nonsmooth nonconvex optimization tasks. Across all cases, high-order momentum consistently outperforms conventional low-order momentum methods, showcasing superior performance in various optimization problems.
Authors: Wei Zhang, Arif Hassan Zidan, Afrar Jahin, Yu Bao, Tianming Liu
Abstract: While gradient-based optimizers that incorporate randomization often showcase superior performance on complex optimization, the theoretical foundations underlying this superiority remain insufficiently understood. A particularly pressing question has emerged: What is the role of randomization in dimension-free nonsmooth nonconvex optimization? To address this gap, we investigate the theoretical and empirical impact of permutation randomization within gradient-based optimization frameworks, using it as a representative case to explore broader implications. From a theoretical perspective, our analyses reveal that permutation randomization disrupts the shrinkage behavior of gradient-based optimizers, facilitating continuous convergence toward the global optimum given a sufficiently large number of iterations. Additionally, we prove that permutation randomization can preserve the convergence rate of the underlying optimizer. On the empirical side, we conduct extensive numerical experiments comparing permutation-randomized optimizer against three baseline methods. These experiments span tasks such as training deep neural networks with stacked architectures and optimizing noisy objective functions. The results not only corroborate our theoretical insights but also highlight the practical benefits of permutation randomization. In summary, this work delivers both rigorous theoretical justification and compelling empirical evidence for the effectiveness of permutation randomization. Our findings and evidence lay a foundation for extending analytics to encompass a wide array of randomization.
Authors: David Chanin, Tom\'a\v{s} Dulka, Adri\`a Garriga-Alonso
Abstract: It is assumed that sparse autoencoders (SAEs) decompose polysemantic activations into interpretable linear directions, as long as the activations are composed of sparse linear combinations of underlying features. However, we find that if an SAE is more narrow than the number of underlying "true features" on which it is trained, and there is correlation between features, the SAE will merge components of correlated features together, thus destroying monosemanticity. In LLM SAEs, these two conditions are almost certainly true. This phenomenon, which we call feature hedging, is caused by SAE reconstruction loss, and is more severe the narrower the SAE. In this work, we introduce the problem of feature hedging and study it both theoretically in toy models and empirically in SAEs trained on LLMs. We suspect that feature hedging may be one of the core reasons that SAEs consistently underperform supervised baselines. Finally, we use our understanding of feature hedging to propose an improved variant of matryoshka SAEs. Our work shows there remain fundamental issues with SAEs, but we are hopeful that that highlighting feature hedging will catalyze future advances that allow SAEs to achieve their full potential of interpreting LLMs at scale.
Authors: Mansi Sakarvadia, Nathaniel Hudson, Tian Li, Ian Foster, Kyle Chard
Abstract: Decentralized learning enables collaborative training of models across naturally distributed data without centralized coordination or maintenance of a global model. Instead, devices are organized in arbitrary communication topologies, in which they can only communicate with neighboring devices. Each device maintains its own local model by training on its local data and integrating new knowledge via model aggregation with neighbors. Therefore, knowledge is propagated across the topology via successive aggregation rounds. We study, in particular, the propagation of out-of-distribution (OOD) knowledge. We find that popular decentralized learning algorithms struggle to propagate OOD knowledge effectively to all devices. Further, we find that both the location of OOD data within a topology, and the topology itself, significantly impact OOD knowledge propagation. We then propose topology-aware aggregation strategies to accelerate (OOD) knowledge propagation across devices. These strategies improve OOD data accuracy, compared to topology-unaware baselines, by 123% on average across models in a topology.
Authors: Haoze Song, Zhihao Li, Xiaobo Zhang, Zecheng Gan, Zhilu Lai, Wei Wang
Abstract: Neural Operators have emerged as powerful tools for learning mappings between function spaces. Among them, the kernel integral operator has been widely validated on universally approximating various operators. Although recent advancements following this definition have developed effective modules to better approximate the kernel function defined on the original domain (with $d$ dimensions, $d=1, 2, 3...$), the unclarified evolving mechanism in the embedding spaces blocks our view to design neural operators that can fully capture the target system evolution. Drawing on recent breakthroughs in quantum simulation of partial differential equations (PDEs), we elucidate the linear evolution process in neural operators. Based on that, we redefine neural operators on a new $d+1$ dimensional domain. Within this framework, we implement our proposed Schr\"odingerised Kernel Neural Operator (SKNO) aligning better with the $d+1$ dimensional evolution. In experiments, our $d+1$ dimensional evolving linear block performs far better than others. Also, we test SKNO's SOTA performance on various benchmark tests and also the zero-shot super-resolution task. In addition, we analyse the impact of different lifting and recovering operators on the prediction within the redefined NO framework, reflecting the alignment between our model and the underlying $d+1$ dimensional evolution.
Authors: Jing Huang, Junyi Tao, Thomas Icard, Diyi Yang, Christopher Potts
Abstract: Interpretability research now offers a variety of techniques for identifying abstract internal mechanisms in neural networks. Can such techniques be used to predict how models will behave on out-of-distribution examples? In this work, we provide a positive answer to this question. Through a diverse set of language modeling tasks--including symbol manipulation, knowledge retrieval, and instruction following--we show that the most robust features for correctness prediction are those that play a distinctive causal role in the model's behavior. Specifically, we propose two methods that leverage causal mechanisms to predict the correctness of model outputs: counterfactual simulation (checking whether key causal variables are realized) and value probing (using the values of those variables to make predictions). Both achieve high AUC-ROC in distribution and outperform methods that rely on causal-agnostic features in out-of-distribution settings, where predicting model behaviors is more crucial. Our work thus highlights a novel and significant application for internal causal analysis of language models.
Authors: Yichen Xu, Ryumei Nakada, Linjun Zhang, Lexin Li
Abstract: Transfer learning typically leverages representations learned from a source domain to improve performance on a target task. A common approach is to extract features from a pre-trained model and directly apply them for target prediction. However, this strategy is prone to negative transfer where the source representation fails to align with the target distribution. In this article, we propose Residual Feature Integration (REFINE), a simple yet effective method designed to mitigate negative transfer. Our approach combines a fixed source-side representation with a trainable target-side encoder and fits a shallow neural network on the resulting joint representation, which adapts to the target domain while preserving transferable knowledge from the source domain. Theoretically, we prove that REFINE is sufficient to prevent negative transfer under mild conditions, and derive the generalization bound demonstrating its theoretical benefit. Empirically, we show that REFINE consistently enhances performance across diverse application and data modalities including vision, text, and tabular data, and outperforms numerous alternative solutions. Our method is lightweight, architecture-agnostic, and robust, making it a valuable addition to the existing transfer learning toolbox.
Authors: Ryan Chen, Youngmin Ko, Zeyu Zhang, Catherine Cho, Sunny Chung, Mauro Giuffr\'e, Dennis L. Shung, Bradly C. Stadie
Abstract: We introduce \textbf{LAMP} (\textbf{L}inear \textbf{A}ttribution \textbf{M}apping \textbf{P}robe), a method that shines light onto a black-box language model's decision surface and studies how reliably a model maps its stated reasons to its predictions through a locally linear model approximating the decision surface. LAMP treats the model's own self-reported explanations as a coordinate system and fits a locally linear surrogate that links those weights to the model's output. By doing so, it reveals which stated factors steer the model's decisions, and by how much. We apply LAMP to three tasks: \textit{sentiment analysis}, \textit{controversial-topic detection}, and \textit{safety-prompt auditing}. Across these tasks, LAMP reveals that many LLMs exhibit locally linear decision landscapes. In addition, these surfaces correlate with human judgments on explanation quality and, on a clinical case-file data set, aligns with expert assessments. Since LAMP operates without requiring access to model gradients, logits, or internal activations, it serves as a practical and lightweight framework for auditing proprietary language models, and enabling assessment of whether a model behaves consistently with the explanations it provides.
Authors: James V. Roggeveen, Erik Y. Wang, Will Flintoft, Peter Donets, Lucy S. Nathwani, Nickholas Gutierrez, David Ettel, Anton Marius Graf, Siddharth Dandavate, Arjun Nageswaran, Raglan Ward, Ava Williamson, Anne Mykland, Kacper K. Migacz, Yijun Wang, Egemen Bostan, Duy Thuc Nguyen, Zhe He, Marc L. Descoteaux, Felix Yeung, Shida Liu, Jorge Garc\'ia Ponce, Luke Zhu, Yuyang Chen, Ekaterina S. Ivshina, Miguel Fernandez, Minjae Kim, Kennan Gumbs, Matthew Scott Tan, Russell Yang, Mai Hoang, David Brown, Isabella A. Silveira, Lavon Sykes, Ahmed Roman, William Fredenberg, Yiming Chen, Lucas Martin, Yixing Tang, Kelly Werker Smith, Hongyu Liao, Logan G. Wilson, Alexander Dazhen Cai, Andrea Elizabeth Biju, Michael P. Brenner
Abstract: Large language models (LLMs) have shown remarkable progress in mathematical problem-solving, but evaluation has largely focused on problems that have exact analytical solutions or involve formal proofs, often overlooking approximation-based problems ubiquitous in applied science and engineering. To fill this gap, we build on prior work and present HARDMath2, a dataset of 211 original problems covering the core topics in an introductory graduate applied math class, including boundary-layer analysis, WKB methods, asymptotic solutions of nonlinear partial differential equations, and the asymptotics of oscillatory integrals. This dataset was designed and verified by the students and instructors of a core graduate applied mathematics course at Harvard. We build the dataset through a novel collaborative environment that challenges students to write and refine difficult problems consistent with the class syllabus, peer-validate solutions, test different models, and automatically check LLM-generated solutions against their own answers and numerical ground truths. Evaluation results show that leading frontier models still struggle with many of the problems in the dataset, highlighting a gap in the mathematical reasoning skills of current LLMs. Importantly, students identified strategies to create increasingly difficult problems by interacting with the models and exploiting common failure modes. This back-and-forth with the models not only resulted in a richer and more challenging benchmark but also led to qualitative improvements in the students' understanding of the course material, which is increasingly important as we enter an age where state-of-the-art language models can solve many challenging problems across a wide domain of fields.
Authors: Jiali Chen, Avijit Mukherjee
Abstract: Self-supervised learning (SSL) on graphs generates node and graph representations (i.e., embeddings) that can be used for downstream tasks such as node classification, node clustering, and link prediction. Graph SSL is particularly useful in scenarios with limited or no labeled data. Existing SSL methods predominantly follow contrastive or generative paradigms, each excelling in different tasks: contrastive methods typically perform well on classification tasks, while generative methods often excel in link prediction. In this paper, we present a novel architecture for graph SSL that integrates the strengths of both approaches. Our framework introduces community-aware node-level contrastive learning, providing more robust and effective positive and negative node pairs generation, alongside graph-level contrastive learning to capture global semantic information. Additionally, we employ a comprehensive augmentation strategy that combines feature masking, node perturbation, and edge perturbation, enabling robust and diverse representation learning. By incorporating these enhancements, our model achieves superior performance across multiple tasks, including node classification, clustering, and link prediction. Evaluations on open benchmark datasets demonstrate that our model outperforms state-of-the-art methods, achieving a performance lift of 0.23%-2.01% depending on the task and dataset.
Authors: Ziyu Zhou, Jiaxi Hu, Qingsong Wen, James T. Kwok, Yuxuan Liang
Abstract: In deep time series forecasting, the Fourier Transform (FT) is extensively employed for frequency representation learning. However, it often struggles in capturing multi-scale, time-sensitive patterns. Although the Wavelet Transform (WT) can capture these patterns through frequency decomposition, its coefficients are insensitive to change points in time series, leading to suboptimal modeling. To mitigate these limitations, we introduce the multi-order Wavelet Derivative Transform (WDT) grounded in the WT, enabling the extraction of time-aware patterns spanning both the overall trend and subtle fluctuations. Compared with the standard FT and WT, which model the raw series, the WDT operates on the derivative of the series, selectively magnifying rate-of-change cues and exposing abrupt regime shifts that are particularly informative for time series modeling. Practically, we embed the WDT into a multi-branch framework named WaveTS, which decomposes the input series into multi-scale time-frequency coefficients, refines them via linear layers, and reconstructs them into the time domain via the inverse WDT. Extensive experiments on ten benchmark datasets demonstrate that WaveTS achieves state-of-the-art forecasting accuracy while retaining high computational efficiency.
Authors: Gina Wong, Drew Prinster, Suchi Saria, Rama Chellappa, Anqi Liu
Abstract: Conformal prediction quantifies the uncertainty of machine learning models by augmenting point predictions with valid prediction sets, assuming exchangeability. For complex scenarios involving multiple trials, models, or data sources, conformal prediction sets can be aggregated to create a prediction set that captures the overall uncertainty, often improving precision. However, aggregating multiple prediction sets with individual $1-\alpha$ coverage inevitably weakens the overall guarantee, typically resulting in $1-2\alpha$ worst-case coverage. In this work, we propose a framework for the weighted aggregation of prediction sets, where weights are assigned to each prediction set based on their contribution. Our framework offers flexible control over how the sets are aggregated, achieving tighter coverage bounds that interpolate between the $1-2\alpha$ guarantee of the combined models and the $1-\alpha$ guarantee of an individual model depending on the distribution of weights. We extend our framework to data-dependent weights, and we derive a general procedure for data-dependent weight aggregation that maintains finite-sample validity. We demonstrate the effectiveness of our methods through experiments on synthetic and real data in the mixture-of-experts setting, and we show that aggregation with data-dependent weights provides a form of adaptive coverage.
Authors: Jesson Wang, Zhanhao Hu, David Wagner
Abstract: Large Language Models (LLMs) are trained with safety alignment to prevent generating malicious content. Although some attacks have highlighted vulnerabilities in these safety-aligned LLMs, they typically have limitations, such as necessitating access to the model weights or the generation process. Since proprietary models through API-calling do not grant users such permissions, these attacks find it challenging to compromise them. In this paper, we propose Jailbreaking Using LLM Introspection (JULI), which jailbreaks LLMs by manipulating the token log probabilities, using a tiny plug-in block, BiasNet. JULI relies solely on the knowledge of the target LLM's predicted token log probabilities. It can effectively jailbreak API-calling LLMs under a black-box setting and knowing only top-$5$ token log probabilities. Our approach demonstrates superior effectiveness, outperforming existing state-of-the-art (SOTA) approaches across multiple metrics.
Authors: Chuang Zhao, Hui Tang, Hongke Zhao, Xiaomeng Li
Abstract: Advanced healthcare predictions offer significant improvements in patient outcomes by leveraging predictive analytics. Existing works primarily utilize various views of Electronic Health Record (EHR) data, such as diagnoses, lab tests, or clinical notes, for model training. These methods typically assume the availability of complete EHR views and that the designed model could fully leverage the potential of each view. However, in practice, random missing views and view laziness present two significant challenges that hinder further improvements in multi-view utilization. To address these challenges, we introduce Diffmv, an innovative diffusion-based generative framework designed to advance the exploitation of multiple views of EHR data. Specifically, to address random missing views, we integrate various views of EHR data into a unified diffusion-denoising framework, enriched with diverse contextual conditions to facilitate progressive alignment and view transformation. To mitigate view laziness, we propose a novel reweighting strategy that assesses the relative advantages of each view, promoting a balanced utilization of various data views within the model. Our proposed strategy achieves superior performance across multiple health prediction tasks derived from three popular datasets, including multi-view and multi-modality scenarios.
Authors: Yang Tan, Wenrui Gou, Bozitao Zhong, Liang Hong, Huiqun Yu, Bingxin Zhou
Abstract: Deep learning models have driven significant progress in predicting protein function and interactions at the protein level. While these advancements have been invaluable for many biological applications such as enzyme engineering and function annotation, a more detailed perspective is essential for understanding protein functional mechanisms and evaluating the biological knowledge captured by models. To address this demand, we introduce VenusX, the first large-scale benchmark for fine-grained functional annotation and function-based protein pairing at the residue, fragment, and domain levels. VenusX comprises three major task categories across six types of annotations, including residue-level binary classification, fragment-level multi-class classification, and pairwise functional similarity scoring for identifying critical active sites, binding sites, conserved sites, motifs, domains, and epitopes. The benchmark features over 878,000 samples curated from major open-source databases such as InterPro, BioLiP, and SAbDab. By providing mixed-family and cross-family splits at three sequence identity thresholds, our benchmark enables a comprehensive assessment of model performance on both in-distribution and out-of-distribution scenarios. For baseline evaluation, we assess a diverse set of popular and open-source models, including pre-trained protein language models, sequence-structure hybrids, structure-based methods, and alignment-based techniques. Their performance is reported across all benchmark datasets and evaluation settings using multiple metrics, offering a thorough comparison and a strong foundation for future research. Code and data are publicly available at https://github.com/ai4protein/VenusX.
Authors: Siliang Zeng, Quan Wei, William Brown, Oana Frunza, Yuriy Nevmyvaka, Mingyi Hong
Abstract: This paper investigates approaches to enhance the reasoning capabilities of Large Language Model (LLM) agents using Reinforcement Learning (RL). Specifically, we focus on multi-turn tool-use scenarios, which can be naturally modeled as Markov Decision Processes (MDPs). While existing approaches often train multi-turn LLM agents with trajectory-level advantage estimation in bandit settings, they struggle with turn-level credit assignment across multiple decision steps, limiting their performance on multi-turn reasoning tasks. To address this, we introduce a fine-grained turn-level advantage estimation strategy to enable more precise credit assignment in multi-turn agent interactions. The strategy is general and can be incorporated into various RL algorithms such as Group Relative Preference Optimization (GRPO). Our experimental evaluation on multi-turn reasoning and search-based tool-use tasks with GRPO implementations highlights the effectiveness of the MDP framework and the turn-level credit assignment in advancing the multi-turn reasoning capabilities of LLM agents in complex decision-making settings. Our method achieves 100% success in tool execution and 50% accuracy in exact answer matching, significantly outperforming baselines, which fail to invoke tools and achieve only 20-30% exact match accuracy.
Authors: Yuhao Sun, Zhenyi Zhang, Zihan Wang, Tiejun Li, Peijie Zhou
Abstract: Recovering the dynamics from a few snapshots of a high-dimensional system is a challenging task in statistical physics and machine learning, with important applications in computational biology. Many algorithms have been developed to tackle this problem, based on frameworks such as optimal transport and the Schr\"odinger bridge. A notable recent framework is Regularized Unbalanced Optimal Transport (RUOT), which integrates both stochastic dynamics and unnormalized distributions. However, since many existing methods do not explicitly enforce optimality conditions, their solutions often struggle to satisfy the principle of least action and meet challenges to converge in a stable and reliable way. To address these issues, we propose Variational RUOT (Var-RUOT), a new framework to solve the RUOT problem. By incorporating the optimal necessary conditions for the RUOT problem into both the parameterization of the search space and the loss function design, Var-RUOT only needs to learn a scalar field to solve the RUOT problem and can search for solutions with lower action. We also examined the challenge of selecting a growth penalty function in the widely used Wasserstein-Fisher-Rao metric and proposed a solution that better aligns with biological priors in Var-RUOT. We validated the effectiveness of Var-RUOT on both simulated data and real single-cell datasets. Compared with existing algorithms, Var-RUOT can find solutions with lower action while exhibiting faster convergence and improved training stability.
Authors: Minsu Kim, Jean-Pierre Falet, Oliver E. Richardson, Xiaoyin Chen, Moksh Jain, Sungjin Ahn, Sungsoo Ahn, Yoshua Bengio
Abstract: Chain-of-Thought (CoT) reasoning has advanced the capabilities and transparency of language models (LMs); however, reasoning chains can contain inaccurate statements that reduce performance and trustworthiness. To address this, we introduce a new self-correction framework that augments each reasoning step in a CoT with a latent variable indicating its veracity, enabling modeling of all possible truth assignments rather than assuming correctness throughout. To efficiently explore this expanded space, we introduce Search Corrector, a discrete search algorithm over boolean-valued veracity assignments. It efficiently performs otherwise intractable inference in the posterior distribution over veracity assignments by leveraging the LM's joint likelihood over veracity and the final answer as a proxy reward. This efficient inference-time correction method facilitates supervised fine-tuning of an Amortized Corrector by providing pseudo-labels for veracity. The Amortized Corrector generalizes self-correction, enabling accurate zero-shot veracity inference in novel contexts. Empirical results demonstrate that Search Corrector reliably identifies errors in logical (ProntoQA) and mathematical reasoning (GSM8K) benchmarks. The Amortized Corrector achieves comparable zero-shot accuracy and improves final answer accuracy by up to 25%.
Authors: Jeremy Budd, Javier Ideami, Benjamin Macdowall Rynne, Keith Duggar, Randall Balestriero
Abstract: Sparse autoencoders (SAEs) have received considerable recent attention as tools for mechanistic interpretability, showing success at extracting interpretable features even from very large LLMs. However, this research has been largely empirical, and there have been recent doubts about the true utility of SAEs. In this work, we seek to enhance the theoretical understanding of SAEs, using the spline theory of deep learning. By situating SAEs in this framework: we discover that SAEs generalise ``$k$-means autoencoders'' to be piecewise affine, but sacrifice accuracy for interpretability vs. the optimal ``$k$-means-esque plus local principal component analysis (PCA)'' piecewise affine autoencoder. We characterise the underlying geometry of (TopK) SAEs using power diagrams. And we develop a novel proximal alternating method SGD (PAM-SGD) algorithm for training SAEs, with both solid theoretical foundations and promising empirical results in MNIST and LLM experiments, particularly in sample efficiency and (in the LLM setting) improved sparsity of codes. All code is available at: https://github.com/splInterp2025/splInterp
Authors: Ziyao Cui, Minxing Zhang, Jian Pei
Abstract: Nowadays, Large Language Models (LLMs) are trained on huge datasets, some including sensitive information. This poses a serious privacy concern because privacy attacks such as Membership Inference Attacks (MIAs) may detect this sensitive information. While knowledge distillation compresses LLMs into efficient, smaller student models, its impact on privacy remains underexplored. In this paper, we investigate how knowledge distillation affects model robustness against MIA. We focus on two questions. First, how is private data protected in teacher and student models? Second, how can we strengthen privacy preservation against MIAs in knowledge distillation? Through comprehensive experiments, we show that while teacher and student models achieve similar overall MIA accuracy, teacher models better protect member data, the primary target of MIA, whereas student models better protect non-member data. To address this vulnerability in student models, we propose 5 privacy-preserving distillation methods and demonstrate that they successfully reduce student models' vulnerability to MIA, with ensembling further stabilizing the robustness, offering a reliable approach for distilling more secure and efficient student models. Our implementation source code is available at https://github.com/richardcui18/MIA_in_KD.
Authors: Huan Li, Yiming Dong, Zhouchen Lin
Abstract: As the default optimizer for training large language models, AdamW has achieved remarkable success in deep learning. However, its convergence behavior is not theoretically well-understood. This paper establishes the convergence rate $\frac{1}{K}\sum_{k=1}^KE\left[\|\nabla f(x^k)\|_1\right]\leq O(\frac{\sqrt{d}C}{K^{1/4}})$ for AdamW measured by $\ell_1$ norm, where $K$ represents the iteration number, $d$ denotes the model dimension, and $C$ matches the constant in the optimal convergence rate of SGD. Theoretically, we have $E\left[\|\nabla f(x)\|_1\right]\geq\sqrt{\frac{2d}{\pi}}E\left[\|\nabla f(x)\|_2\right]$ when each element of $\nabla f(x)$ is generated from Gaussian distribution $\mathcal N(0,1)$. Empirically, our experimental results on real-world deep learning tasks reveal $\|\nabla f(x)\|_1=\varTheta(\sqrt{d})\|\nabla f(x)\|_2$. Both support that our convergence rate can be considered to be analogous to the optimal $\frac{1}{K}\sum_{k=1}^KE\left[\|\nabla f(x^k)\|_2\right]\leq O(\frac{C}{K^{1/4}})$ convergence rate of SGD.
Authors: Vahid Shahverdi, Giovanni Luca Marchetti, Kathl\'en Kohn
Abstract: Deep neural networks often infer sparse representations, converging to a subnetwork during the learning process. In this work, we theoretically analyze subnetworks and their bias through the lens of algebraic geometry. We consider fully-connected networks with polynomial activation functions, and focus on the geometry of the function space they parametrize, often referred to as neuromanifold. First, we compute the dimension of the subspace of the neuromanifold parametrized by subnetworks. Second, we show that this subspace is singular. Third, we argue that such singularities often correspond to critical points of the training dynamics. Lastly, we discuss convolutional networks, for which subnetworks and singularities are similarly related, but the bias does not arise.
Authors: Sizhe Ma, Katherine A. Flanigan, Mario Berg\'es
Abstract: Digital twins (DTs) enable powerful predictive analytics, but persistent discrepancies between simulations and real systems--known as the reality gap--undermine their reliability. Coined in robotics, the term now applies to DTs, where discrepancies stem from context mismatches, cross-domain interactions, and multi-scale dynamics. Among these, context mismatch is pressing and underexplored, as DT accuracy depends on capturing operational context, often only partially observable. However, DTs have a key advantage: simulators can systematically vary contextual factors and explore scenarios difficult or impossible to observe empirically, informing inference and model alignment. While sim-to-real transfer like domain adaptation shows promise in robotics, their application to DTs poses two key challenges. First, unlike one-time policy transfers, DTs require continuous calibration across an asset's lifecycle--demanding structured information flow, timely detection of out-of-sync states, and integration of historical and new data. Second, DTs often perform inverse modeling, inferring latent states or faults from observations that may reflect multiple evolving contexts. These needs strain purely data-driven models and risk violating physical consistency. Though some approaches preserve validity via reduced-order model, most domain adaptation techniques still lack such constraints. To address this, we propose a Reality Gap Analysis (RGA) module for DTs that continuously integrates new sensor data, detects misalignments, and recalibrates DTs via a query-response framework. Our approach fuses domain-adversarial deep learning with reduced-order simulator guidance to improve context inference and preserve physical consistency. We illustrate the RGA module in a structural health monitoring case study on a steel truss bridge in Pittsburgh, PA, showing faster calibration and better real-world alignment.
Authors: Kalyan Cherukuri, Aarav Lala, Yash Yardi
Abstract: We propose Q-Policy, a hybrid quantum-classical reinforcement learning (RL) framework that mathematically accelerates policy evaluation and optimization by exploiting quantum computing primitives. Q-Policy encodes value functions in quantum superposition, enabling simultaneous evaluation of multiple state-action pairs via amplitude encoding and quantum parallelism. We introduce a quantum-enhanced policy iteration algorithm with provable polynomial reductions in sample complexity for the evaluation step, under standard assumptions. To demonstrate the technical feasibility and theoretical soundness of our approach, we validate Q-Policy on classical emulations of small discrete control tasks. Due to current hardware and simulation limitations, our experiments focus on showcasing proof-of-concept behavior rather than large-scale empirical evaluation. Our results support the potential of Q-Policy as a theoretical foundation for scalable RL on future quantum devices, addressing RL scalability challenges beyond classical approaches.
Authors: Kalyan Cherukuri, Aarav Lala
Abstract: As generative agents become increasingly capable, alignment of their behavior with complex human values remains a fundamental challenge. Existing approaches often simplify human intent through reduction to a scalar reward, overlooking the multi-faceted nature of human feedback. In this work, we introduce a theoretical framework for preference-based Multi-Objective Inverse Reinforcement Learning (MO-IRL), where human preferences are modeled as latent vector-valued reward functions. We formalize the problem of recovering a Pareto-optimal reward representation from noisy preference queries and establish conditions for identifying the underlying multi-objective structure. We derive tight sample complexity bounds for recovering $\epsilon$-approximations of the Pareto front and introduce a regret formulation to quantify suboptimality in this multi-objective setting. Furthermore, we propose a provably convergent algorithm for policy optimization using preference-inferred reward cones. Our results bridge the gap between practical alignment techniques and theoretical guarantees, providing a principled foundation for learning aligned behaviors in a high-dimension and value-pluralistic environment.
Authors: Chi-Min Chan, Chunpu Xu, Jiaming Ji, Zhen Ye, Pengcheng Wen, Chunyang Jiang, Yaodong Yang, Wei Xue, Sirui Han, Yike Guo
Abstract: The current focus of AI research is shifting from emphasizing model training towards enhancing evaluation quality, a transition that is crucial for driving further advancements in AI systems. Traditional evaluation methods typically rely on reward models assigning scalar preference scores to outputs. Although effective, such approaches lack interpretability, leaving users often uncertain about why a reward model rates a particular response as high or low. The advent of LLM-as-a-Judge provides a more scalable and interpretable method of supervision, offering insights into the decision-making process. Moreover, with the emergence of large reasoning models, which consume more tokens for deeper thinking and answer refinement, scaling test-time computation in the LLM-as-a-Judge paradigm presents an avenue for further boosting performance and providing more interpretability through reasoning traces. In this paper, we introduce $\textbf{J1-7B}$, which is first supervised fine-tuned on reflection-enhanced datasets collected via rejection-sampling and subsequently trained using Reinforcement Learning (RL) with verifiable rewards. At inference time, we apply Simple Test-Time Scaling (STTS) strategies for additional performance improvement. Experimental results demonstrate that $\textbf{J1-7B}$ surpasses the previous state-of-the-art LLM-as-a-Judge by $ \textbf{4.8}$\% and exhibits a $ \textbf{5.1}$\% stronger scaling trend under STTS. Additionally, we present three key findings: (1) Existing LLM-as-a-Judge does not inherently exhibit such scaling trend. (2) Model simply fine-tuned on reflection-enhanced datasets continues to demonstrate similarly weak scaling behavior. (3) Significant scaling trend emerges primarily during the RL phase, suggesting that effective STTS capability is acquired predominantly through RL training.
Authors: Yifan Dai (College of Computer Science and Electronic Engineering, Hunan University), Xuanbai Ren (College of Computer Science and Electronic Engineering, Hunan University), Tengfei Ma (College of Computer Science and Electronic Engineering, Hunan University), Qipeng Yan (School of Biomedical Science, Hunan University), Yiping Liu (College of Computer Science and Electronic Engineering, Hunan University), Yuansheng Liu (College of Computer Science and Electronic Engineering, Hunan University), Xiangxiang Zeng (College of Computer Science and Electronic Engineering, Hunan University)
Abstract: Accurate molecular property prediction (MPP) is a critical step in modern drug development. However, the scarcity of experimental validation data poses a significant challenge to AI-driven research paradigms. Under few-shot learning scenarios, the quality of molecular representations directly dictates the theoretical upper limit of model performance. We present AdaptMol, a prototypical network integrating Adaptive multimodal fusion for Molecular representation. This framework employs a dual-level attention mechanism to dynamically integrate global and local molecular features derived from two modalities: SMILES sequences and molecular graphs. (1) At the local level, structural features such as atomic interactions and substructures are extracted from molecular graphs, emphasizing fine-grained topological information; (2) At the global level, the SMILES sequence provides a holistic representation of the molecule. To validate the necessity of multimodal adaptive fusion, we propose an interpretable approach based on identifying molecular active substructures to demonstrate that multimodal adaptive fusion can efficiently represent molecules. Extensive experiments on three commonly used benchmarks under 5-shot and 10-shot settings demonstrate that AdaptMol achieves state-of-the-art performance in most cases. The rationale-extracted method guides the fusion of two modalities and highlights the importance of both modalities.
Authors: Zihuan Qiu, Yi Xu, Chiyuan He, Fanman Meng, Linfeng Xu, Qingbo Wu, Hongliang Li
Abstract: Continual model merging integrates independently fine-tuned models sequentially without access to original training data, providing a scalable and efficient solution to continual learning. However, current methods still face critical challenges, notably parameter interference among tasks and limited adaptability to evolving test distributions. The former causes catastrophic forgetting of integrated tasks, while the latter hinders effective adaptation to new tasks. To address these, we propose MINGLE, a novel framework for test-time continual model merging, which leverages test-time adaptation using a small set of unlabeled test samples from the current task to dynamically guide the merging process. MINGLE employs a mixture-of-experts architecture composed of parameter-efficient, low-rank experts, enabling efficient adaptation and improving robustness to distribution shifts. To mitigate catastrophic forgetting, we propose Null-Space Constrained Gating, which restricts gating updates to subspaces orthogonal to prior task representations. This suppresses activations on old task inputs and preserves model behavior on past tasks. To further balance stability and adaptability, we design an Adaptive Relaxation Strategy, which dynamically adjusts the constraint strength based on interference signals captured during test-time adaptation. Extensive experiments on standard continual merging benchmarks demonstrate that MINGLE achieves robust generalization, reduces forgetting significantly, and consistently surpasses previous state-of-the-art methods by 7-9\% on average across diverse task orders.
Authors: Josh Alman, Zhao Song
Abstract: The transformer architecture has been widely applied to many machine learning tasks. A main bottleneck in the time to perform transformer computations is a task called attention computation. [Alman and Song, NeurIPS 2023] have shown that in the bounded entry regime, there is an almost linear time algorithm to approximate the attention computation. They also proved that the bounded entry assumption is necessary for a fast algorithm assuming the popular Strong Exponential Time Hypothesis. A new version of transformer which uses position embeddings has recently been very successful. At a high level, position embedding enables the model to capture the correlations between tokens while taking into account their position in the sequence. Perhaps the most popular and effective version is Rotary Position Embedding (RoPE), which was proposed by [Su, Lu, Pan, Murtadha, Wen, and Liu, Neurocomputing 2024]. A main downside of RoPE is that it complicates the attention computation problem, so that previous techniques for designing almost linear time algorithms no longer seem to work. In this paper, we show how to overcome this issue, and give a new algorithm to compute the RoPE attention in almost linear time in the bounded entry regime. (Again, known lower bounds imply that bounded entries are necessary.) Our new algorithm combines two techniques in a novel way: the polynomial method, which was used in prior fast attention algorithms, and the Fast Fourier Transform.
Authors: Chenwei Lou, Zewei Sun, Xinnian Liang, Meng Qu, Wei Shen, Wenqi Wang, Yuntao Li, Qingping Yang, Shuangzhi Wu
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities but often face challenges with tasks requiring sophisticated reasoning. While Chain-of-Thought (CoT) prompting significantly enhances reasoning, it indiscriminately generates lengthy reasoning steps for all queries, leading to substantial computational costs and inefficiency, especially for simpler inputs. To address this critical issue, we introduce AdaCoT (Adaptive Chain-of-Thought), a novel framework enabling LLMs to adaptively decide when to invoke CoT. AdaCoT framed adaptive reasoning as a Pareto optimization problem that seeks to balance model performance with the costs associated with CoT invocation (both frequency and computational overhead). We propose a reinforcement learning (RL) based method, specifically utilizing Proximal Policy Optimization (PPO), to dynamically control the CoT triggering decision boundary by adjusting penalty coefficients, thereby allowing the model to determine CoT necessity based on implicit query complexity. A key technical contribution is Selective Loss Masking (SLM), designed to counteract decision boundary collapse during multi-stage RL training, ensuring robust and stable adaptive triggering. Experimental results demonstrate that AdaCoT successfully navigates the Pareto frontier, achieving substantial reductions in CoT usage for queries not requiring elaborate reasoning. For instance, on our production traffic testset, AdaCoT reduced CoT triggering rates to as low as 3.18\% and decreased average response tokens by 69.06%, while maintaining high performance on complex tasks.
Authors: Jiang You, Xiaozhen Wang, Arben Cela
Abstract: We formulate time series tasks as input-output mappings under varying objectives, where the same input may yield different outputs. This challenges a model's generalization and adaptability. To study this, we construct a synthetic dataset with numerous conflicting subtasks to evaluate adaptation under frequent task shifts. Existing static models consistently fail in such settings. We propose a dynamic perturbed adaptive method based on a trunk-branch architecture, where the trunk evolves slowly to capture long-term structure, and branch modules are re-initialized and updated for each task. This enables continual test-time adaptation and cross-task transfer without relying on explicit task labels. Theoretically, we show that this architecture has strictly higher functional expressivity than static models and LoRA. We also establish exponential convergence of branch adaptation under the Polyak-Lojasiewicz condition. Experiments demonstrate that our method significantly outperforms competitive baselines in complex and conflicting task environments, exhibiting fast adaptation and progressive learning capabilities.
Authors: Louis Mahon, Mirella Lapata
Abstract: Clustering is a widely used and powerful machine learning technique, but its effectiveness is often limited by the need to specify the number of clusters, k, or by relying on thresholds that implicitly determine k. We introduce k*-means, a novel clustering algorithm that eliminates the need to set k or any other parameters. Instead, it uses the minimum description length principle to automatically determine the optimal number of clusters, k*, by splitting and merging clusters while also optimising the standard k-means objective. We prove that k*-means is guaranteed to converge and demonstrate experimentally that it significantly outperforms existing methods in scenarios where k is unknown. We also show that it is accurate in estimating k, and that empirically its runtime is competitive with existing methods, and scales well with dataset size.
Authors: Paul Saves, Nicolas Verstaevel, Beno\^it Gaudou
Abstract: Multi-agent simulations enables the modeling and analyses of the dynamic behaviors and interactions of autonomous entities evolving in complex environments. Agent-based models (ABM) are widely used to study emergent phenomena arising from local interactions. However, their high computational cost poses a significant challenge, particularly for large-scale simulations requiring extensive parameter exploration, optimization, or uncertainty quantification. The increasing complexity of ABM limits their feasibility for real-time decision-making and large-scale scenario analysis. To address these limitations, surrogate models offer an efficient alternative by learning approximations from sparse simulation data. These models provide cheap-to-evaluate predictions, significantly reducing computational costs while maintaining accuracy. Various machine learning techniques, including regression models, neural networks, random forests and Gaussian processes, have been applied to construct robust surrogates. Moreover, uncertainty quantification and sensitivity analysis play a crucial role in enhancing model reliability and interpretability. This article explores the motivations, methods, and applications of surrogate modeling for ABM, emphasizing the trade-offs between accuracy, computational efficiency, and interpretability. Through a case study on a segregation model, we highlight the challenges associated with building and validating surrogate models, comparing different approaches and evaluating their performance. Finally, we discuss future perspectives on integrating surrogate models within ABM to improve scalability, explainability, and real-time decision support across various fields such as ecology, urban planning and economics.
Authors: Zhiheng Chen, Ruofan Wu, Guanhua Fang
Abstract: The transformer architecture has demonstrated remarkable capabilities in modern artificial intelligence, among which the capability of implicitly learning an internal model during inference time is widely believed to play a key role in the under standing of pre-trained large language models. However, most recent works have been focusing on studying supervised learning topics such as in-context learning, leaving the field of unsupervised learning largely unexplored. This paper investigates the capabilities of transformers in solving Gaussian Mixture Models (GMMs), a fundamental unsupervised learning problem through the lens of statistical estimation. We propose a transformer-based learning framework called TGMM that simultaneously learns to solve multiple GMM tasks using a shared transformer backbone. The learned models are empirically demonstrated to effectively mitigate the limitations of classical methods such as Expectation-Maximization (EM) or spectral algorithms, at the same time exhibit reasonable robustness to distribution shifts. Theoretically, we prove that transformers can approximate both the EM algorithm and a core component of spectral methods (cubic tensor power iterations). These results bridge the gap between practical success and theoretical understanding, positioning transformers as versatile tools for unsupervised learning.
Authors: Nikolai West, Jochen Deuse
Abstract: This paper presents a comprehensive collection of industrial screw driving datasets designed to advance research in manufacturing process monitoring and quality control. The collection comprises six distinct datasets with over 34,000 individual screw driving operations conducted under controlled experimental conditions, capturing the multifaceted nature of screw driving processes in plastic components. Each dataset systematically investigates specific aspects: natural thread degradation patterns through repeated use (s01), variations in surface friction conditions including contamination and surface treatments (s02), diverse assembly faults with up to 27 error types (s03-s04), and fabrication parameter variations in both upper and lower workpieces through modified injection molding settings (s05-s06). We detail the standardized experimental setup used across all datasets, including hardware specifications, process phases, and data acquisition methods. The hierarchical data model preserves the temporal and operational structure of screw driving processes, facilitating both exploratory analysis and the development of machine learning models. To maximize accessibility, we provide dual access pathways: raw data through Zenodo with a persistent DOI, and a purpose-built Python library (PyScrew) that offers consistent interfaces for data loading, preprocessing, and integration with common analysis workflows. These datasets serve diverse research applications including anomaly detection, predictive maintenance, quality control system development, feature extraction methodology evaluation, and classification of specific error conditions. By addressing the scarcity of standardized, comprehensive datasets in industrial manufacturing, this collection enables reproducible research and fair comparison of analytical approaches in an area of growing importance for industrial automation.
Authors: Marco S\"alzer, Przemys{\l}aw Andrzej Wa{\l}\k{e}ga, Martin Lange
Abstract: In recent years, the expressive power of various neural architectures -- including graph neural networks (GNNs), transformers, and recurrent neural networks -- has been characterised using tools from logic and formal language theory. As the capabilities of basic architectures are becoming well understood, increasing attention is turning to models that combine multiple architectural paradigms. Among them particularly important, and challenging to analyse, are temporal extensions of GNNs, which integrate both spatial (graph-structure) and temporal (evolution over time) dimensions. In this paper, we initiate the study of logical characterisation of temporal GNNs by connecting them to two-dimensional product logics. We show that the expressive power of temporal GNNs depends on how graph and temporal components are combined. In particular, temporal GNNs that apply static GNNs recursively over time can capture all properties definable in the product logic of (past) propositional temporal logic PTL and the modal logic K. In contrast, architectures such as graph-and-time TGNNs and global TGNNs can only express restricted fragments of this logic, where the interaction between temporal and spatial operators is syntactically constrained. These results yield the first logical characterisations of temporal GNNs and establish new relative expressiveness results for temporal GNNs.
Authors: Jingren Liu, Zhong Ji, Xiangyu Chen
Abstract: While diffusion models have achieved remarkable success in static data generation, their deployment in streaming or continual learning (CL) scenarios faces a major challenge: catastrophic forgetting (CF), where newly acquired generative capabilities overwrite previously learned ones. To systematically address this, we introduce a formal Continual Diffusion Generation (CDG) paradigm that characterizes and redefines CL in the context of generative diffusion models. Prior efforts often adapt heuristic strategies from continual classification tasks but lack alignment with the underlying diffusion process. In this work, we develop the first theoretical framework for CDG by analyzing cross-task dynamics in diffusion-based generative modeling. Our analysis reveals that the retention and stability of generative knowledge across tasks are governed by three key consistency criteria: inter-task knowledge consistency (IKC), unconditional knowledge consistency (UKC), and label knowledge consistency (LKC). Building on these insights, we propose Continual Consistency Diffusion (CCD), a principled framework that integrates these consistency objectives into training via hierarchical loss terms $\mathcal{L}_{IKC}$, $\mathcal{L}_{UKC}$, and $\mathcal{L}_{LKC}$. This promotes effective knowledge retention while enabling the assimilation of new generative capabilities. Extensive experiments on four benchmark datasets demonstrate that CCD achieves state-of-the-art performance under continual settings, with substantial gains in Mean Fidelity (MF) and Incremental Mean Fidelity (IMF), particularly in tasks with rich cross-task knowledge overlap.
Authors: Puning Yang, Qizhou Wang, Zhuo Huang, Tongliang Liu, Chengqi Zhang, Bo Han
Abstract: Loss reweighting has shown significant benefits for machine unlearning with large language models (LLMs). However, their exact functionalities are left unclear and the optimal strategy remains an open question, thus impeding the understanding and improvement of existing methodologies. In this paper, we identify two distinct goals of loss reweighting, namely, Saturation and Importance -- the former indicates that those insufficiently optimized data should be emphasized, while the latter stresses some critical data that are most influential for loss minimization. To study their usefulness, we design specific reweighting strategies for each goal and evaluate their respective effects on unlearning. We conduct extensive empirical analyses on well-established benchmarks, and summarize some important observations as follows: (i) Saturation enhances efficacy more than importance-based reweighting, and their combination can yield additional improvements. (ii) Saturation typically allocates lower weights to data with lower likelihoods, whereas importance-based reweighting does the opposite. (iii) The efficacy of unlearning is also largely influenced by the smoothness and granularity of the weight distributions. Based on these findings, we propose SatImp, a simple reweighting method that combines the advantages of both saturation and importance. Empirical results on extensive datasets validate the efficacy of our method, potentially bridging existing research gaps and indicating directions for future research. Our code is available at https://github.com/Puning97/SatImp-for-LLM-Unlearning.
URLs: https://github.com/Puning97/SatImp-for-LLM-Unlearning.
Authors: Daniyar Zakarin, Sidak Pal Singh
Abstract: Recent work has highlighted a surprising alignment between gradients and the top eigenspace of the Hessian -- termed the Dominant subspace -- during neural network training. Concurrently, there has been growing interest in the distinct roles of sharp and flat directions in the Hessian spectrum. In this work, we study Bulk-SGD, a variant of SGD that restricts updates to the orthogonal complement of the Dominant subspace. Through ablation studies, we characterize the stability properties of Bulk-SGD and identify critical hyperparameters that govern its behavior. We show that updates along the Bulk subspace, corresponding to flatter directions in the loss landscape, can accelerate convergence but may compromise stability. To balance these effects, we introduce interpolated gradient methods that unify SGD, Dom-SGD, and Bulk-SGD. Finally, we empirically connect this subspace decomposition to the Generalized Gauss-Newton and Functional Hessian terms, showing that curvature energy is largely concentrated in the Dominant subspace. Our findings suggest a principled approach to designing curvature-aware optimizers.
Authors: Zihao Zheng (Eric), Ziyao Wang (Eric), Xiuping Cui (Eric), Maoliang Li (Eric), Jiayu Chen (Eric), Yun (Eric), Liang, Ang Li, Xiang Chen
Abstract: Federated Learning (FL) is a decentralized model training approach that preserves data privacy but struggles with low efficiency. Quantization, a powerful training optimization technique, has been widely explored for integration into FL. However, many studies fail to consider the distinct performance attribution between particular quantization strategies, such as post-training quantization (PTQ) or quantization-aware training (QAT). As a result, existing FL quantization methods rely solely on either PTQ or QAT, optimizing for speed or accuracy while compromising the other. To efficiently accelerate FL and maintain distributed convergence accuracy across various FL settings, this paper proposes a hybrid quantitation approach combining PTQ and QAT for FL systems. We conduct case studies to validate the effectiveness of using hybrid quantization in FL. To solve the difficulty of modeling speed and accuracy caused by device and data heterogeneity, we propose a hardware-related analysis and data-distribution-related analysis to help identify the trade-off boundaries for strategy selection. Based on these, we proposed a novel framework named FedHQ to automatically adopt optimal hybrid strategy allocation for FL systems. Specifically, FedHQ develops a coarse-grained global initialization and fine-grained ML-based adjustment to ensure efficiency and robustness. Experiments show that FedHQ achieves up to 2.47x times training acceleration and up to 11.15% accuracy improvement and negligible extra overhead.
Authors: Sabrina Khurshid, Gourab Ghatak, Mohammad Shahid Abdulla
Abstract: This paper focuses on selecting the arm with the highest variance from a set of $K$ independent arms. Specifically, we focus on two settings: (i) regret setting, that penalizes the number of pulls of suboptimal arms in terms of variance, and (ii) fixed-budget \ac{BAI} setting, that evaluates the ability of an algorithm to determine the arm with the highest variance after a fixed number of pulls. We develop a novel online algorithm called \texttt{UCB-VV} for the regret setting and show that its upper bound on regret for bounded rewards evolves as $\mathcal{O}\left(\log{n}\right)$ where $n$ is the horizon. By deriving the lower bound on the regret, we show that \texttt{UCB-VV} is order optimal. For the fixed budget \ac{BAI} setting and propose the \texttt{SHVV} algorithm. We show that the upper bound of the error probability of \texttt{SHVV} evolves as $\exp\left(-\frac{n}{\log(K) H}\right)$, where $H$ represents the complexity of the problem, and this rate matches the corresponding lower bound. We extend the framework from bounded distributions to sub-Gaussian distributions using a novel concentration inequality on the sample variance. Leveraging the same, we derive a concentration inequality for the empirical Sharpe ratio (SR) for sub-Gaussian distributions, which was previously unknown in the literature. Empirical simulations show that \texttt{UCB-VV} consistently outperforms \texttt{$\epsilon$-greedy} across different sub-optimality gaps though it is surpassed by \texttt{VTS}, which exhibits the lowest regret, albeit lacking in theoretical guarantees. We also illustrate the superior performance of \texttt{SHVV}, for a fixed budget setting under 6 different setups against uniform sampling. Finally, we conduct a case study to empirically evaluate the performance of the \texttt{UCB-VV} and \texttt{SHVV} in call option trading on $100$ stocks generated using \ac{GBM}.
Authors: Prashant Shivaram Bhat, Shakib Yazdani, Elahe Arani, Bahram Zonooz
Abstract: Catastrophic forgetting has remained a critical challenge for deep neural networks in Continual Learning (CL) as it undermines consolidated knowledge when learning new tasks. Parameter efficient fine tuning CL techniques are gaining traction for their effectiveness in addressing catastrophic forgetting with a lightweight training schedule while avoiding degradation of consolidated knowledge in pre-trained models. However, low rank adapters (LoRA) in these approaches are highly sensitive to rank selection which can lead to sub-optimal resource allocation and performance. To this end, we introduce PEARL, a rehearsal-free CL framework that entails dynamic rank allocation for LoRA components during CL training. Specifically, PEARL leverages reference task weights and adaptively determines the rank of task-specific LoRA components based on the current tasks' proximity to reference task weights in parameter space. To demonstrate the versatility of PEARL, we evaluate it across three vision architectures (ResNet, Separable Convolutional Network and Vision Transformer) and a multitude of CL scenarios, and show that PEARL outperforms all considered baselines by a large margin.
Authors: Davide Murari, Takashi Furuya, Carola-Bibiane Sch\"onlieb
Abstract: 1-Lipschitz neural networks are fundamental for generative modelling, inverse problems, and robust classifiers. In this paper, we focus on 1-Lipschitz residual networks (ResNets) based on explicit Euler steps of negative gradient flows and study their approximation capabilities. Leveraging the Restricted Stone-Weierstrass Theorem, we first show that these 1-Lipschitz ResNets are dense in the set of scalar 1-Lipschitz functions on any compact domain when width and depth are allowed to grow. We also show that these networks can exactly represent scalar piecewise affine 1-Lipschitz functions. We then prove a stronger statement: by inserting norm-constrained linear maps between the residual blocks, the same density holds when the hidden width is fixed. Because every layer obeys simple norm constraints, the resulting models can be trained with off-the-shelf optimisers. This paper provides the first universal approximation guarantees for 1-Lipschitz ResNets, laying a rigorous foundation for their practical use.
Authors: Xi Han, Jingwei Zhang, Dimitris Samaras, Fei Hou, Hong Qin
Abstract: The neural operator (NO) framework has emerged as a powerful tool for solving partial differential equations (PDEs). Recent NOs are dominated by the Transformer architecture, which offers NOs the capability to capture long-range dependencies in PDE dynamics. However, existing Transformer-based NOs suffer from quadratic complexity, lack geometric rigor, and thus suffer from sub-optimal performance on regular grids. As a remedy, we propose the Geometric Mamba Neural Operator (GeoMaNO) framework, which empowers NOs with Mamba's modeling capability, linear complexity, plus geometric rigor. We evaluate GeoMaNO's performance on multiple standard and popularly employed PDE benchmarks, spanning from Darcy flow problems to Navier-Stokes problems. GeoMaNO improves existing baselines in solution operator approximation by as much as 58.9%.
Authors: Praveen Venkateswaran, Danish Contractor
Abstract: In many real-world applications, users rely on natural language instructions to guide large language models (LLMs) across a wide range of tasks. These instructions are often complex, diverse, and subject to frequent change. However, LLMs do not always attend to these instructions reliably, and users lack simple mechanisms to emphasize their importance beyond modifying prompt wording or structure. To address this, we present an inference-time method that enables users to emphasize specific parts of their prompt by steering the model's attention toward them, aligning the model's perceived importance of different prompt tokens with user intent. Unlike prior approaches that are limited to static instructions, require significant offline profiling, or rely on fixed biases, we dynamically update the proportion of model attention given to the user-specified parts--ensuring improved instruction following without performance degradation. We demonstrate that our approach improves instruction following across a variety of tasks involving multiple instructions and generalizes across models of varying scales.
Authors: Jianxiang Yu, Jiapeng Zhu, Hao Qian, Ziqi Liu, Zhiqiang Zhang, Xiang Li
Abstract: In recent years, large language models (LLMs) have demonstrated remarkable generalization capabilities across various natural language processing (NLP) tasks. Similarly, graph foundation models (GFMs) have emerged as a promising direction in graph learning, aiming to generalize across diverse datasets through large-scale pre-training. However, unlike language models that rely on explicit token representations, graphs lack a well-defined unit for generalization, making it challenging to design effective pre-training strategies. In this work, we propose REEF, a novel framework that leverages relation tokens as the basic units for GFMs. Inspired by the token vocabulary in LLMs, we construct a relation vocabulary of relation tokens to store relational information within graphs. To accommodate diverse relations, we introduce two hypernetworks that adaptively generate the parameters of aggregators and classifiers in graph neural networks based on relation tokens. In addition, we design another hypernetwork to construct dataset-specific projectors and incorporate a dataset-level feature bias into the initial node representations, enhancing flexibility across different datasets with the same relation. Further, we adopt graph data augmentation and a mixed-dataset pre-training strategy, allowing REEF to capture relational diversity more effectively and exhibit strong generalization capabilities. Extensive experiments show that REEF significantly outperforms existing methods on both pre-training and transfer learning tasks, underscoring its potential as a powerful foundation model for graph-based applications.
Authors: Jiashuo Jiang, Yiming Zong, Yinyu Ye
Abstract: Reinforcement learning (RL) problems are fundamental in online decision-making and have been instrumental in finding an optimal policy for Markov decision processes (MDPs). Function approximations are usually deployed to handle large or infinite state-action space. In our work, we consider the RL problems with function approximation and we develop a new algorithm to solve it efficiently. Our algorithm is based on the linear programming (LP) reformulation and it resolves the LP at each iteration improved with new data arrival. Such a resolving scheme enables our algorithm to achieve an instance-dependent sample complexity guarantee, more precisely, when we have $N$ data, the output of our algorithm enjoys an instance-dependent $\tilde{O}(1/N)$ suboptimality gap. In comparison to the $O(1/\sqrt{N})$ worst-case guarantee established in the previous literature, our instance-dependent guarantee is tighter when the underlying instance is favorable, and the numerical experiments also reveal the efficient empirical performances of our algorithms.
Authors: Ning Lu, Shengcai Liu, Jiahao Wu, Weiyu Chen, Zhirui Zhang, Yew-Soon Ong, Qi Wang, Ke Tang
Abstract: Large language models (LLMs) have shown great potential as general-purpose AI assistants across various domains. To fully leverage this potential in specific applications, many companies provide fine-tuning API services, enabling users to upload their own data for LLM customization. However, fine-tuning services introduce a new safety threat: user-uploaded data, whether harmful or benign, can break the model's alignment, leading to unsafe outputs. Moreover, existing defense methods struggle to address the diversity of fine-tuning datasets (e.g., varying sizes, tasks), often sacrificing utility for safety or vice versa. To address this issue, we propose Safe Delta, a safety-aware post-training defense method that adjusts the delta parameters (i.e., the parameter change before and after fine-tuning). Specifically, Safe Delta estimates the safety degradation, selects delta parameters to maximize utility while limiting overall safety loss, and applies a safety compensation vector to mitigate residual safety loss. Through extensive experiments on four diverse datasets with varying settings, our approach consistently preserves safety while ensuring that the utility gain from benign datasets remains unaffected.
Authors: James Jackaman, Oliver Sutton
Abstract: In this paper we design a neural interpolation operator to improve the boundary data for regional weather models, which is a challenging problem as we are required to map multi-scale dynamics between grid resolutions. In particular, we expose a methodology for approaching the problem through the study of a simplified model, with a view to generalise the results in this work to the dynamical core of regional weather models. Our approach will exploit a combination of techniques from image super-resolution with convolutional neural networks (CNNs) and residual networks, in addition to building the flow of atmospheric dynamics into the neural network
Authors: Haixu Wu, Minghao Guo, Yuezhou Ma, Yuanxu Sun, Jianmin Wang, Wojciech Matusik, Mingsheng Long
Abstract: Attention mechanism has emerged as a foundation module of modern deep learning models and has also empowered many milestones in various domains. Moreover, FlashAttention with IO-aware speedup resolves the efficiency issue of standard attention, further promoting its practicality. Beyond canonical attention, attention with bias also widely exists, such as relative position bias in vision and language models and pair representation bias in AlphaFold. In these works, prior knowledge is introduced as an additive bias term of attention weights to guide the learning process, which has been proven essential for model performance. Surprisingly, despite the common usage of attention with bias, its targeted efficiency optimization is still absent, which seriously hinders its wide applications in complex tasks. Diving into the computation of FlashAttention, we prove that its optimal efficiency is determined by the rank of the attention weight matrix. Inspired by this theoretical result, this paper presents FlashBias based on the low-rank compressed sensing theory, which can provide fast-exact computation for many widely used attention biases and a fast-accurate approximation for biases in general formalization. FlashBias can fully take advantage of the extremely optimized matrix multiplication operation in modern GPUs, achieving 1.5$\times$ speedup for AlphaFold, and over 2$\times$ speedup for attention with bias in vision and language models without loss of accuracy.
Authors: Andreas Hadjipieris, Neofytos Dimitriou, Ognjen Arandjelovi\'c
Abstract: Port berthing sites are regions of high interest for monitoring and optimizing port operations. Data sourced from the Automatic Identification System (AIS) can be superimposed on berths enabling their real-time monitoring and revealing long-term utilization patterns. Ultimately, insights from multiple berths can uncover bottlenecks, and lead to the optimization of the underlying supply chain of the port and beyond. However, publicly available documentation of port berths, even when available, is frequently incomplete - e.g. there may be missing berths or inaccuracies such as incorrect boundary boxes - necessitating a more robust, data-driven approach to port berth localization. In this context, we propose an unsupervised spatial modeling method that leverages AIS data clustering and hyperparameter optimization to identify berthing sites. Trained on one month of freely available AIS data and evaluated across ports of varying sizes, our models significantly outperform competing methods, achieving a mean Bhattacharyya distance of 0.85 when comparing Gaussian Mixture Models (GMMs) trained on separate data splits, compared to 13.56 for the best existing method. Qualitative comparison with satellite images and existing berth labels further supports the superiority of our method, revealing more precise berth boundaries and improved spatial resolution across diverse port environments.
Authors: Mehran Shakerinava, Siamak Ravanbakhsh, Adam Oberman
Abstract: Recent work has formalized the reward hypothesis through the lens of expected utility theory, by interpreting reward as utility. Hausner's foundational work showed that dropping the continuity axiom leads to a generalization of expected utility theory where utilities are lexicographically ordered vectors of arbitrary dimension. In this paper, we extend this result by identifying a simple and practical condition under which preferences cannot be represented by scalar rewards, necessitating a 2-dimensional reward function. We provide a full characterization of such reward functions, as well as the general d-dimensional case, in Markov Decision Processes (MDPs) under a memorylessness assumption on preferences. Furthermore, we show that optimal policies in this setting retain many desirable properties of their scalar-reward counterparts, while in the Constrained MDP (CMDP) setting -- another common multiobjective setting -- they do not.
Authors: Jianke Yang, Manu Bhat, Bryan Hu, Yadi Cao, Nima Dehmamy, Robin Walters, Rose Yu
Abstract: Discovering symbolic differential equations from data uncovers fundamental dynamical laws underlying complex systems. However, existing methods often struggle with the vast search space of equations and may produce equations that violate known physical laws. In this work, we address these problems by introducing the concept of \textit{symmetry invariants} in equation discovery. We leverage the fact that differential equations admitting a symmetry group can be expressed in terms of differential invariants of symmetry transformations. Thus, we propose to use these invariants as atomic entities in equation discovery, ensuring the discovered equations satisfy the specified symmetry. Our approach integrates seamlessly with existing equation discovery methods such as sparse regression and genetic programming, improving their accuracy and efficiency. We validate the proposed method through applications to various physical systems, such as fluid and reaction-diffusion, demonstrating its ability to recover parsimonious and interpretable equations that respect the laws of physics.
Authors: M Ruhul Amin
Abstract: This paper introduces Attribution Projection Calculus (AP-Calculus), a novel mathematical framework for determining causal relationships in structured Bayesian networks. We investigate a specific network architecture with source nodes connected to destination nodes through intermediate nodes, where each input maps to a single label with maximum marginal probability. We prove that for each label, exactly one intermediate node acts as a deconfounder while others serve as confounders, enabling optimal attribution of features to their corresponding labels. The framework formalizes the dual nature of intermediate nodes as both confounders and deconfounders depending on the context, and establishes separation functions that maximize distinctions between intermediate representations. We demonstrate that the proposed network architecture is optimal for causal inference compared to alternative structures, including those based on Pearl's causal framework. AP-Calculus provides a comprehensive mathematical foundation for analyzing feature-label attributions, managing spurious correlations, quantifying information gain, ensuring fairness, and evaluating uncertainty in prediction models, including large language models. Theoretical verification shows that AP-Calculus not only extends but can also subsume traditional do-calculus for many practical applications, offering a more direct approach to causal inference in supervised learning contexts.
Authors: Alberto Bassi, Carlo Albert, Aurelien Lucchi, Marco Baity-Jesi, Emanuele Francazi
Abstract: Understanding the statistical properties of deep neural networks (DNNs) at initialization is crucial for elucidating both their trainability and the intrinsic architectural biases they encode prior to data exposure. Mean-field (MF) analyses have demonstrated that the parameter distribution in randomly initialized networks dictates whether gradients vanish or explode. Concurrently, untrained DNNs were found to exhibit an initial-guessing bias (IGB), in which large regions of the input space are assigned to a single class. In this work, we derive a theoretical proof establishing the correspondence between IGB and previous MF theories, thereby connecting a network prejudice toward specific classes with the conditions for fast and accurate learning. This connection yields the counter-intuitive conclusion: the initialization that optimizes trainability is necessarily biased, rather than neutral. Furthermore, we extend the MF/IGB framework to multi-node activation functions, offering practical guidelines for designing initialization schemes that ensure stable optimization in architectures employing max- and average-pooling layers.
Authors: Matthew Landers, Taylor W. Killian, Thomas Hartvigsen, Afsaneh Doryab
Abstract: The combinatorial structure of many real-world action spaces leads to exponential growth in the number of possible actions, limiting the effectiveness of conventional reinforcement learning algorithms. Recent approaches for combinatorial action spaces impose factorized or sequential structures over sub-actions, failing to capture complex joint behavior. We introduce the Sub-Action Interaction Network using Transformers (SAINT), a novel policy architecture that represents multi-component actions as unordered sets and models their dependencies via self-attention conditioned on the global state. SAINT is permutation-invariant, sample-efficient, and compatible with standard policy optimization algorithms. In 15 distinct combinatorial environments across three task domains, including environments with nearly 17 million joint actions, SAINT consistently outperforms strong baselines.
Authors: Yueqi Cao, Anthea Monod
Abstract: We propose new graph kernels grounded in the study of metric graphs via tropical algebraic geometry. In contrast to conventional graph kernels that are based on graph combinatorics such as nodes, edges, and subgraphs, our graph kernels are purely based on the geometry and topology of the underlying metric space. A key characterizing property of our construction is its invariance under edge subdivision, making the kernels intrinsically well-suited for comparing graphs that represent different underlying spaces. We develop efficient algorithms for computing these kernels and analyze their complexity, showing that it depends primarily on the genus of the input graphs. Empirically, our kernels outperform existing methods in label-free settings, as demonstrated on both synthetic and real-world benchmark datasets. We further highlight their practical utility through an urban road network classification task.
Authors: Can Polat, Hasan Kurban, Erchin Serpedin, Mustafa Kurban
Abstract: Molecular graph neural networks (GNNs) often focus exclusively on XYZ-based geometric representations and thus overlook valuable chemical context available in public databases like PubChem. This work introduces a multimodal framework that integrates textual descriptors, such as IUPAC names, molecular formulas, physicochemical properties, and synonyms, alongside molecular graphs. A gated fusion mechanism balances geometric and textual features, allowing models to exploit complementary information. Experiments on benchmark datasets indicate that adding textual data yields notable improvements for certain electronic properties, while gains remain limited for others. Furthermore, the GNN architectures display similar performance patterns (improving and deteriorating on analogous targets), suggesting they learn comparable representations rather than distinctly different physical insights.
Authors: Fei Lu, Yue Yu
Abstract: Transformers have shown a remarkable ability for in-context learning (ICL), making predictions based on contextual examples. However, while theoretical analyses have explored this prediction capability, the nature of the inferred context and its utility for downstream predictions remain open questions. This paper aims to address these questions by examining ICL for inverse linear regression (ILR), where context inference can be characterized by unsupervised learning of underlying weight vectors. Focusing on the challenging scenario of rank-deficient inverse problems, where context length is smaller than the number of unknowns in the weight vectors and regularization is necessary, we introduce a linear transformer to learn the inverse mapping from contextual examples to the underlying weight vector. Our findings reveal that the transformer implicitly learns both a prior distribution and an effective regularization strategy, outperforming traditional ridge regression and regularization methods. A key insight is the necessity of low task dimensionality relative to the context length for successful learning. Furthermore, we numerically verify that the error of the transformer estimator scales linearly with the noise level, the ratio of task dimension to context length, and the condition number of the input data. These results not only demonstrate the potential of transformers for solving ill-posed inverse problems, but also provide a new perspective towards understanding the knowledge extraction mechanism within transformers.
Authors: Arun Kumar, Paul Schrater
Abstract: Invariant representations are core to representation learning, yet a central challenge remains: uncovering invariants that are stable and transferable without suppressing task-relevant signals. This raises fundamental questions, requiring further inquiry, about the appropriate level of abstraction at which such invariants should be defined, and which aspects of a system they should characterize. Interpretation of the environment relies on abstract knowledge structures to make sense of the current state, which leads to interactions, essential drivers of learning and knowledge acquisition. We posit that interpretation operates at the level of higher-order relational knowledge; hence, invariant structures must be where knowledge resides, specifically, as partitions defined by the closure of relational paths within an abstract knowledge space. These partitions serve as the core invariant representations, forming the structural substrate where knowledge is stored and learning occurs. On the other hand, inter-partition connectors enable the deployment of these knowledge partitions encoding task-relevant transitions. Thus, invariant partitions provide the foundational primitives of structured representation. We formalize the computational foundations for structured representation of the invariant partitions based on closed semiring, a relational algebraic structure.
Authors: Nikolaos-Lysias Kosioris, Sotirios Nikoletseas, Gavrilis Filios, Stefanos Panagiotou
Abstract: The rapid increase in computing power and the ability to store Big Data in the infrastructure has enabled predictions in a large variety of domains by Machine Learning. However, in many cases, existing Machine Learning tools are considered insufficient or incorrect since they exploit only probabilistic dependencies rather than inference logic. Causal Machine Learning methods seem to close this gap. In this paper, two prevalent tools based on Causal Machine Learning methods are compared, as well as their mathematical underpinning background. The operation of the tools is demonstrated by examining their response to 18 queries, based on the IDEAL Household Energy Dataset, published by the University of Edinburgh. First, it was important to evaluate the causal relations assumption that allowed the use of this approach; this was based on the preexisting scientific knowledge of the domain and was implemented by use of the in-built validation tools. Results were encouraging and may easily be extended to other domains.
Authors: Andr\'es Guzm\'an-Cordero, Felix Dangel, Gil Goldshlager, Marius Zeinhofer
Abstract: Natural gradient methods significantly accelerate the training of Physics-Informed Neural Networks (PINNs), but are often prohibitively costly. We introduce a suite of techniques to improve the accuracy and efficiency of energy natural gradient descent (ENGD) for PINNs. First, we leverage the Woodbury formula to dramatically reduce the computational complexity of ENGD. Second, we adapt the Subsampled Projected-Increment Natural Gradient Descent algorithm from the variational Monte Carlo literature to accelerate the convergence. Third, we explore the use of randomized algorithms to further reduce the computational cost in the case of large batch sizes. We find that randomization accelerates progress in the early stages of training for low-dimensional problems, and we identify key barriers to attaining acceleration in other scenarios. Our numerical experiments demonstrate that our methods outperform previous approaches, achieving the same $L^2$ error as the original ENGD up to $75\times$ faster.
Authors: Alex Heyman, Joel Zylberberg
Abstract: Large language models have recently made great strides in reasoning task performance through chain-of-thought (CoT) strategies trained via reinforcement learning; however, these "reasoning large language models" (RLLMs) remain imperfect reasoners, and understanding the frequencies and causes of their failure modes is important for both users and developers. We test o1-mini, o3-mini, DeepSeek-R1, Claude 3.7 Sonnet, Gemini 2.5 Pro Preview, and Grok 3 Mini Beta on graph coloring as a variable-complexity constraint-satisfaction logic problem, and find evidence from both error rate comparisons and CoT/explanation text analysis that RLLMs are prone to hallucinate edges not specified in the prompt's description of the graph. This phenomenon persists across multiple problem complexity levels and semantic frames, and it appears to account for a significant fraction of the incorrect answers from every tested model, and the vast majority of them for some models. Our results indicate that RLLMs may possess broader issues with misrepresentation of problem specifics, and we offer suggestions for design choices to mitigate this weakness.
Authors: Yue Deng, Asadullah Hill Galib, Xin Lan, Pang-Ning Tan, Lifeng Luo
Abstract: Deep learning-based weather forecasting models have recently demonstrated significant performance improvements over gold-standard physics-based simulation tools. However, these models are vulnerable to adversarial attacks, which raises concerns about their trustworthiness. In this paper, we first investigate the feasibility of applying existing adversarial attack methods to weather forecasting models. We argue that a successful attack should (1) not modify significantly its original inputs, (2) be faithful, i.e., achieve the desired forecast at targeted locations with minimal changes to non-targeted locations, and (3) be geospatio-temporally realistic. However, balancing these criteria is a challenge as existing methods are not designed to preserve the geospatio-temporal dependencies of the original samples. To address this challenge, we propose a novel framework called FABLE (Forecast Alteration By Localized targeted advErsarial attack), which employs a 3D discrete wavelet decomposition to extract the varying components of the geospatio-temporal data. By regulating the magnitude of adversarial perturbations across different components, FABLE can generate adversarial inputs that maintain geospatio-temporal coherence while remaining faithful and closely aligned with the original inputs. Experimental results on multiple real-world datasets demonstrate the effectiveness of our framework over baseline methods across various metrics.
Authors: Jared Boyer, T. Konstantin Rusch, Daniela Rus
Abstract: State-space models (SSMs) are a class of networks for sequence learning that benefit from fixed state size and linear complexity with respect to sequence length, contrasting the quadratic scaling of typical attention mechanisms. Inspired from observations in neuroscience, Linear Oscillatory State-Space models (LinOSS) are a recently proposed class of SSMs constructed from layers of discretized forced harmonic oscillators. Although these models perform competitively, leveraging fast parallel scans over diagonal recurrent matrices and achieving state-of-the-art performance on tasks with sequence length up to 50k, LinOSS models rely on rigid energy dissipation ("forgetting") mechanisms that are inherently coupled to the timescale of state evolution. As forgetting is a crucial mechanism for long-range reasoning, we demonstrate the representational limitations of these models and introduce Damped Linear Oscillatory State-Space models (D-LinOSS), a more general class of oscillatory SSMs that learn to dissipate latent state energy on multiple timescales. We analyze the spectral distribution of the model's recurrent matrices and prove that the SSM layers exhibit stable dynamics under simple, flexible parameterizations. D-LinOSS consistently outperforms previous LinOSS methods on long-range learning tasks, without introducing additional complexity, and simultaneously reduces the hyperparameter search space by 50%.
Authors: Yuhui Wang, Rongyi Zhu, Ting Wang
Abstract: Harmful fine-tuning attacks pose a major threat to the security of large language models (LLMs), allowing adversaries to compromise safety guardrails with minimal harmful data. While existing defenses attempt to reinforce LLM alignment, they fail to address models' inherent "trainability" on harmful data, leaving them vulnerable to stronger attacks with increased learning rates or larger harmful datasets. To overcome this critical limitation, we introduce SEAM, a novel alignment-enhancing defense that transforms LLMs into self-destructive models with intrinsic resilience to misalignment attempts. Specifically, these models retain their capabilities for legitimate tasks while exhibiting substantial performance degradation when fine-tuned on harmful data. The protection is achieved through a novel loss function that couples the optimization trajectories of benign and harmful data, enhanced with adversarial gradient ascent to amplify the self-destructive effect. To enable practical training, we develop an efficient Hessian-free gradient estimate with theoretical error bounds. Extensive evaluation across LLMs and datasets demonstrates that SEAM creates a no-win situation for adversaries: the self-destructive models achieve state-of-the-art robustness against low-intensity attacks and undergo catastrophic performance collapse under high-intensity attacks, rendering them effectively unusable. (warning: this paper contains potentially harmful content generated by LLMs.)
Authors: Riad Hossain, Muhammad Ashad Kabir, Arat Ibne Golam Mowla, Animesh Chandra Roy, Ranjit Kumar Ghosh
Abstract: Parkinson's disease (PD) poses a growing global health challenge, with Bangladesh experiencing a notable rise in PD-related mortality. Early detection of PD remains particularly challenging in resource-constrained settings, where voice-based analysis has emerged as a promising non-invasive and cost-effective alternative. However, existing studies predominantly focus on English or other major languages; notably, no voice dataset for PD exists for Bengali - posing a significant barrier to culturally inclusive and accessible healthcare solutions. Moreover, most prior studies employed only a narrow set of acoustic features, with limited or no hyperparameter tuning and feature selection strategies, and little attention to model explainability. This restricts the development of a robust and generalizable machine learning model. To address this gap, we present BenSparX, the first Bengali conversational speech dataset for PD detection, along with a robust and explainable machine learning framework tailored for early diagnosis. The proposed framework incorporates diverse acoustic feature categories, systematic feature selection methods, and state-of-the-art machine learning algorithms with extensive hyperparameter optimization. Furthermore, to enhance interpretability and trust in model predictions, the framework incorporates SHAP (SHapley Additive exPlanations) analysis to quantify the contribution of individual acoustic features toward PD detection. Our framework achieves state-of-the-art performance, yielding an accuracy of 95.77%, F1 score of 95.57%, and AUC-ROC of 0.982. We further externally validated our approach by applying the framework to existing PD datasets in other languages, where it consistently outperforms state-of-the-art approaches. To facilitate further research and reproducibility, the dataset has been made publicly available at https://github.com/Riad071/BenSParX.
Authors: Zhenghao Li, Shengbo Wang, Nian Si
Abstract: Distributionally robust reinforcement learning (DR-RL) has recently gained significant attention as a principled approach that addresses discrepancies between training and testing environments. To balance robustness, conservatism, and computational traceability, the literature has introduced DR-RL models with SA-rectangular and S-rectangular adversaries. While most existing statistical analyses focus on SA-rectangular models, owing to their algorithmic simplicity and the optimality of deterministic policies, S-rectangular models more accurately capture distributional discrepancies in many real-world applications and often yield more effective robust randomized policies. In this paper, we study the empirical value iteration algorithm for divergence-based S-rectangular DR-RL and establish near-optimal sample complexity bounds of $\widetilde{O}(|\mathcal{S}||\mathcal{A}|(1-\gamma)^{-4}\varepsilon^{-2})$, where $\varepsilon$ is the target accuracy, $|\mathcal{S}|$ and $|\mathcal{A}|$ denote the cardinalities of the state and action spaces, and $\gamma$ is the discount factor. To the best of our knowledge, these are the first sample complexity results for divergence-based S-rectangular models that achieve optimal dependence on $|\mathcal{S}|$, $|\mathcal{A}|$, and $\varepsilon$ simultaneously. We further validate this theoretical dependence through numerical experiments on a robust inventory control problem and a theoretical worst-case example, demonstrating the fast learning performance of our proposed algorithm.
Authors: Shuo Han, German Espinosa, Junda Huang, Daniel A. Dombeck, Malcolm A. MacIver, Bradly C. Stadie
Abstract: Recent advances in reinforcement learning (RL) have demonstrated impressive capabilities in complex decision-making tasks. This progress raises a natural question: how do these artificial systems compare to biological agents, which have been shaped by millions of years of evolution? To help answer this question, we undertake a comparative study of biological mice and RL agents in a predator-avoidance maze environment. Through this analysis, we identify a striking disparity: RL agents consistently demonstrate a lack of self-preservation instinct, readily risking ``death'' for marginal efficiency gains. These risk-taking strategies are in contrast to biological agents, which exhibit sophisticated risk-assessment and avoidance behaviors. Towards bridging this gap between the biological and artificial, we propose two novel mechanisms that encourage more naturalistic risk-avoidance behaviors in RL agents. Our approach leads to the emergence of naturalistic behaviors, including strategic environment assessment, cautious path planning, and predator avoidance patterns that closely mirror those observed in biological systems.
Authors: Wenhui Liu, Zhijian Wu, Jingchao Wang, Dingjiang Huang, Shuigeng Zhou
Abstract: Offline reinforcement learning seeks to derive improved policies entirely from historical data but often struggles with over-optimistic value estimates for out-of-distribution (OOD) actions. This issue is typically mitigated via policy constraint or conservative value regularization methods. However, these approaches may impose overly constraints or biased value estimates, potentially limiting performance improvements. To balance exploitation and restriction, we propose an Imagination-Limited Q-learning (ILQ) method, which aims to maintain the optimism that OOD actions deserve within appropriate limits. Specifically, we utilize the dynamics model to imagine OOD action-values, and then clip the imagined values with the maximum behavior values. Such design maintains reasonable evaluation of OOD actions to the furthest extent, while avoiding its over-optimism. Theoretically, we prove the convergence of the proposed ILQ under tabular Markov decision processes. Particularly, we demonstrate that the error bound between estimated values and optimality values of OOD state-actions possesses the same magnitude as that of in-distribution ones, thereby indicating that the bias in value estimates is effectively mitigated. Empirically, our method achieves state-of-the-art performance on a wide range of tasks in the D4RL benchmark.
Authors: Yuhan Zhang (Ret.), Yishu Wei (Ret.), Yanshan Wang (Ret.), Yunyu Xiao (Ret.), COL (Ret.), Ronald K. Poropatich, Gretchen L. Haas, Yiye Zhang, Chunhua Weng, Jinze Liu, Lisa A. Brenner, James M. Bjork, Yifan Peng
Abstract: Suicide remains one of the main preventable causes of death among active service members and veterans. Early detection and prediction are crucial in suicide prevention. Machine learning techniques have yielded promising results in this area recently. This study aims to assess and summarize current research and provides a comprehensive review regarding the application of machine learning techniques in assessing and predicting suicidal ideation, attempts, and mortality among members of military and veteran populations. A keyword search using PubMed, IEEE, ACM, and Google Scholar was conducted, and the PRISMA protocol was adopted for relevant study selection. Thirty-two articles met the inclusion criteria. These studies consistently identified risk factors relevant to mental health issues such as depression, post-traumatic stress disorder (PTSD), suicidal ideation, prior attempts, physical health problems, and demographic characteristics. Machine learning models applied in this area have demonstrated reasonable predictive accuracy. However, additional research gaps still exist. First, many studies have overlooked metrics that distinguish between false positives and negatives, such as positive predictive value and negative predictive value, which are crucial in the context of suicide prevention policies. Second, more dedicated approaches to handling survival and longitudinal data should be explored. Lastly, most studies focused on machine learning methods, with limited discussion of their connection to clinical rationales. In summary, machine learning analyses have identified a wide range of risk factors associated with suicide in military populations. The diversity and complexity of these factors also demonstrates that effective prevention strategies must be comprehensive and flexible.
Authors: Jizhou Guo, Zhaomin Wu, Philip S. Yu
Abstract: High-quality reward models are crucial for unlocking the reasoning potential of large language models (LLMs), with best-of-N voting demonstrating significant performance gains. However, current reward models, which typically operate on the textual output of LLMs, are computationally expensive and parameter-heavy, limiting their real-world applications. We introduce the Efficient Linear Hidden State Reward (ELHSR) model - a novel, highly parameter-efficient approach that leverages the rich information embedded in LLM hidden states to address these issues. ELHSR systematically outperform baselines with less than 0.005% of the parameters of baselines, requiring only a few samples for training. ELHSR also achieves orders-of-magnitude efficiency improvement with significantly less time and fewer FLOPs per sample than baseline reward models. Moreover, ELHSR exhibits robust performance even when trained only on logits, extending its applicability to some closed-source LLMs. In addition, ELHSR can also be combined with traditional reward models to achieve additional performance gains.
Authors: Jianheng Tang, Huiping Zhuang, Di Fang, Jiaxu Li, Feijiang Han, Yajiang Huang, Kejia Fan, Leye Wang, Zhanxing Zhu, Shanghang Zhang, Houbing Herbert Song, Yunhuai Liu
Abstract: The development of artificial intelligence demands that models incrementally update knowledge by Continual Learning (CL) to adapt to open-world environments. To meet privacy and security requirements, Continual Unlearning (CU) emerges as an important problem, aiming to sequentially forget particular knowledge acquired during the CL phase. However, existing unlearning methods primarily focus on single-shot joint forgetting and face significant limitations when applied to CU. First, most existing methods require access to the retained dataset for re-training or fine-tuning, violating the inherent constraint in CL that historical data cannot be revisited. Second, these methods often suffer from a poor trade-off between system efficiency and model fidelity, making them vulnerable to being overwhelmed or degraded by adversaries through deliberately frequent requests. In this paper, we identify that the limitations of existing unlearning methods stem fundamentally from their reliance on gradient-based updates. To bridge the research gap at its root, we propose a novel gradient-free method for CU, named Analytic Continual Unlearning (ACU), for efficient and exact forgetting with historical data privacy preservation. In response to each unlearning request, our ACU recursively derives an analytical (i.e., closed-form) solution in an interpretable manner using the least squares method. Theoretical and experimental evaluations validate the superiority of our ACU on unlearning effectiveness, model fidelity, and system efficiency.
Authors: Jianheng Tang, Huiping Zhuang, Jingyu He, Run He, Jingchao Wang, Kejia Fan, Anfeng Liu, Tian Wang, Leye Wang, Zhanxing Zhu, Shanghang Zhang, Houbing Herbert Song, Yunhuai Liu
Abstract: Federated Continual Learning (FCL) enables distributed clients to collaboratively train a global model from online task streams in dynamic real-world scenarios. However, existing FCL methods face challenges of both spatial data heterogeneity among distributed clients and temporal data heterogeneity across online tasks. Such data heterogeneity significantly degrades the model performance with severe spatial-temporal catastrophic forgetting of local and past knowledge. In this paper, we identify that the root cause of this issue lies in the inherent vulnerability and sensitivity of gradients to non-IID data. To fundamentally address this issue, we propose a gradient-free method, named Analytic Federated Continual Learning (AFCL), by deriving analytical (i.e., closed-form) solutions from frozen extracted features. In local training, our AFCL enables single-epoch learning with only a lightweight forward-propagation process for each client. In global aggregation, the server can recursively and efficiently update the global model with single-round aggregation. Theoretical analyses validate that our AFCL achieves spatio-temporal invariance of non-IID data. This ideal property implies that, regardless of how heterogeneous the data are distributed across local clients and online tasks, the aggregated model of our AFCL remains invariant and identical to that of centralized joint learning. Extensive experiments show the consistent superiority of our AFCL over state-of-the-art baselines across various benchmark datasets and settings.
Authors: Yuhan Guo, Lizhong Ding, Yuwan Yang, Xuewei Guo
Abstract: Kernelized attention extends the attention mechanism by modeling sequence correlations through kernel functions, making significant progresses in optimizing attention. Under the guarantee of harmonic analysis theory, kernel functions can be expanded with basis functions, inspiring random feature-based approaches to enhance the efficiency of kernelized attention while maintaining predictive performance. However, current random feature-based works are limited to the Fourier basis expansions under Bochner's theorem. We propose Schoenberg's theorem-based attention (SchoenbAt), which approximates dot-product kernelized attention with the polynomial basis under Schoenberg's theorem via random Maclaurin features and applies a two-stage regularization to constrain the input space and restore the output scale, acting as a drop-in replacement of dot-product kernelized attention. Our theoretical proof of the unbiasedness and concentration error bound of SchoenbAt supports its efficiency and accuracy as a kernelized attention approximation, which is also empirically validated under various random feature dimensions. Evaluations on real-world datasets demonstrate that SchoenbAt significantly enhances computational speed while preserving competitive performance in terms of precision, outperforming several efficient attention methods.
Authors: Wen-Chao Hu, Qi-Jie Li, Lin-Han Jia, Cunjing Ge, Yu-Feng Li, Yuan Jiang, Zhi-Hua Zhou
Abstract: Abductive Learning (ABL) integrates machine learning with logical reasoning in a loop: a learning model predicts symbolic concept labels from raw inputs, which are revised through abduction using domain knowledge and then fed back for retraining. However, due to the nondeterminism of abduction, the training process often suffers from instability, especially when the knowledge base is large and complex, resulting in a prohibitively large abduction space. While prior works focus on improving candidate selection within this space, they typically treat the knowledge base as a static black box. In this work, we propose Curriculum Abductive Learning (C-ABL), a method that explicitly leverages the internal structure of the knowledge base to address the ABL training challenges. C-ABL partitions the knowledge base into a sequence of sub-bases, progressively introduced during training. This reduces the abduction space throughout training and enables the model to incorporate logic in a stepwise, smooth way. Experiments across multiple tasks show that C-ABL outperforms previous ABL implementations, significantly improves training stability, convergence speed, and final accuracy, especially under complex knowledge setting.
Authors: Zhen Zhao, Wenqi Huang, Zicheng Wang, Jiaxuan Hou, Peng Li, Lei Bai
Abstract: Power flow estimation plays a vital role in ensuring the stability and reliability of electrical power systems, particularly in the context of growing network complexities and renewable energy integration. However, existing studies often fail to adequately address the unique characteristics of power systems, such as the sparsity of network connections and the critical importance of the unique Slack node, which poses significant challenges in achieving high-accuracy estimations. In this paper, we present SenseFlow, a novel physics-informed and self-ensembling iterative framework that integrates two main designs, the Physics-Informed Power Flow Network (FlowNet) and Self-Ensembling Iterative Estimation (SeIter), to carefully address the unique properties of the power system and thereby enhance the power flow estimation. Specifically, SenseFlow enforces the FlowNet to gradually predict high-precision voltage magnitudes and phase angles through the iterative SeIter process. On the one hand, FlowNet employs the Virtual Node Attention and Slack-Gated Feed-Forward modules to facilitate efficient global-local communication in the face of network sparsity and amplify the influence of the Slack node on angle predictions, respectively. On the other hand, SeIter maintains an exponential moving average of FlowNet's parameters to create a robust ensemble model that refines power state predictions throughout the iterative fitting process. Experimental results demonstrate that SenseFlow outperforms existing methods, providing a promising solution for high-accuracy power flow estimation across diverse grid configurations.
Authors: Feng Yu, Jia Hu, Geyong Min
Abstract: Federated Parameter-Efficient Fine-Tuning (FedPEFT) reduces communication and computation costs in federated fine-tuning of pre-trained models by updating only a small subset of model parameters. However, existing approaches assume static data distributions, failing to adequately address real-world scenarios where new classes continually emerge, particularly in Federated Class Incremental Learning (FCIL). FCIL faces two key challenges: catastrophic forgetting and performance degradation caused by non-IID data across clients. Unlike current methods that maintain separate task-specific components or suffer from aggregation noise during parameter aggregation, we propose Federated Task-agnostic Low-rank Residual Adaptation (Fed-TaLoRA), a novel parameter-efficient approach for fine-tuning in resource-constrained FCIL scenarios. Specifically, we fine-tune only shared task-agnostic LoRA parameters across sequential tasks, effectively mitigating catastrophic forgetting while enabling efficient knowledge transfer among clients. Based on a theoretical analysis of aggregation, we develop a novel residual weight update mechanism that ensures accurate knowledge consolidation with minimal overhead. Our methodological innovations are attributed to three key strategies: task-agnostic adaptation, post-aggregation model calibration, and strategic placement of LoRA modules. Extensive experiments on multiple benchmark datasets demonstrate that Fed-TaLoRA consistently outperforms state-of-the-art methods in diverse data heterogeneity scenarios while substantially reducing resource requirements.
Authors: Ali Gholamzadeh, Noor Sajid
Abstract: Foundation models have demonstrated remarkable performance across modalities such as language and vision. However, model reuse across distinct modalities (e.g., text and vision) remains limited due to the difficulty of aligning internal representations. Existing methods require extensive paired training data or are constrained to specific domains. We introduce a semi-supervised approach for model alignment via conditional flow matching. The conditional flow between latent spaces of different modalities (e.g., text-to-image or biological-to-artificial neuronal activity) can be learned in two settings: ($1$) solving a (balanced or unbalanced) optimal transport problem with an inter-space bridge cost, and ($2$) performing memory-efficient alignment using labelled exemplars. Despite being constrained by the original models' capacity, our method--under both settings--matches downstream task performance of end-to-end trained models on object recognition and image generation tasks across MNIST, ImageNet, and \cite{majaj2015simple} datasets, particularly when labelled training data is scarce ($<20\%$). Our method provides a data-efficient solution for inter-modal model alignment with minimal supervision.
Authors: Mohit Kataria, Nikita Malik, Sandeep Kumar, Jayadeva
Abstract: Graph structure learning is a core problem in graph-based machine learning, essential for uncovering latent relationships and ensuring model interpretability. However, most existing approaches are ill-suited for large-scale and dynamically evolving graphs, as they often require complete re-learning of the structure upon the arrival of new nodes and incur substantial computational and memory costs. In this work, we propose GraphFLEx: a unified and scalable framework for Graph Structure Learning in Large and Expanding Graphs. GraphFLEx mitigates the scalability bottlenecks by restricting edge formation to structurally relevant subsets of nodes identified through a combination of clustering and coarsening techniques. This dramatically reduces the search space and enables efficient, incremental graph updates. The framework supports 48 flexible configurations by integrating diverse choices of learning paradigms, coarsening strategies, and clustering methods, making it adaptable to a wide range of graph settings and learning objectives. Extensive experiments across 26 diverse datasets and Graph Neural Network architectures demonstrate that GraphFLEx achieves state-of-the-art performance with significantly improved scalability.
Authors: Chaolong Ying, Yingqi Ruan, Xuemin Chen, Yaomin Wang, Tianshu Yu
Abstract: The Maximum Common Edge Subgraph (MCES) problem is a crucial challenge with significant implications in domains such as biology and chemistry. Traditional approaches, which include transformations into max-clique and search-based algorithms, suffer from scalability issues when dealing with larger instances. This paper introduces ``Neural Graduated Assignment'' (NGA), a simple, scalable, unsupervised-training-based method that addresses these limitations by drawing inspiration from the classical Graduated Assignment (GA) technique. Central to NGA is stacking of neural components that closely resemble the GA process, but with the reparameterization of learnable temperature into higher dimension. We further theoretically analyze the learning dynamics of NGA, showing its design leads to fast convergence, better exploration-exploitation tradeoff, and ability to escape local optima. Extensive experiments across MCES computation, graph similarity estimation, and graph retrieval tasks reveal that NGA not only significantly improves computation time and scalability on large instances but also enhances performance compared to existing methodologies. The introduction of NGA marks a significant advancement in the computation of MCES and offers insights into other assignment problems.
Authors: Kai Tang, Jinhao You, Xiuqi Ge, Hanze Li, Yichen Guo, Xiande Huang
Abstract: Despite the impressive capabilities of Large Vision-Language Models (LVLMs), they remain susceptible to hallucinations-generating content that is inconsistent with the input image. Existing training-free hallucination mitigation methods often suffer from unstable performance and high sensitivity to hyperparameter settings, limiting their practicality and broader adoption. In this paper, we propose a novel decoding mechanism, Decoding with Inter-layer Consistency via Layer Aggregation (DCLA), which requires no retraining, fine-tuning, or access to external knowledge bases. Specifically, our approach constructs a dynamic semantic reference by aggregating representations from previous layers, and corrects semantically deviated layers to enforce inter-layer consistency. The method allows DCLA to robustly mitigate hallucinations across multiple LVLMs. Experiments on hallucination benchmarks such as MME and POPE demonstrate that DCLA effectively reduces hallucinations while enhancing the reliability and performance of LVLMs.
Authors: Han Wang
Abstract: The intensive care unit (ICU) manages critically ill patients, many of whom face a high risk of mortality. Early and accurate prediction of in-hospital mortality within the first 24 hours of ICU admission is crucial for timely clinical interventions, resource optimization, and improved patient outcomes. Traditional scoring systems, while useful, often have limitations in predictive accuracy and adaptability. Objective: This review aims to systematically evaluate and benchmark innovative methodologies that leverage data available within the first day of ICU admission for predicting in-hospital mortality. We focus on advancements in machine learning, novel biomarker applications, and the integration of diverse data types.
Authors: Georgiy Malaniya, Anton Bolychev, Grigory Yaremenko, Anastasia Krasnaya, Pavel Osinenko
Abstract: We introduce Multi-CALF, an algorithm that intelligently combines reinforcement learning policies based on their relative value improvements. Our approach integrates a standard RL policy with a theoretically-backed alternative policy, inheriting formal stability guarantees while often achieving better performance than either policy individually. We prove that our combined policy converges to a specified goal set with known probability and provide precise bounds on maximum deviation and convergence time. Empirical validation on control tasks demonstrates enhanced performance while maintaining stability guarantees.
Authors: Prakash Palanivelu Rajmohan, Fred Roosta
Abstract: While norm-based and leverage-score-based methods have been extensively studied for identifying "important" data points in linear models, analogous tools for nonlinear models remain significantly underdeveloped. By introducing the concept of the adjoint operator of a nonlinear map, we address this gap and generalize norm-based and leverage-score-based importance sampling to nonlinear settings. We demonstrate that sampling based on these generalized notions of norm and leverage scores provides approximation guarantees for the underlying nonlinear mapping, similar to linear subspace embeddings. As direct applications, these nonlinear scores not only reduce the computational complexity of training nonlinear models by enabling efficient sampling over large datasets but also offer a novel mechanism for model explainability and outlier detection. Our contributions are supported by both theoretical analyses and experimental results across a variety of supervised learning scenarios.
Authors: Anton Bolychev, Georgiy Malaniya, Grigory Yaremenko, Anastasia Krasnaya, Pavel Osinenko
Abstract: We introduce a universal policy wrapper for reinforcement learning agents that ensures formal goal-reaching guarantees. In contrast to standard reinforcement learning algorithms that excel in performance but lack rigorous safety assurances, our wrapper selectively switches between a high-performing base policy -- derived from any existing RL method -- and a fallback policy with known convergence properties. Base policy's value function supervises this switching process, determining when the fallback policy should override the base policy to ensure the system remains on a stable path. The analysis proves that our wrapper inherits the fallback policy's goal-reaching guarantees while preserving or improving upon the performance of the base policy. Notably, it operates without needing additional system knowledge or online constrained optimization, making it readily deployable across diverse reinforcement learning architectures and tasks.
Authors: Abrar Rahman Abir, Haz Sameen Shahgir, Md Rownok Zahan Ratul, Md Toki Tahmid, Greg Ver Steeg, Yue Dong
Abstract: Complementarity Determining Regions (CDRs) are critical segments of an antibody that facilitate binding to specific antigens. Current computational methods for CDR design utilize reconstruction losses and do not jointly optimize binding energy, a crucial metric for antibody efficacy. Rather, binding energy optimization is done through computationally expensive Online Reinforcement Learning (RL) pipelines rely heavily on unreliable binding energy estimators. In this paper, we propose AbFlowNet, a novel generative framework that integrates GFlowNet with Diffusion models. By framing each diffusion step as a state in the GFlowNet framework, AbFlowNet jointly optimizes standard diffusion losses and binding energy by directly incorporating energy signals into the training process, thereby unifying diffusion and reward optimization in a single procedure. Experimental results show that AbFlowNet outperforms the base diffusion model by 3.06% in amino acid recovery, 20.40% in geometric reconstruction (RMSD), and 3.60% in binding energy improvement ratio. ABFlowNet also decreases Top-1 total energy and binding energy errors by 24.8% and 38.1% without pseudo-labeling the test dataset or using computationally expensive online RL regimes.
Authors: Yichen Guo, Hanze Li, Zonghao Zhang, Jinhao You, Kai Tang, Xiande Huang
Abstract: Although large vision-language models (LVLMs) leverage rich visual token representations to achieve strong performance on multimodal tasks, these tokens also introduce significant computational overhead during inference. Existing training-free token pruning methods typically adopt a single-stage strategy, focusing either on visual self-attention or visual-textual cross-attention. However, such localized perspectives often overlook the broader information flow across the model, leading to substantial performance degradation, especially under high pruning ratios. In this work, we propose STAR (Stage-wise Attention-guided token Reduction), a training-free, plug-and-play framework that approaches token pruning from a global perspective. Instead of pruning at a single point, STAR performs attention-guided reduction in two complementary stages: an early-stage pruning based on visual self-attention to remove redundant low-level features, and a later-stage pruning guided by cross-modal attention to discard task-irrelevant tokens. This holistic approach allows STAR to significantly reduce computational cost while better preserving task-critical information. Extensive experiments across multiple LVLM architectures and benchmarks show that STAR achieves strong acceleration while maintaining comparable, and in some cases even improved performance.
Authors: Gang Li, Ming Lin, Tomer Galanti, Zhengzhong Tu, Tianbao Yang
Abstract: The recent success and openness of DeepSeek-R1 have brought widespread attention to Group Relative Policy Optimization (GRPO) as a reinforcement learning method for large reasoning models (LRMs). In this work, we analyze the GRPO objective under a binary reward setting and reveal an inherent limitation of question-level difficulty bias. We also identify a connection between GRPO and traditional discriminative methods in supervised learning. Motivated by these insights, we introduce a new Discriminative Constrained Optimization (DisCO) framework for reinforcing LRMs, grounded in the principle of discriminative learning. The main differences between DisCO and GRPO and its recent variants are: (1) it replaces the group relative objective with a discriminative objective defined by a scoring function; (2) it abandons clipping-based surrogates in favor of non-clipping RL surrogate objectives used as scoring functions; (3) it employs a simple yet effective constrained optimization approach to enforce the KL divergence constraint, ensuring stable training. As a result, DisCO offers notable advantages over GRPO and its variants: (i) it completely eliminates difficulty bias by adopting discriminative objectives; (ii) it addresses the entropy instability in GRPO and its variants through the use of non-clipping scoring functions and a constrained optimization approach; (iii) it allows the incorporation of advanced discriminative learning techniques to address data imbalance, where a significant number of questions have more negative than positive generated answers during training. Our experiments on enhancing the mathematical reasoning capabilities of SFT-finetuned models show that DisCO significantly outperforms GRPO and its improved variants such as DAPO, achieving average gains of 7\% over GRPO and 6\% over DAPO across six benchmark tasks for an 1.5B model.
Authors: Han Weng, Boyi Liu, Yuanfeng Song, Dun Zeng, Yingxiang Yang, Yi Zhan, Longjie Cui, Xiaoming Yin, Yang Sun
Abstract: Reinforcement learning (RL) has been widely adopted to enhance the performance of large language models (LLMs) on Text-to-SQL tasks. However, existing methods often rely on execution-based or LLM-based Bradley-Terry reward models. The former suffers from high execution latency caused by repeated database calls, whereas the latter imposes substantial GPU memory overhead, both of which significantly hinder the efficiency and scalability of RL pipelines. To this end, we propose a novel Text-to-SQL RL fine-tuning framework named Graph-Reward-SQL, which employs the GMNScore outcome reward model. We leverage SQL graph representations to provide accurate reward signals while significantly reducing inference time and GPU memory usage. Building on this foundation, we further introduce StepRTM, a stepwise reward model that provides intermediate supervision over Common Table Expression (CTE) subqueries. This encourages both functional correctness and structural clarity of SQL. Extensive comparative and ablation experiments on standard benchmarks, including Spider and BIRD, demonstrate that our method consistently outperforms existing reward models.
Authors: Liu Ziyin, Yizhou Xu, Isaac Chuang
Abstract: With the rapid discovery of emergent phenomena in deep learning and large language models, explaining and understanding their cause has become an urgent need. Here, we propose a rigorous entropic-force theory for understanding the learning dynamics of neural networks trained with stochastic gradient descent (SGD) and its variants. Building on the theory of parameter symmetries and an entropic loss landscape, we show that representation learning is crucially governed by emergent entropic forces arising from stochasticity and discrete-time updates. These forces systematically break continuous parameter symmetries and preserve discrete ones, leading to a series of gradient balance phenomena that resemble the equipartition property of thermal systems. These phenomena, in turn, (a) explain the universal alignment of neural representations between AI models and lead to a proof of the Platonic Representation Hypothesis, and (b) reconcile the seemingly contradictory observations of sharpness- and flatness-seeking behavior of deep learning optimization. Our theory and experiments demonstrate that a combination of entropic forces and symmetry breaking is key to understanding emergent phenomena in deep learning.
Authors: Su Yeong Jo, Sanghyeon Park, Seungchan Ko, Jongcheon Park, Hosung Kim, Sangseung Lee, Joongoo Jeon
Abstract: The Saint-Venant torsion theory is a classical theory for analyzing the torsional behavior of structural components, and it remains critically important in modern computational design workflows. Conventional numerical methods, including the finite element method (FEM), typically rely on mesh-based approaches to obtain approximate solutions. However, these methods often require complex and computationally intensive techniques to overcome the limitations of approximation, leading to significant increases in computational cost. The objective of this study is to develop a series of novel numerical methods based on physics-informed neural networks (PINN) for solving the Saint-Venant torsion equations. Utilizing the expressive power and the automatic differentiation capability of neural networks, the PINN can solve partial differential equations (PDEs) along with boundary conditions without the need for intricate computational techniques. First, a PINN solver was developed to compute the torsional constant for bars with arbitrary cross-sectional geometries. This was followed by the development of a solver capable of handling cases with sharp geometric transitions; variable-scaling PINN (VS-PINN). Finally, a parametric PINN was constructed to address the limitations of conventional single-instance PINN. The results from all three solvers showed good agreement with reference solutions, demonstrating their accuracy and robustness. Each solver can be selectively utilized depending on the specific requirements of torsional behavior analysis.
Authors: Udaya Shreyas, L. N. Aadarsh
Abstract: Image Generation models are a trending topic nowadays, with many people utilizing Artificial Intelligence models in order to generate images. There are many such models which, given a prompt of a text, will generate an image which depicts said prompt. There are many image generation models, such as Latent Diffusion Models, Denoising Diffusion Probabilistic Models, Generative Adversarial Networks and many more. When generating images, these models can generate sensitive image data, which can be threatening to privacy or may violate copyright laws of private entities. Machine unlearning aims at removing the influence of specific data subsets from the trained models and in the case of image generation models, remove the influence of a concept such that the model is unable to generate said images of the concept when prompted. Conventional retraining of the model can take upto days, hence fast algorithms are the need of the hour. In this paper we propose an algorithm that aims to remove the influence of concepts in diffusion models through updating the gradients of the final layers of the text encoders. Using a weighted loss function, we utilize backpropagation in order to update the weights of the final layers of the Text Encoder componet of the Stable Diffusion Model, removing influence of the concept from the text-image embedding space, such that when prompted, the result is an image not containing the concept. The weighted loss function makes use of Textual Inversion and Low-Rank Adaptation.We perform our experiments on Latent Diffusion Models, namely the Stable Diffusion v2 model, with an average concept unlearning runtime of 50 seconds using 4-5 images.
Authors: Piotr Pi\k{e}kos, Subhradeep Kayal, Alexandros Karatzoglou
Abstract: Hierarchical data arise in countless domains, from biological taxonomies and organizational charts to legal codes and knowledge graphs. Residual Quantization (RQ) is widely used to generate discrete, multitoken representations for such data by iteratively quantizing residuals in a multilevel codebook. However, its reliance on Euclidean geometry can introduce fundamental mismatches that hinder modeling of hierarchical branching, necessary for faithful representation of hierarchical data. In this work, we propose Hyperbolic Residual Quantization (HRQ), which embeds data natively in a hyperbolic manifold and performs residual quantization using hyperbolic operations and distance metrics. By adapting the embedding network, residual computation, and distance metric to hyperbolic geometry, HRQ imparts an inductive bias that aligns naturally with hierarchical branching. We claim that HRQ in comparison to RQ can generate more useful for downstream tasks discrete hierarchical representations for data with latent hierarchies. We evaluate HRQ on two tasks: supervised hierarchy modeling using WordNet hypernym trees, where the model is supervised to learn the latent hierarchy - and hierarchy discovery, where, while latent hierarchy exists in the data, the model is not directly trained or evaluated on a task related to the hierarchy. Across both scenarios, HRQ hierarchical tokens yield better performance on downstream tasks compared to Euclidean RQ with gains of up to $20\%$ for the hierarchy modeling task. Our results demonstrate that integrating hyperbolic geometry into discrete representation learning substantially enhances the ability to capture latent hierarchies.
Authors: Harel Mendelman, Haggai Maron, Ronen Talmon
Abstract: Graph Neural Networks (GNNs) excel at analyzing graph-structured data but struggle on heterophilic graphs, where connected nodes often belong to different classes. While this challenge is commonly addressed with specialized GNN architectures, graph rewiring remains an underexplored strategy in this context. We provide theoretical foundations linking edge homophily, GNN embedding smoothness, and node classification performance, motivating the need to enhance homophily. Building on this insight, we introduce a rewiring framework that increases graph homophily using a reference graph, with theoretical guarantees on the homophily of the rewired graph. To broaden applicability, we propose a label-driven diffusion approach for constructing a homophilic reference graph from node features and training labels. Through extensive simulations, we analyze how the homophily of both the original and reference graphs influences the rewired graph homophily and downstream GNN performance. We evaluate our method on 11 real-world heterophilic datasets and show that it outperforms existing rewiring techniques and specialized GNNs for heterophilic graphs, achieving improved node classification accuracy while remaining efficient and scalable to large graphs.
Authors: Jiahan Zhang, Tao Luo, Yaoyu Zhang
Abstract: Understanding the convergence points and optimization landscape of neural networks is crucial, particularly for homogeneous networks where Karush-Kuhn-Tucker (KKT) points of the associated maximum-margin problem often characterize solutions. This paper investigates the relationship between such KKT points across networks of different widths generated via neuron splitting. We introduce and formalize the \textbf{KKT point embedding principle}, establishing that KKT points of a homogeneous network's max-margin problem ($P_{\Phi}$) can be embedded into the KKT points of a larger network's problem ($P_{\tilde{\Phi}}$) via specific linear isometric transformations corresponding to neuron splitting. We rigorously prove this principle holds for neuron splitting in both two-layer and deep homogeneous networks. Furthermore, we connect this static embedding to the dynamics of gradient flow training with smooth losses. We demonstrate that trajectories initiated from appropriately mapped points remain mapped throughout training and that the resulting $\omega$-limit sets of directions are correspondingly mapped ($T(L(\theta(0))) = L(\boldsymbol{\eta}(0))$), thereby preserving the alignment with KKT directions dynamically when directional convergence occurs. Our findings offer insights into the effects of network width, parameter redundancy, and the structural connections between solutions found via optimization in homogeneous networks of varying sizes.
Authors: Emanuele La Malfa, Jon Vadillo, Marco Molinari, Michael Wooldridge
Abstract: This paper introduces a formal notion of fixed point explanations, inspired by the "why regress" principle, to assess, through recursive applications, the stability of the interplay between a model and its explainer. Fixed point explanations satisfy properties like minimality, stability, and faithfulness, revealing hidden model behaviours and explanatory weaknesses. We define convergence conditions for several classes of explainers, from feature-based to mechanistic tools like Sparse AutoEncoders, and we report quantitative and qualitative results.
Authors: Rafael Florencio, Julio Guerrero
Abstract: Recently, innovative adaptations of the Ritz Method incorporating deep learning have been developed, known as the Deep Ritz Method. This approach employs a neural network as the test function for variational problems. However, the neural network does not inherently satisfy the boundary conditions of the variational problem. To resolve this issue, the Deep Ritz Method introduces a penalty term into the functional of the variational problem, which can lead to misleading results during the optimization process. In this work, an ansatz is proposed that inherently satisfies the boundary conditions of the variational problem. The results demonstrate that the proposed ansatz not only eliminates misleading outcomes but also reduces complexity while maintaining accuracy, showcasing its practical effectiveness in addressing variational problems.
Authors: Zirun Guo, Minjie Hong, Tao Jin
Abstract: Reinforcement Learning (RL) has shown promise in improving the reasoning abilities of Large Language Models (LLMs). However, the specific challenges of adapting RL to multimodal data and formats remain relatively unexplored. In this work, we present Observe-R1, a novel framework aimed at enhancing the reasoning capabilities of multimodal large language models (MLLMs). We draw inspirations from human learning progression--from simple to complex and easy to difficult, and propose a gradual learning paradigm for MLLMs. To this end, we construct the NeuraLadder dataset, which is organized and sampled according to the difficulty and complexity of data samples for RL training. To tackle multimodal tasks, we introduce a multimodal format constraint that encourages careful observation of images, resulting in enhanced visual abilities and clearer and more structured responses. Additionally, we implement a bonus reward system that favors concise, correct answers within a length constraint, alongside a dynamic weighting mechanism that prioritizes uncertain and medium-difficulty problems, ensuring that more informative samples have a greater impact on training. Our experiments with the Qwen2.5-VL-3B and Qwen2.5-VL-7B models on 20k samples from the NeuraLadder dataset show that Observe-R1 outperforms a series of larger reasoning models on both reasoning and general benchmarks, achieving superior clarity and conciseness in reasoning chains. Ablation studies validate the effectiveness of our strategies, highlighting the robustness and generalization of our approach. The dataset and code will be released at https://github.com/zrguo/Observe-R1.
Authors: Wenqiao Zhu, Ji Liu, Lulu Wang, Jun Wu, Yulun Zhang
Abstract: Direct Preference Optimization (DPO) is broadly utilized for aligning Large Language Models (LLMs) with human values because of its flexibility. Despite its effectiveness, it has been observed that the capability of DPO to generate human-preferred response is limited and the results of DPO are far from resilient. To address these limitations, in this paper we propose a novel Self-Guided Direct Preference Optimization algorithm, i.e., SGDPO, which incorporates a pilot term to steer the gradient flow during the optimization process, allowing for fine-grained control over the updates of chosen and rejected rewards. We provide a detailed theoretical analysis of our proposed method and elucidate its operational mechanism. Furthermore, we conduct comprehensive experiments on various models and benchmarks. The extensive experimental results demonstrate the consistency between the empirical results and our theoretical analysis and confirm the effectiveness of our proposed approach (up to 9.19% higher score).
Authors: Michele Fontanesi, Alessio Micheli, Marco Podda, Domenico Tortorella
Abstract: While Graph Neural Networks (GNNs) have become the de facto model for learning from structured data, their decisional process remains opaque to the end user, undermining their deployment in safety-critical applications. In the case of graph classification, Explainable Artificial Intelligence (XAI) techniques address this major issue by identifying sub-graph motifs that explain predictions. However, advancements in this field are hindered by a chronic scarcity of benchmark datasets with known ground-truth motifs to assess the explanations' quality. Current graph XAI benchmarks are limited to synthetic data or a handful of real-world tasks hand-curated by domain experts. In this paper, we propose a general method to automate the construction of XAI benchmarks for graph classification from real-world datasets. We provide both 15 ready-made benchmarks, as well as the code to generate more than 2000 additional XAI benchmarks with our method. As a use case, we employ our benchmarks to assess the effectiveness of some popular graph explainers.
Authors: Xin Yu, Yujia Wang, Jinghui Chen, Lingzhou Xue
Abstract: Low-Rank Adaptation (LoRA) has emerged as an effective technique for reducing memory overhead in fine-tuning large language models. However, it often suffers from sub-optimal performance compared with full fine-tuning since the update is constrained in the low-rank space. Recent variants such as LoRA-Pro attempt to mitigate this by adjusting the gradients of the low-rank matrices to approximate the full gradient. However, LoRA-Pro's solution is not unique, and different solutions can lead to significantly varying performance in ablation studies. Besides, to incorporate momentum or adaptive optimization design, approaches like LoRA-Pro must first compute the equivalent gradient, causing a higher memory cost close to full fine-tuning. A key challenge remains in integrating momentum properly into the low-rank space with lower memory cost. In this work, we propose AltLoRA, an alternating projection method that avoids the difficulties in gradient approximation brought by the joint update design, meanwhile integrating momentum without higher memory complexity. Our theoretical analysis provides convergence guarantees and further shows that AltLoRA enables stable feature learning and robustness to transformation invariance. Extensive experiments across multiple tasks demonstrate that AltLoRA outperforms LoRA and its variants, narrowing the gap toward full fine-tuning while preserving superior memory efficiency.
Authors: Yang Zhao, Kai Xiong, Xiao Ding, Li Du, YangouOuyang, Zhouhao Sun, Jiannan Guan, Wenbin Zhang, Bin Liu, Dong Hu, Bing Qin, Ting Liu
Abstract: Scaling RL for LLMs is computationally expensive, largely due to multi-sampling for policy optimization and evaluation, making efficient data selection crucial. Inspired by the Zone of Proximal Development (ZPD) theory, we hypothesize LLMs learn best from data within their potential comprehension zone. Addressing the limitation of conventional, computationally intensive multi-sampling methods for data assessment, we introduce UFO-RL. This novel framework uses a computationally efficient single-pass uncertainty estimation to identify informative data instances, achieving up to 185x faster data evaluation. UFO-RL leverages this metric to select data within the estimated ZPD for training. Experiments show that training with just 10% of data selected by UFO-RL yields performance comparable to or surpassing full-data training, reducing overall training time by up to 16x while enhancing stability and generalization. UFO-RL offers a practical and highly efficient strategy for scaling RL fine-tuning of LLMs by focusing learning on valuable data.
Authors: Asher Labovich
Abstract: Modern gradient-boosted decision trees (GBDTs) accelerate split finding with histogram-based binning, which reduces complexity from O(N) to O(B) given a fixed bin budget B. However, the predominant quantile binning strategy-designed to distribute data points evenly among bins-may overlook critical boundary values that could enhance predictive performance. In this work, we propose replacing quantile binning with a k-means discretizer initialized with quantile bins. We test this swap on 33 OpenML tasks plus synthetics that control for modality, skew, and bin budget. Across 18 regression datasets, k-means shows no statistically significant losses at the 5% level and wins in four cases-most strikingly a 55% MSE drop on one particularly skewed dataset-even though k-means' mean reciprocal rank (MRR) is slightly lower (0.65 vs 0.72). On the 15 classification datasets the two methods are statistically tied (MRR 0.70 vs 0.68) with gaps $\leq$0.2 pp. Synthetic experiments confirm consistently large MSE gains-typically >20% and rising to 90% as outlier magnitude increases or bin budget drops. We find that k-means keeps error on par with exact splitting when extra cuts add little value, yet still recovers key split points that quantile overlooks. As such, we advocate for a built-in bin_method=k-means flag, especially in regression tasks and in tight-budget settings such as the 32-64-bin GPU regime-because it is a "safe default" with large upside, yet adds only a one-off, cacheable overhead ($\approx$ 2s to bin 10M rows on one core).
Authors: Zachary Roch, Chi Zhang, George Atia, Yue Wang
Abstract: Robust reinforcement learning (RL) under the average-reward criterion is crucial for long-term decision making under potential environment mismatches, yet its finite-sample complexity study remains largely unexplored. Existing works offer algorithms with asymptotic guarantees, but the absence of finite-sample analysis hinders its principled understanding and practical deployment, especially in data-limited settings. We close this gap by proposing Robust Halpern Iteration (RHI), the first algorithm with provable finite-sample complexity guarantee. Under standard uncertainty sets -- including contamination sets and $\ell_p$-norm balls -- RHI attains an $\epsilon$-optimal policy with near-optimal sample complexity of $\tilde{\mathcal O}\left(\frac{SA\mathcal H^{2}}{\epsilon^{2}}\right)$, where $S$ and $A$ denote the numbers of states and actions, and $\mathcal H$ is the robust optimal bias span. This result gives the first polynomial sample complexity guarantee for robust average-reward RL. Moreover, our RHI's independence from prior knowledge distinguishes it from many previous average-reward RL studies. Our work thus constitutes a significant advancement in enhancing the practical applicability of robust average-reward methods to complex, real-world problems.
Authors: Junzhe Jiang, Chang Yang, Xinrun Wang, Zhiming Li, Xiao Huang, Bo Li
Abstract: The latency of the exchanges in Market Making (MM) is inevitable due to hardware limitations, system processing times, delays in receiving data from exchanges, the time required for order transmission to reach the market, etc. Existing reinforcement learning (RL) methods for Market Making (MM) overlook the impact of these latency, which can lead to unintended order cancellations due to price discrepancies between decision and execution times and result in undesired inventory accumulation, exposing MM traders to increased market risk. Therefore, these methods cannot be applied in real MM scenarios. To address these issues, we first build a realistic MM environment with random delays of 30-100 milliseconds for order placement and market information reception, and implement a batch matching mechanism that collects orders within every 500 milliseconds before matching them all at once, simulating the batch auction mechanisms adopted by some exchanges. Then, we propose Relaver, an RL-based method for MM to tackle the latency and inventory risk issues. The three main contributions of Relaver are: i) we introduce an augmented state-action space that incorporates order hold time alongside price and volume, enabling Relaver to optimize execution strategies under latency constraints and time-priority matching mechanisms, ii) we leverage dynamic programming (DP) to guide the exploration of RL training for better policies, iii) we train a market trend predictor, which can guide the agent to intelligently adjust the inventory to reduce the risk. Extensive experiments and ablation studies on four real-world datasets demonstrate that \textsc{Relaver} significantly improves the performance of state-of-the-art RL-based MM strategies across multiple metrics.
Authors: Hugues Van Assel, Mark Ibrahim, Tommaso Biancalani, Aviv Regev, Randall Balestriero
Abstract: Reconstruction and joint embedding have emerged as two leading paradigms in Self Supervised Learning (SSL). Reconstruction methods focus on recovering the original sample from a different view in input space. On the other hand, joint embedding methods align the representations of different views in latent space. Both approaches offer compelling advantages, yet practitioners lack clear guidelines for choosing between them. In this work, we unveil the core mechanisms that distinguish each paradigm. By leveraging closed form solutions for both approaches, we precisely characterize how the view generation process, e.g. data augmentation, impacts the learned representations. We then demonstrate that, unlike supervised learning, both SSL paradigms require a minimal alignment between augmentations and irrelevant features to achieve asymptotic optimality with increasing sample size. Our findings indicate that in scenarios where these irrelevant features have a large magnitude, joint embedding methods are preferable because they impose a strictly weaker alignment condition compared to reconstruction based methods. These results not only clarify the trade offs between the two paradigms but also substantiate the empirical success of joint embedding approaches on real world challenging datasets.
Authors: Rongwei Lu, Yutong Jiang, Jinrui Zhang, Chunyang Li, Yifei Zhu, Bin Chen, Zhi Wang
Abstract: Gradient compression can effectively alleviate communication bottlenecks in Federated Learning (FL). Contemporary state-of-the-art sparse compressors, such as Top-$k$, exhibit high computational complexity, up to $\mathcal{O}(d\log_2{k})$, where $d$ is the number of model parameters. The hard-threshold compressor, which simply transmits elements with absolute values higher than a fixed threshold, is thus proposed to reduce the complexity to $\mathcal{O}(d)$. However, the hard-threshold compression causes accuracy degradation in FL, where the datasets are non-IID and the stepsize $\gamma$ is decreasing for model convergence. The decaying stepsize reduces the updates and causes the compression ratio of the hard-threshold compression to drop rapidly to an aggressive ratio. At or below this ratio, the model accuracy has been observed to degrade severely. To address this, we propose $\gamma$-FedHT, a stepsize-aware low-cost compressor with Error-Feedback to guarantee convergence. Given that the traditional theoretical framework of FL does not consider Error-Feedback, we introduce the fundamental conversation of Error-Feedback. We prove that $\gamma$-FedHT has the convergence rate of $\mathcal{O}(\frac{1}{T})$ ($T$ representing total training iterations) under $\mu$-strongly convex cases and $\mathcal{O}(\frac{1}{\sqrt{T}})$ under non-convex cases, \textit{same as FedAVG}. Extensive experiments demonstrate that $\gamma$-FedHT improves accuracy by up to $7.42\%$ over Top-$k$ under equal communication traffic on various non-IID image datasets.
Authors: Zongkai Liu, Fanqing Meng, Lingxiao Du, Zhixiang Zhou, Chao Yu, Wenqi Shao, Qiaosheng Zhang
Abstract: Recent advances in rule-based reinforcement learning (RL) have significantly improved the reasoning capability of language models (LMs) with rule-based rewards. However, existing RL methods -- such as GRPO, REINFORCE++, and RLOO -- often suffer from training instability, where large policy updates and improper clipping can lead to training collapse. To address this issue, we propose Clipped Policy Gradient Optimization with Policy Drift (CPGD), a novel algorithm designed to stabilize policy learning in LMs. CPGD introduces a policy drift constraint based on KL divergence to dynamically regularize policy updates, and leverages a clip mechanism on the logarithm of the ratio to prevent excessive policy updates. We provide theoretical justification for CPGD and demonstrate through empirical analysis that it mitigates the instability observed in prior approaches. Furthermore, we show that CPGD significantly improves performance while maintaining training stability. Our implementation balances theoretical rigor with practical usability, offering a robust alternative for RL in the post-training of LMs. We release our code at https://github.com/ModalMinds/MM-EUREKA.
Authors: Yotam Norman, Ron Meir
Abstract: We propose a novel unsupervised framework for \emph{Invariant Risk Minimization} (IRM), extending the concept of invariance to settings where labels are unavailable. Traditional IRM methods rely on labeled data to learn representations that are robust to distributional shifts across environments. In contrast, our approach redefines invariance through feature distribution alignment, enabling robust representation learning from unlabeled data. We introduce two methods within this framework: Principal Invariant Component Analysis (PICA), a linear method that extracts invariant directions under Gaussian assumptions, and Variational Invariant Autoencoder (VIAE), a deep generative model that disentangles environment-invariant and environment-dependent latent factors. Our approach is based on a novel ``unsupervised'' structural causal model and supports environment-conditioned sample-generation and intervention. Empirical evaluations on synthetic dataset and modified versions of MNIST demonstrate the effectiveness of our methods in capturing invariant structure, preserving relevant information, and generalizing across environments without access to labels.
Authors: Rubens O. Moraes, Quazi Asif Sadmine, Hendrik Baier, Levi H. S. Lelis
Abstract: Outside of transfer learning settings, reinforcement learning agents start their learning process from a clean slate. As a result, such agents have to go through a slow process to learn even the most obvious skills required to solve a problem. In this paper, we present InnateCoder, a system that leverages human knowledge encoded in foundation models to provide programmatic policies that encode "innate skills" in the form of temporally extended actions, or options. In contrast to existing approaches to learning options, InnateCoder learns them from the general human knowledge encoded in foundation models in a zero-shot setting, and not from the knowledge the agent gains by interacting with the environment. Then, InnateCoder searches for a programmatic policy by combining the programs encoding these options into larger and more complex programs. We hypothesized that InnateCoder's way of learning and using options could improve the sampling efficiency of current methods for learning programmatic policies. Empirical results in MicroRTS and Karel the Robot support our hypothesis, since they show that InnateCoder is more sample efficient than versions of the system that do not use options or learn them from experience.
Authors: Junhao Liu, Haonan Yu, Xin Zhang
Abstract: With Large language models (LLMs) becoming increasingly prevalent in various applications, the need for interpreting their predictions has become a critical challenge. As LLMs vary in architecture and some are closed-sourced, model-agnostic techniques show great promise without requiring access to the model's internal parameters. However, existing model-agnostic techniques need to invoke LLMs many times to gain sufficient samples for generating faithful explanations, which leads to high economic costs. In this paper, we show that it is practical to generate faithful explanations for large-scale LLMs by sampling from some budget-friendly models through a series of empirical studies. Moreover, we show that such proxy explanations also perform well on downstream tasks. Our analysis provides a new paradigm of model-agnostic explanation methods for LLMs, by including information from budget-friendly models.
Authors: Truman Hickok
Abstract: Future deep learning models will be distinguished by systems that perpetually learn through interaction, imagination, and cooperation, blurring the line between training and inference. This makes continual learning a critical challenge, as methods that efficiently maximize bidirectional transfer across learning trajectories will be essential. Replay is on track to play a foundational role in continual learning, allowing models to directly reconcile new information with past knowledge. In practice, however, replay is quite unscalable, doubling the cost of continual learning when applied naively. Moreover, the continual learning literature has not fully synchronized with the multi-task fine-tuning literature, having not fully integrated highly scalable techniques like model merging and low rank adaptation into a replay-enabled toolset that can produce a unified model in the face of many sequential tasks. In this paper, we begin by applying and analyzing low rank adaptation in a continual learning setting. Next, we introduce consolidation, a phasic approach to replay which leads to up to 55\% less replay samples being needed for a given performance target. Then, we propose sequential merging, an offshoot of task arithmetic which is tailored to the continual learning setting and is shown to work well in combination with replay. Finally, we demonstrate that the developed strategies can operate synergistically, resulting in a highly scalable toolset that outperforms standalone variants.
Authors: Hanlin Zhu, Shibo Hao, Zhiting Hu, Jiantao Jiao, Stuart Russell, Yuandong Tian
Abstract: Large Language Models (LLMs) have demonstrated remarkable performance in many applications, including challenging reasoning problems via chain-of-thoughts (CoTs) techniques that generate ``thinking tokens'' before answering the questions. While existing theoretical works demonstrate that CoTs with discrete tokens boost the capability of LLMs, recent work on continuous CoTs lacks a theoretical understanding of why it outperforms discrete counterparts in various reasoning tasks such as directed graph reachability, a fundamental graph reasoning problem that includes many practical domain applications as special cases. In this paper, we prove that a two-layer transformer with $D$ steps of continuous CoTs can solve the directed graph reachability problem, where $D$ is the diameter of the graph, while the best known result of constant-depth transformers with discrete CoTs requires $O(n^2)$ decoding steps where $n$ is the number of vertices ($D
Authors: Josh Millar, Hamed Haddadi, Anil Madhavapeddy
Abstract: The use of deep learning (DL) on Internet of Things (IoT) and mobile devices offers numerous advantages over cloud-based processing. However, such devices face substantial energy constraints to prolong battery-life, or may even operate intermittently via energy-harvesting. Consequently, \textit{energy-aware} approaches for optimizing DL inference and training on such resource-constrained devices have garnered recent interest. We present an overview of such approaches, outlining their methodologies, implications for energy consumption and system-level efficiency, and their limitations in terms of supported network types, hardware platforms, and application scenarios. We hope our review offers a clear synthesis of the evolving energy-aware DL landscape and serves as a foundation for future research in energy-constrained computing.
Authors: Alexander Panyshev, Dmitry Vinichenko, Oleg Travkin, Roman Alferov, Alexey Zaytsev
Abstract: Temporal Graph Networks (TGNs), while being accurate, face significant training inefficiencies due to irregular supervision signals in dynamic graphs, which induce sparse gradient updates. We first theoretically establish that aggregating historical node interactions into pseudo-labels reduces gradient variance, accelerating convergence. Building on this analysis, we propose History-Averaged Labels (HAL), a method that dynamically enriches training batches with pseudo-targets derived from historical label distributions. HAL ensures continuous parameter updates without architectural modifications by converting idle computation into productive learning steps. Experiments on the Temporal Graph Benchmark (TGB) validate our findings and an assumption about slow change of user preferences: HAL accelerates TGNv2 training by up to 15x while maintaining competitive performance. Thus, this work offers an efficient, lightweight, architecture-agnostic, and theoretically motivated solution to label sparsity in temporal graph learning.
Authors: Yutian He, Yankun Huang, Yao Yao, Qihang Lin
Abstract: Fairness in machine learning has become a critical concern, particularly in high-stakes applications. Existing approaches often focus on achieving full fairness across all score ranges generated by predictive models, ensuring fairness in both high and low-scoring populations. However, this stringent requirement can compromise predictive performance and may not align with the practical fairness concerns of stakeholders. In this work, we propose a novel framework for building partially fair machine learning models, which enforce fairness within a specific score range of interest, such as the middle range where decisions are most contested, while maintaining flexibility in other regions. We introduce two statistical metrics to rigorously evaluate partial fairness within a given score range, such as the top 20%-40% of scores. To achieve partial fairness, we propose an in-processing method by formulating the model training problem as constrained optimization with difference-of-convex constraints, which can be solved by an inexact difference-of-convex algorithm (IDCA). We provide the complexity analysis of IDCA for finding a nearly KKT point. Through numerical experiments on real-world datasets, we demonstrate that our framework achieves high predictive performance while enforcing partial fairness where it matters most.
Authors: Adrian Mirza, Nawaf Alampara, Marti\~no R\'ios-Garc\'ia, Mohamed Abdelalim, Jack Butler, Bethany Connolly, Tunca Dogan, Marianna Nezhurina, B\"unyamin \c{S}en, Santosh Tirunagari, Mark Worrall, Adamo Young, Philippe Schwaller, Michael Pieler, Kevin Maik Jablonka
Abstract: Foundation models have shown remarkable success across scientific domains, yet their impact in chemistry remains limited due to the absence of diverse, large-scale, high-quality datasets that reflect the field's multifaceted nature. We present the ChemPile, an open dataset containing over 75 billion tokens of curated chemical data, specifically built for training and evaluating general-purpose models in the chemical sciences. The dataset mirrors the human learning journey through chemistry -- from educational foundations to specialized expertise -- spanning multiple modalities and content types including structured data in diverse chemical representations (SMILES, SELFIES, IUPAC names, InChI, molecular renderings), scientific and educational text, executable code, and chemical images. ChemPile integrates foundational knowledge (textbooks, lecture notes), specialized expertise (scientific articles and language-interfaced data), visual understanding (molecular structures, diagrams), and advanced reasoning (problem-solving traces and code) -- mirroring how human chemists develop expertise through diverse learning materials and experiences. Constructed through hundreds of hours of expert curation, the ChemPile captures both foundational concepts and domain-specific complexity. We provide standardized training, validation, and test splits, enabling robust benchmarking. ChemPile is openly released via HuggingFace with a consistent API, permissive license, and detailed documentation. We hope the ChemPile will serve as a catalyst for chemical AI, enabling the development of the next generation of chemical foundation models.
Authors: Rishi Jha, Collin Zhang, Vitaly Shmatikov, John X. Morris
Abstract: We introduce the first method for translating text embeddings from one vector space to another without any paired data, encoders, or predefined sets of matches. Our unsupervised approach translates any embedding to and from a universal latent representation (i.e., a universal semantic structure conjectured by the Platonic Representation Hypothesis). Our translations achieve high cosine similarity across model pairs with different architectures, parameter counts, and training datasets. The ability to translate unknown embeddings into a different space while preserving their geometry has serious implications for the security of vector databases. An adversary with access only to embedding vectors can extract sensitive information about the underlying documents, sufficient for classification and attribute inference.
Authors: Manolis Zampetakis, Felix Zhou
Abstract: We introduce a novel framework for differentially private (DP) statistical estimation via data truncation, addressing a key challenge in DP estimation when the data support is unbounded. Traditional approaches rely on problem-specific sensitivity analysis, limiting their applicability. By leveraging techniques from truncated statistics, we develop computationally efficient DP estimators for exponential family distributions, including Gaussian mean and covariance estimation, achieving near-optimal sample complexity. Previous works on exponential families only consider bounded or one-dimensional families. Our approach mitigates sensitivity through truncation while carefully correcting for the introduced bias using maximum likelihood estimation and DP stochastic gradient descent. Along the way, we establish improved uniform convergence guarantees for the log-likelihood function of exponential families, which may be of independent interest. Our results provide a general blueprint for DP algorithm design via truncated statistics.
Authors: Mohammad R. Rezaei, Adji Bousso Dieng
Abstract: Alternators have recently been introduced as a framework for modeling time-dependent data. They often outperform other popular frameworks, such as state-space models and diffusion models, on challenging time-series tasks. This paper introduces a new Alternator model, called Alternator++, which enhances the flexibility of traditional Alternators by explicitly modeling the noise terms used to sample the latent and observed trajectories, drawing on the idea of noise models from the diffusion modeling literature. Alternator++ optimizes the sum of the Alternator loss and a noise-matching loss. The latter forces the noise trajectories generated by the two noise models to approximate the noise trajectories that produce the observed and latent trajectories. We demonstrate the effectiveness of Alternator++ in tasks such as density estimation, time series imputation, and forecasting, showing that it outperforms several strong baselines, including Mambas, ScoreGrad, and Dyffusion.
Authors: Taniya Kapoor, Abhishek Chandra, Anastasios Stamou, Stephen J Roberts
Abstract: Real-world systems, from aerospace to railway engineering, are modeled with partial differential equations (PDEs) describing the physics of the system. Estimating robust solutions for such problems is essential. Deep learning-based architectures, such as neural PDE solvers, have recently gained traction as a reliable solution method. The current state of development of these approaches, however, primarily focuses on improving accuracy. The environmental impact of excessive computation, leading to increased carbon emissions, has largely been overlooked. This paper introduces a carbon emission measure for a range of PDE solvers. Our proposed metric, EcoL2, balances model accuracy with emissions across data collection, model training, and deployment. Experiments across both physics-informed machine learning and operator learning architectures demonstrate that the proposed metric presents a holistic assessment of model performance and emission cost. As such solvers grow in scale and deployment, EcoL2 represents a step toward building performant scientific machine learning systems with lower long-term environmental impact.
Authors: Leyang Xue, Yao Fu, Luo Mai, Mahesh K. Marina
Abstract: Giant Deep Neural Networks (DNNs), have become indispensable for accurate and robust support of large-scale cloud based AI services. However, serving giant DNNs is prohibitively expensive from an energy consumption viewpoint easily exceeding that of training, due to the enormous scale of GPU clusters needed to hold giant DNN model partitions and replicas. Existing approaches can either optimize energy efficiency or inference accuracy but not both. To overcome this status quo, we propose HybridServe, a novel hybrid DNN model serving system that leverages multiple sized versions (small to giant) of the model to be served in tandem. Through a confidence based hybrid model serving dataflow, HybridServe prefers to serve inference requests with energy-efficient smaller models so long as accuracy is not compromised, thereby reducing the number of replicas needed for giant DNNs. HybridServe also features a dataflow planner for efficient partitioning and replication of candidate models to maximize serving system throughput. Experimental results using a prototype implementation of HybridServe show that it reduces energy footprint by up to 19.8x compared to the state-of-the-art DNN model serving systems while matching the accuracy of serving solely with giant DNNs.
Authors: Kiran Kokilepersaud, Mohit Prabhushankar, Ghassan AlRegib
Abstract: A key factor in effective Self-Supervised learning (SSL) is preventing dimensional collapse, which is where higher-dimensional representation spaces span a lower-dimensional subspace. Therefore, SSL optimization strategies involve guiding a model to produce representations ($R$) with a higher dimensionality. Dimensionality is either optimized through a dimension-contrastive approach that encourages feature decorrelation or through a sample-contrastive method that promotes a uniform spread of sample representations. Both families of SSL algorithms also utilize a projection head that maps $R$ into a lower-dimensional embedding space $Z$. Recent work has characterized the projection head as a filter of irrelevant features from the SSL objective by reducing mutual information, $I(R;Z)$. Therefore, the current literature's view is that a good SSL representation space should have a high $H(R)$ and a low $I(R;Z)$. However, this view of the problem is lacking in terms of an understanding of the underlying training dynamics that influences both terms, as well as how the values of $H(R)$ and $I(R;Z)$ arrived at the end of training reflect the downstream performance of an SSL model. We address both gaps in the literature by demonstrating that increases in $H(R)$ due to feature decorrelation at the start of training lead to a higher $I(R;Z)$, while increases in $H(R)$ due to samples distributing uniformly in a high-dimensional space at the end of training cause $I(R;Z)$ to plateau or decrease. Furthermore, our analysis shows that the best performing SSL models do not have the highest $H(R)$ nor the lowest $I(R;Z)$, but arrive at an optimal intermediate point for both. We develop a method called AdaDim to exploit these observed training dynamics by adaptively weighting between losses based on feature decorrelation and uniform sample spread.
Authors: Shiyun Xu, Zhiqi Bu
Abstract: Parameter-efficient fine-tuning (PEFT) is a highly effective approach for adapting large pre-trained models to downstream tasks with minimal computational overhead. At the core, PEFT methods freeze most parameters and only trains a small subset (say $<0.1\%$ of total parameters). Notably, different PEFT methods select different subsets, resulting in varying levels of performance. This variation prompts a key question: how to effectively select the most influential subset to train? We formulate the subset selection as a multi-task problem: maximizing the performance and minimizing the number of trainable parameters. We leverage a series of transformations -- including $\epsilon$-constraint method and second-order Taylor approximation -- to arrive at the classical 0-1 knapsack problem, which we solve through the lens of Pareto optimality. Consequently, we propose AdaPEFT, a Hessian-informed PEFT that adapts to various tasks and models, in which the selected subset empirically transfers across training horizons and model sizes.
Authors: Lucas M. Dorneles, Luan Fonseca Garcia, Joel Lu\'is Carbonera
Abstract: Neural networks have become increasingly popular in the last few years as an effective tool for the task of image classification due to the impressive performance they have achieved on this task. In image classification tasks, it is common to use data augmentation strategies to increase the robustness of trained networks to changes in the input images and to avoid overfitting. Although data augmentation is a widely adopted technique, the literature lacks a body of research analyzing the effects data augmentation methods have on the patterns learned by neural network models working on complex datasets. The primary objective of this work is to propose a methodology and set of metrics that may allow a quantitative approach to analyzing the effects of data augmentation in convolutional networks applied to image classification. An important tool used in the proposed approach lies in the concept of class activation maps for said models, which allow us to identify and measure the importance these models assign to each individual pixel in an image when executing the classification task. From these maps, we may then extract metrics over the similarities and differences between maps generated by these models trained on a given dataset with different data augmentation strategies. Experiments made using this methodology suggest that the effects of these data augmentation techniques not only can be analyzed in this way but also allow us to identify different impact profiles over the trained models.
Authors: En Yu, Jie Lu, Xiaoyu Yang, Guangquan Zhang, Zhen Fang
Abstract: Modern machine learning models struggle to maintain performance in dynamic environments where temporal distribution shifts, \emph{i.e., concept drift}, are prevalent. Temporal Domain Generalization (TDG) seeks to enable model generalization across evolving domains, yet existing approaches typically assume smooth incremental changes, struggling with complex real-world drifts involving long-term structure (incremental evolution/periodicity) and local uncertainties. To overcome these limitations, we introduce FreKoo, which tackles these challenges via a novel frequency-domain analysis of parameter trajectories. It leverages the Fourier transform to disentangle parameter evolution into distinct spectral bands. Specifically, low-frequency component with dominant dynamics are learned and extrapolated using the Koopman operator, robustly capturing diverse drift patterns including both incremental and periodicity. Simultaneously, potentially disruptive high-frequency variations are smoothed via targeted temporal regularization, preventing overfitting to transient noise and domain uncertainties. In addition, this dual spectral strategy is rigorously grounded through theoretical analysis, providing stability guarantees for the Koopman prediction, a principled Bayesian justification for the high-frequency regularization, and culminating in a multiscale generalization bound connecting spectral dynamics to improved generalization. Extensive experiments demonstrate FreKoo's significant superiority over SOTA TDG approaches, particularly excelling in real-world streaming scenarios with complex drifts and uncertainties.
Authors: Sanggeon Yun, Ryozo Masukawa, Hyunwoo Oh, Nathaniel D. Bastian, Mohsen Imani
Abstract: Deep neural networks (DNNs) are highly susceptible to adversarial examples--subtle, imperceptible perturbations that can lead to incorrect predictions. While detection-based defenses offer a practical alternative to adversarial training, many existing methods depend on external models, complex architectures, heavy augmentations, or adversarial data, limiting their efficiency and generalizability. We introduce a lightweight, plug-in detection framework that leverages internal layer-wise inconsistencies within the target model itself, requiring only benign data for calibration. Our approach is grounded in the A Few Large Shifts Assumption, which posits that adversarial perturbations typically induce large representation shifts in a small subset of layers. Building on this, we propose two complementary strategies--Recovery Testing (RT) and Logit-layer Testing (LT)--to expose internal disruptions caused by adversaries. Evaluated on CIFAR-10, CIFAR-100, and ImageNet under both standard and adaptive threat models, our method achieves state-of-the-art detection performance with negligible computational overhead and no compromise to clean accuracy.
Authors: Yang Li
Abstract: As large language models (LLMs) grow in scale and specialization, routing--selecting the best model for a given input--has become essential for efficient and effective deployment. While recent methods rely on complex learned routing strategies, their dependence on disparate training data and evaluation setups makes comparison and generalization difficult. In this work, we revisit LLM routing through the lens of simplicity. We show that a well-tuned k-Nearest Neighbors (kNN) approach not only matches but often outperforms state-of-the-art learned routers across diverse tasks. To support systematic evaluation, we introduce a suite of standardized routing benchmarks spanning instruction-following, question-answering, and reasoning tasks, as well as the first multi-modal routing dataset involving visual inputs. Our findings reveal that the locality properties of model performance in embedding space enable simple non-parametric methods to achieve strong routing decisions with lower sample complexity than parametric approaches. This challenges the prevailing trend toward sophisticated architectures and highlights the importance of thoroughly evaluating simple baselines before investing in complex solutions. To support reproducibility and further exploration, we will release all benchmarks and code upon publication.
Authors: Grant C. Forbes, Jianxun Wang, Leonardo Villalobos-Arias, Arnav Jhala, David L. Roberts
Abstract: Recent RL research has utilized reward shaping--particularly complex shaping rewards such as intrinsic motivation (IM)--to encourage agent exploration in sparse-reward environments. While often effective, ``reward hacking'' can lead to the shaping reward being optimized at the expense of the extrinsic reward, resulting in a suboptimal policy. Potential-Based Reward Shaping (PBRS) techniques such as Generalized Reward Matching (GRM) and Policy-Invariant Explicit Shaping (PIES) have mitigated this. These methods allow for implementing IM without altering optimal policies. In this work we show that they are effectively unsuitable for complex, exploration-heavy environments with long-duration episodes. To remedy this, we introduce Action-Dependent Optimality Preserving Shaping (ADOPS), a method of converting intrinsic rewards to an optimality-preserving form that allows agents to utilize IM more effectively in the extremely sparse environment of Montezuma's Revenge. We also prove ADOPS accommodates reward shaping functions that cannot be written in a potential-based form: while PBRS-based methods require the cumulative discounted intrinsic return be independent of actions, ADOPS allows for intrinsic cumulative returns to be dependent on agents' actions while still preserving the optimal policy set. We show how action-dependence enables ADOPS's to preserve optimality while learning in complex, sparse-reward environments where other methods struggle.
Authors: Pengfei Ding, Yan Wang, Guanfeng Liu, Jiajie Zhu
Abstract: Graph unlearning, which deletes graph elements such as nodes and edges from trained graph neural networks (GNNs), is crucial for real-world applications where graph data may contain outdated, inaccurate, or privacy-sensitive information. However, existing methods often suffer from (1) incomplete or over unlearning due to neglecting the distinct objectives of different unlearning tasks, and (2) inaccurate identification of neighbors affected by deleted elements across various GNN architectures. To address these limitations, we propose AGU, a novel Adaptive Graph Unlearning framework that flexibly adapts to diverse unlearning tasks and GNN architectures. AGU ensures the complete forgetting of deleted elements while preserving the integrity of the remaining graph. It also accurately identifies affected neighbors for each GNN architecture and prioritizes important ones to enhance unlearning performance. Extensive experiments on seven real-world graphs demonstrate that AGU outperforms existing methods in terms of effectiveness, efficiency, and unlearning capability.
Authors: Wanfu Gao, Zengyao Man, Hanlin Pan, Kunpeng Liu
Abstract: Feature generation involves creating new features from raw data to capture complex relationships among the original features, improving model robustness and machine learning performance. Current methods using reinforcement learning for feature generation have made feature exploration more flexible and efficient. However, several challenges remain: first, during feature expansion, a large number of redundant features are generated. When removing them, current methods only retain the best features each round, neglecting those that perform poorly initially but could improve later. Second, the state representation used by current methods fails to fully capture complex feature relationships. Third, there are significant differences between discrete and continuous features in tabular data, requiring different operations for each type. To address these challenges, we propose a novel dual-agent reinforcement learning method for feature generation. Two agents are designed: the first generates new features, and the second determines whether they should be preserved. A self-attention mechanism enhances state representation, and diverse operations distinguish interactions between discrete and continuous features. The experimental results on multiple datasets demonstrate that the proposed method is effective. The code is available at https://github.com/extess0/DARL.
Authors: Yuchang Sun, Yanxi Chen, Yaliang Li, Bolin Ding
Abstract: Augmenting large language models (LLMs) with auxiliary tokens has emerged as a promising strategy for enhancing model performance. In this work, we introduce a lightweight method termed latent tokens; these are dummy tokens that may be non-interpretable in natural language but steer the autoregressive decoding process of a Transformer-based LLM via the attention mechanism. The proposed latent tokens can be seamlessly integrated with a pre-trained Transformer, trained in a parameter-efficient manner, and applied flexibly at inference time, while adding minimal complexity overhead to the existing infrastructure of standard Transformers. We propose several hypotheses about the underlying mechanisms of latent tokens and design synthetic tasks accordingly to verify them. Numerical results confirm that the proposed method noticeably outperforms the baselines, particularly in the out-of-distribution generalization scenarios, highlighting its potential in improving the adaptability of LLMs.
Authors: Jung Hoon Lee, Sujith Vijayan
Abstract: Deep learning (DL) can automatically construct intelligent agents, deep neural networks (alternatively, DL models), that can outperform humans in certain tasks. However, the operating principles of DL remain poorly understood, making its decisions incomprehensible. As a result, it poses a great risk to deploy DL in high-stakes domains in which mistakes or errors may lead to critical consequences. Here, we aim to develop an algorithm that can help DL models make more robust decisions by allowing them to abstain from answering when they are uncertain. Our algorithm, named `Two out of Three (ToT)', is inspired by the sensitivity of the human brain to conflicting information. ToT creates two alternative predictions in addition to the original model prediction and uses the alternative predictions to decide whether it should provide an answer or not.
Authors: Jung Hoon Lee, Sujith Vijayan
Abstract: Deep learning (DL) is a powerful tool that can solve complex problems, and thus, it seems natural to assume that DL can be used to enhance the security of wireless communication. However, deploying DL models to edge devices in wireless networks is challenging, as they require significant amounts of computing and power resources. Notably, Spiking Neural Networks (SNNs) are known to be efficient in terms of power consumption, meaning they can be an alternative platform for DL models for edge devices. In this study, we ask if SNNs can be used in physical layer authentication. Our evaluation suggests that SNNs can learn unique physical properties (i.e., `fingerprints') of RF transmitters and use them to identify individual devices. Furthermore, we find that SNNs are also vulnerable to adversarial attacks and that an autoencoder can be used clean out adversarial perturbations to harden SNNs against them.
Authors: Tonglong Wei, Yan Lin, Zeyu Zhou, Haomin Wen, Jilin Hu, Shengnan Guo, Youfang Lin, Gao Cong, Huaiyu Wan
Abstract: Vehicle GPS trajectories provide valuable movement information that supports various downstream tasks and applications. A desirable trajectory learning model should be able to transfer across regions and tasks without retraining, avoiding the need to maintain multiple specialized models and subpar performance with limited training data. However, each region has its unique spatial features and contexts, which are reflected in vehicle movement patterns and difficult to generalize. Additionally, transferring across different tasks faces technical challenges due to the varying input-output structures required for each task. Existing efforts towards transferability primarily involve learning embedding vectors for trajectories, which perform poorly in region transfer and require retraining of prediction modules for task transfer. To address these challenges, we propose TransferTraj, a vehicle GPS trajectory learning model that excels in both region and task transferability. For region transferability, we introduce RTTE as the main learnable module within TransferTraj. It integrates spatial, temporal, POI, and road network modalities of trajectories to effectively manage variations in spatial context distribution across regions. It also introduces a TRIE module for incorporating relative information of spatial features and a spatial context MoE module for handling movement patterns in diverse contexts. For task transferability, we propose a task-transferable input-output scheme that unifies the input-output structure of different tasks into the masking and recovery of modalities and trajectory points. This approach allows TransferTraj to be pre-trained once and transferred to different tasks without retraining. Extensive experiments on three real-world vehicle trajectory datasets under task transfer, zero-shot, and few-shot region transfer, validating TransferTraj's effectiveness.
Authors: Hana Satou, Alan Mitkiy
Abstract: Transfer learning across domains with distribution shift remains a fundamental challenge in building robust and adaptable machine learning systems. While adversarial perturbations are traditionally viewed as threats that expose model vulnerabilities, recent studies suggest that they can also serve as constructive tools for data augmentation. In this work, we systematically investigate the role of adversarial data augmentation (ADA) in enhancing both robustness and adaptivity in transfer learning settings. We analyze how adversarial examples, when used strategically during training, improve domain generalization by enriching decision boundaries and reducing overfitting to source-domain-specific features. We further propose a unified framework that integrates ADA with consistency regularization and domain-invariant representation learning. Extensive experiments across multiple benchmark datasets -- including VisDA, DomainNet, and Office-Home -- demonstrate that our method consistently improves target-domain performance under both unsupervised and few-shot domain adaptation settings. Our results highlight a constructive perspective of adversarial learning, transforming perturbation from a destructive attack into a regularizing force for cross-domain transferability.
Authors: Yun-Yun Tsai, Chuan Guo, Junfeng Yang, Laurens van der Maaten
Abstract: AI developers are releasing large language models (LLMs) under a variety of different licenses. Many of these licenses restrict the ways in which the models or their outputs may be used. This raises the question how license violations may be recognized. In particular, how can we identify that an API or product uses (an adapted version of) a particular LLM? We present a new method that enable model developers to perform such identification via fingerprints: statistical patterns that are unique to the developer's model and robust to common alterations of that model. Our method permits model identification in a black-box setting using a limited number of queries, enabling identification of models that can only be accessed via an API or product. The fingerprints are non-invasive: our method does not require any changes to the model during training, hence by design, it does not impact model quality. Empirically, we find our method provides a high degree of robustness to common changes in the model or inference settings. In our experiments, it substantially outperforms prior art, including invasive methods that explicitly train watermarks into the model.
Authors: Yihong Huang, Chen Chu
Abstract: Key feature fields need bigger embedding dimensionality, others need smaller. This demands automated dimension allocation. Existing approaches, such as pruning or Neural Architecture Search (NAS), require training a memory-intensive SuperNet that enumerates all possible dimension combinations, which is infeasible for large feature spaces. We propose DimGrow, a lightweight approach that eliminates the SuperNet requirement. Starting training model from one dimension per feature field, DimGrow can progressively expand/shrink dimensions via importance scoring. Dimensions grow only when their importance consistently exceed a threshold, ensuring memory efficiency. Experiments on three recommendation datasets verify the effectiveness of DimGrow while it reduces training memory compared to SuperNet-based methods.
Authors: Yinlin Zhu, Xunkai Li, Jishuo Jia, Miao Hu, Di Wu, Meikang Qiu
Abstract: Recent advances in graph machine learning have shifted to data-centric paradigms, driven by two emerging fields: (1) Federated graph learning (FGL) enables multi-client collaboration but faces challenges from data and task heterogeneity, limiting its practicality; (2) Graph foundation models (GFM) offer strong domain generalization but are usually trained on single machines, missing out on cross-silo data and resources. These paradigms are complementary, and their integration brings notable benefits. Motivated by this, we propose FedGFM, a novel decentralized GFM training paradigm. However, a key challenge is knowledge entanglement, where multi-domain knowledge merges into indistinguishable representations, hindering downstream adaptation. To address this, we present FedGFM+, an enhanced framework with two core modules to reduce knowledge entanglement: (1) AncDAI: A global anchor-based domain-aware initialization strategy. Before pre-training, each client encodes its local graph into domain-specific prototypes that serve as semantic anchors. Synthetic embeddings around these anchors initialize the global model. We theoretically prove these prototypes are distinguishable across domains, providing a strong inductive bias to disentangle domain-specific knowledge. (2) AdaDPP: A local adaptive domain-sensitive prompt pool. Each client learns a lightweight graph prompt capturing domain semantics during pre-training. During fine-tuning, prompts from all clients form a pool from which the GFM selects relevant prompts to augment target graph attributes, improving downstream adaptation. FedGFM+ is evaluated on 8 diverse benchmarks across multiple domains and tasks, outperforming 20 baselines from supervised learning, FGL, and federated GFM variants.
Authors: Seungmin Kim, Sohee Park, Donghyun Kim, Jisu Lee, Daeseon Choi
Abstract: With the advancement of AI-based speech synthesis technologies such as Deep Voice, there is an increasing risk of voice spoofing attacks, including voice phishing and fake news, through unauthorized use of others' voices. Existing defenses that inject adversarial perturbations directly into audio signals have limited effectiveness, as these perturbations can easily be neutralized by speech enhancement methods. To overcome this limitation, we propose RoVo (Robust Voice), a novel proactive defense technique that injects adversarial perturbations into high-dimensional embedding vectors of audio signals, reconstructing them into protected speech. This approach effectively defends against speech synthesis attacks and also provides strong resistance to speech enhancement models, which represent a secondary attack threat. In extensive experiments, RoVo increased the Defense Success Rate (DSR) by over 70% compared to unprotected speech, across four state-of-the-art speech synthesis models. Specifically, RoVo achieved a DSR of 99.5% on a commercial speaker-verification API, effectively neutralizing speech synthesis attack. Moreover, RoVo's perturbations remained robust even under strong speech enhancement conditions, outperforming traditional methods. A user study confirmed that RoVo preserves both naturalness and usability of protected speech, highlighting its effectiveness in complex and evolving threat scenarios.
Authors: Shuyang Dong, Shangtong Zhang, Lu Feng
Abstract: Reinforcement Learning (RL) has shown great promise in domains like healthcare and robotics but often struggles with adoption due to its lack of interpretability. Counterfactual explanations, which address "what if" scenarios, provide a promising avenue for understanding RL decisions but remain underexplored for continuous action spaces. We propose a novel approach for generating counterfactual explanations in continuous action RL by computing alternative action sequences that improve outcomes while minimizing deviations from the original sequence. Our approach leverages a distance metric for continuous actions and accounts for constraints such as adhering to predefined policies in specific states. Evaluations in two RL domains, Diabetes Control and Lunar Lander, demonstrate the effectiveness, efficiency, and generalization of our approach, enabling more interpretable and trustworthy RL applications.
Authors: Yingchen He, Christian D. Weilbach, Martyna E. Wojciechowska, Yuxuan Zhang, Frank Wood
Abstract: Advances in deep generative modelling have made it increasingly plausible to train human-level embodied agents. Yet progress has been limited by the absence of large-scale, real-time, multi-modal, and socially interactive datasets that reflect the sensory-motor complexity of natural environments. To address this, we present PLAICraft, a novel data collection platform and dataset capturing multiplayer Minecraft interactions across five time-aligned modalities: video, game output audio, microphone input audio, mouse, and keyboard actions. Each modality is logged with millisecond time precision, enabling the study of synchronous, embodied behaviour in a rich, open-ended world. The dataset comprises over 10,000 hours of gameplay from more than 10,000 global participants.\footnote{We have done a privacy review for the public release of an initial 200-hour subset of the dataset, with plans to release most of the dataset over time.} Alongside the dataset, we provide an evaluation suite for benchmarking model capabilities in object recognition, spatial awareness, language grounding, and long-term memory. PLAICraft opens a path toward training and evaluating agents that act fluently and purposefully in real time, paving the way for truly embodied artificial intelligence.
Authors: Zhichen Zeng, Ruizhong Qiu, Wenxuan Bao, Tianxin Wei, Xiao Lin, Yuchen Yan, Tarek F. Abdelzaher, Jiawei Han, Hanghang Tong
Abstract: Graph neural networks, despite their impressive performance, are highly vulnerable to distribution shifts on graphs. Existing graph domain adaptation (graph DA) methods often implicitly assume a \textit{mild} shift between source and target graphs, limiting their applicability to real-world scenarios with \textit{large} shifts. Gradual domain adaptation (GDA) has emerged as a promising approach for addressing large shifts by gradually adapting the source model to the target domain via a path of unlabeled intermediate domains. Existing GDA methods exclusively focus on independent and identically distributed (IID) data with a predefined path, leaving their extension to \textit{non-IID graphs without a given path} an open challenge. To bridge this gap, we present Gadget, the first GDA framework for non-IID graph data. First (\textit{theoretical foundation}), the Fused Gromov-Wasserstein (FGW) distance is adopted as the domain discrepancy for non-IID graphs, based on which, we derive an error bound revealing that the target domain error is proportional to the length of the path. Second (\textit{optimal path}), guided by the error bound, we identify the FGW geodesic as the optimal path, which can be efficiently generated by our proposed algorithm. The generated path can be seamlessly integrated with existing graph DA methods to handle large shifts on graphs, improving state-of-the-art graph DA methods by up to 6.8\% in node classification accuracy on real-world datasets.
Authors: Yingkai Kang, Jiawen Kang, Jinbo Wen, Tao Zhang, Zhaohui Yang, Dusit Niyato, Yan Zhang
Abstract: Vehicular metaverses are an emerging paradigm that merges intelligent transportation systems with virtual spaces, leveraging advanced digital twin and Artificial Intelligence (AI) technologies to seamlessly integrate vehicles, users, and digital environments. In this paradigm, vehicular AI agents are endowed with environment perception, decision-making, and action execution capabilities, enabling real-time processing and analysis of multi-modal data to provide users with customized interactive services. Since vehicular AI agents require substantial resources for real-time decision-making, given vehicle mobility and network dynamics conditions, the AI agents are deployed in RoadSide Units (RSUs) with sufficient resources and dynamically migrated among them. However, AI agent migration requires frequent data exchanges, which may expose vehicular metaverses to potential cyber attacks. To this end, we propose a reliable vehicular AI agent migration framework, achieving reliable dynamic migration and efficient resource scheduling through cooperation between vehicles and RSUs. Additionally, we design a trust evaluation model based on the theory of planned behavior to dynamically quantify the reputation of RSUs, thereby better accommodating the personalized trust preferences of users. We then model the vehicular AI agent migration process as a partially observable markov decision process and develop a Confidence-regulated Generative Diffusion Model (CGDM) to efficiently generate AI agent migration decisions. Numerical results demonstrate that the CGDM algorithm significantly outperforms baseline methods in reducing system latency and enhancing robustness against cyber attacks.
Authors: Zeyi Ren, Jingreng Lei, Yichen Jin, Ermo Hua, Qingfeng Lin, Chen Zhang, Bowen Zhou, Yik-Chung Wu
Abstract: The development of edge computing places critical demands on energy-efficient model deployment for multiple-input multiple-output (MIMO) detection tasks. Deploying deep unfolding models such as PGD-Nets and ADMM-Nets into resource-constrained edge devices using quantization methods is challenging. Existing quantization methods based on quantization aware training (QAT) suffer from performance degradation due to their reliance on parametric distribution assumption of activations and static quantization step sizes. To address these challenges, this paper proposes a novel kernel-based adaptive quantization (KAQ) framework for deep unfolding networks. By utilizing a joint kernel density estimation (KDE) and maximum mean discrepancy (MMD) approach to align activation distributions between full-precision and quantized models, the need for prior distribution assumptions is eliminated. Additionally, a dynamic step size updating method is introduced to adjust the quantization step size based on the channel conditions of wireless networks. Extensive simulations demonstrate that the accuracy of proposed KAQ framework outperforms traditional methods and successfully reduces the model's inference latency.
Authors: Hongjoon Ahn, Heewoong Choi, Jisu Han, Taesup Moon
Abstract: Offline goal-conditioned reinforcement learning (GCRL) offers a practical learning paradigm where goal-reaching policies are trained from abundant unlabeled (reward-free) datasets without additional environment interaction. However, offline GCRL still struggles with long-horizon tasks, even with recent advances that employ hierarchical policy structures, such as HIQL. By identifying the root cause of this challenge, we observe the following insights: First, performance bottlenecks mainly stem from the high-level policy's inability to generate appropriate subgoals. Second, when learning the high-level policy in the long-horizon regime, the sign of the advantage signal frequently becomes incorrect. Thus, we argue that improving the value function to produce a clear advantage signal for learning the high-level policy is essential. In this paper, we propose a simple yet effective solution: Option-aware Temporally Abstracted value learning, dubbed OTA, which incorporates temporal abstraction into the temporal-difference learning process. By modifying the value update to be option-aware, the proposed learning scheme contracts the effective horizon length, enabling better advantage estimates even in long-horizon regimes. We experimentally show that the high-level policy extracted using the OTA value function achieves strong performance on complex tasks from OGBench, a recently proposed offline GCRL benchmark, including maze navigation and visual robotic manipulation environments.
Authors: Chenghua Gong, Rui Sun, Yuhao Zheng, Juyuan Zhang, Tianjun Gu, Liming Pan, Linyuan Lv
Abstract: Advanced epidemic forecasting is critical for enabling precision containment strategies, highlighting its strategic importance for public health security. While recent advances in Large Language Models (LLMs) have demonstrated effectiveness as foundation models for domain-specific tasks, their potential for epidemic forecasting remains largely unexplored. In this paper, we introduce EpiLLM, a novel LLM-based framework tailored for spatio-temporal epidemic forecasting. Considering the key factors in real-world epidemic transmission: infection cases and human mobility, we introduce a dual-branch architecture to achieve fine-grained token-level alignment between such complex epidemic patterns and language tokens for LLM adaptation. To unleash the multi-step forecasting and generalization potential of LLM architectures, we propose an autoregressive modeling paradigm that reformulates the epidemic forecasting task into next-token prediction. To further enhance LLM perception of epidemics, we introduce spatio-temporal prompt learning techniques, which strengthen forecasting capabilities from a data-driven perspective. Extensive experiments show that EpiLLM significantly outperforms existing baselines on real-world COVID-19 datasets and exhibits scaling behavior characteristic of LLMs.
Authors: Dong Kyu Cho, Inwoo Hwang, Sanghack Lee
Abstract: Data augmentation is a popular tool for single source domain generalization, which expands the source domain by generating simulated ones, improving generalization on unseen target domains. In this work, we show that the performance of such augmentation-based methods in the target domains universally fluctuates during training, posing challenges in model selection under realistic scenarios. We argue that the fluctuation stems from the inability of the model to accumulate the knowledge learned from diverse augmentations, exacerbating feature distortion during training. Based on this observation, we propose a novel generalization method, coined Parameter-Space Ensemble with Entropy Regularization (PEER), that uses a proxy model to learn the augmented data on behalf of the main model. The main model is updated by averaging its parameters with the proxy model, progressively accumulating knowledge over the training steps. Maximizing the mutual information between the output representations of the two models guides the learning process of the proxy model, mitigating feature distortion during training. Experimental results demonstrate the effectiveness of PEER in reducing the OOD performance fluctuation and enhancing generalization across various datasets, including PACS, Digits, Office-Home, and VLCS. Notably, our method with simple random augmentation achieves state-of-the-art performance, surpassing prior approaches on sDG that utilize complex data augmentation strategies.
Authors: Filippo Leveni
Abstract: Anomaly detection is a fundamental problem in domains such as healthcare, manufacturing, and cybersecurity. This thesis proposes new unsupervised methods for anomaly detection in both structured and streaming data settings. In the first part, we focus on structure-based anomaly detection, where normal data follows low-dimensional manifolds while anomalies deviate from them. We introduce Preference Isolation Forest (PIF), which embeds data into a high-dimensional preference space via manifold fitting, and isolates outliers using two variants: Voronoi-iForest, based on geometric distances, and RuzHash-iForest, leveraging Locality Sensitive Hashing for scalability. We also propose Sliding-PIF, which captures local manifold information for streaming scenarios. Our methods outperform existing techniques on synthetic and real datasets. We extend this to structure-based clustering with MultiLink, a novel method for recovering multiple geometric model families in noisy data. MultiLink merges clusters via a model-aware linkage strategy, enabling robust multi-class structure recovery. It offers key advantages over existing approaches, such as speed, reduced sensitivity to thresholds, and improved robustness to poor initial sampling. The second part of the thesis addresses online anomaly detection in evolving data streams. We propose Online Isolation Forest (Online-iForest), which uses adaptive, multi-resolution histograms and dynamically updates tree structures to track changes over time. It avoids retraining while achieving accuracy comparable to offline models, with superior efficiency for real-time applications. Finally, we tackle anomaly detection in cybersecurity via open-set recognition for malware classification. We enhance a Gradient Boosting classifier with MaxLogit to detect unseen malware families, a method now integrated into Cleafy's production system.
Authors: Wenya Guo, Zhengkun Zhang, Xumeng Liu, Ying Zhang, Ziyu Lu, Haoze Zhu, Xubo Liu, Ruxue Yan
Abstract: Instruction data selection aims to identify a high-quality subset from the training set that matches or exceeds the performance of the full dataset on target tasks. Existing methods focus on the instruction-to-response mapping, but neglect the human preference for diverse responses. In this paper, we propose Preference-oriented Data Selection method (ProDS) that scores training samples based on their alignment with preferences observed in the target set. Our key innovation lies in shifting the data selection criteria from merely estimating features for accurate response generation to explicitly aligning training samples with human preferences in target tasks. Specifically, direct preference optimization (DPO) is employed to estimate human preferences across diverse responses. Besides, a bidirectional preference synthesis strategy is designed to score training samples according to both positive preferences and negative preferences. Extensive experimental results demonstrate our superiority to existing task-agnostic and targeted methods.
Authors: Haochen Yuan, Minting Pan, Yunbo Wang, Siyu Gao, Philip S. Yu, Xiaokang Yang
Abstract: Reinforcement learning (RL) has shown significant promise for sequential portfolio optimization tasks, such as stock trading, where the objective is to maximize cumulative returns while minimizing risks using historical data. However, traditional RL approaches often produce policies that merely memorize the optimal yet impractical buying and selling behaviors within the fixed dataset. These offline policies are less generalizable as they fail to account for the non-stationary nature of the market. Our approach, MetaTrader, frames portfolio optimization as a new type of partial-offline RL problem and makes two technical contributions. First, MetaTrader employs a bilevel learning framework that explicitly trains the RL agent to improve both in-domain profits on the original dataset and out-of-domain performance across diverse transformations of the raw financial data. Second, our approach incorporates a new temporal difference (TD) method that approximates worst-case TD estimates from a batch of transformed TD targets, addressing the value overestimation issue that is particularly challenging in scenarios with limited offline data. Our empirical results on two public stock datasets show that MetaTrader outperforms existing methods, including both RL-based approaches and traditional stock prediction models.
Authors: Donghwa Shin, Edwin Zhang
Abstract: Transformers have recently gained popularity in time series forecasting due to their ability to capture long-term dependencies. However, many existing models focus only on capturing temporal dependencies while omitting intricate relationships between variables. Recent models have tried tackling this by explicitly modeling both cross-time and cross-variate dependencies through a sequential or unified attention mechanism, but they are entirely channel dependent (CD) across all layers, making them potentially susceptible to overfitting. To address this, we propose Cross-Variate Patch Embeddings (CVPE), a lightweight CD module that injects cross-variate context into channel-independent (CI) models by simply modifying the patch embedding process. We achieve this by adding a learnable positional encoding and a lightweight router-attention block to the vanilla patch embedding layer. We then integrate CVPE into Time-LLM, a multimodal CI forecasting model, to demonstrate its effectiveness in capturing cross-variate dependencies and enhance the CI model's performance. Extensive experimental results on seven real-world datasets show that our enhanced Time-LLM outperforms the original baseline model simply by incorporating the CVPE module, with no other changes.
Authors: Sunghwan Kim, Dongjin Kang, Taeyoon Kwon, Hyungjoo Chae, Dongha Lee, Jinyoung Yeo
Abstract: Reward models (RMs) play a crucial role in reinforcement learning from human feedback (RLHF), aligning model behavior with human preferences. However, existing benchmarks for reward models show a weak correlation with the performance of optimized policies, suggesting that they fail to accurately assess the true capabilities of RMs. To bridge this gap, we explore several evaluation designs through the lens of reward overoptimization\textemdash a phenomenon that captures both how well the reward model aligns with human preferences and the dynamics of the learning signal it provides to the policy. The results highlight three key findings on how to construct a reliable benchmark: (i) it is important to minimize differences between chosen and rejected responses beyond correctness, (ii) evaluating reward models requires multiple comparisons across a wide range of chosen and rejected responses, and (iii) given that reward models encounter responses with diverse representations, responses should be sourced from a variety of models. However, we also observe that a extremely high correlation with degree of overoptimization leads to comparatively lower correlation with certain downstream performance. Thus, when designing a benchmark, it is desirable to use the degree of overoptimization as a useful tool, rather than the end goal.
Authors: Seanie Lee, Sangwoo Park, Dong Bok Lee, Dominik Wagner, Haebin Seong, Tobias Bocklet, Juho Lee, Sung Ju Hwang
Abstract: Low-Rank Adaptation (LoRA), which introduces a product of two trainable low-rank matrices into frozen pre-trained weights, is widely used for efficient fine-tuning of language models in federated learning (FL). However, when combined with differentially private stochastic gradient descent (DP-SGD), LoRA faces substantial noise amplification: DP-SGD perturbs per-sample gradients, and the matrix multiplication of the LoRA update ($BA$) intensifies this effect. Freezing one matrix (e.g., $A$) reduces the noise but restricts model expressiveness, often resulting in suboptimal adaptation. To address this, we propose FedSVD, a simple yet effective method that introduces a global reparameterization based on singular value decomposition (SVD). In our approach, each client optimizes only the $B$ matrix and transmits it to the server. The server aggregates the $B$ matrices, computes the product $BA$ using the previous $A$, and refactorizes the result via SVD. This yields a new adaptive $A$ composed of the orthonormal right singular vectors of $BA$, and an updated $B$ containing the remaining SVD components. This reparameterization avoids quadratic noise amplification, while allowing $A$ to better capture the principal directions of the aggregate updates. Moreover, the orthonormal structure of $A$ bounds the gradient norms of $B$ and preserves more signal under DP-SGD, as confirmed by our theoretical analysis. As a result, FedSVD consistently improves stability and performance across a variety of privacy settings and benchmarks, outperforming relevant baselines under both private and non-private regimes.
Authors: Nishant Suresh Aswani, Saif Eddin Jabari
Abstract: This paper explores a simple question: can we model the internal transformations of a neural network using dynamical systems theory? We introduce Koopman autoencoders to capture how neural representations evolve through network layers, treating these representations as states in a dynamical system. Our approach learns a surrogate model that predicts how neural representations transform from input to output, with two key advantages. First, by way of lifting the original states via an autoencoder, it operates in a linear space, making editing the dynamics straightforward. Second, it preserves the topologies of the original representations by regularizing the autoencoding objective. We demonstrate that these surrogate models naturally replicate the progressive topological simplification observed in neural networks. As a practical application, we show how our approach enables targeted class unlearning in the Yin-Yang and MNIST classification tasks.
Authors: Qin-Cheng Zheng, Shao-Qun Zhang, Shen-Huan Lyu, Yuan Jiang, Zhi-Hua Zhou
Abstract: Isolation Forest (iForest) stands out as a widely-used unsupervised anomaly detector valued for its exceptional runtime efficiency and performance on large-scale tasks. Despite its widespread adoption, a theoretical foundation explaining iForest's success remains unclear. This paper theoretically investigates the conditions and extent of iForest's effectiveness by analyzing its inductive bias through the formulation of depth functions and growth processes. Since directly analyzing the depth function proves intractable due to iForest's random splitting mechanism, we model the growth process of iForest as a random walk, enabling us to derive the expected depth function using transition probabilities. Our case studies reveal key inductive biases: iForest exhibits lower sensitivity to central anomalies while demonstrating greater parameter adaptability compared to $k$-Nearest Neighbor anomaly detectors. Our study provides theoretical understanding of the effectiveness of iForest and establishes a foundation for further theoretical exploration.
Authors: Zheng Wu, Pengzhou Cheng, Zongru Wu, Lingzhong Dong, Zhuosheng Zhang
Abstract: Graphical user interface (GUI) agents have recently emerged as an intriguing paradigm for human-computer interaction, capable of automatically executing user instructions to operate intelligent terminal devices. However, when encountering out-of-distribution (OOD) instructions that violate environmental constraints or exceed the current capabilities of agents, GUI agents may suffer task breakdowns or even pose security threats. Therefore, effective OOD detection for GUI agents is essential. Traditional OOD detection methods perform suboptimally in this domain due to the complex embedding space and evolving GUI environments. In this work, we observe that the in-distribution input semantic space of GUI agents exhibits a clustering pattern with respect to the distance from the centroid. Based on the finding, we propose GEM, a novel method based on fitting a Gaussian mixture model over input embedding distances extracted from the GUI Agent that reflect its capability boundary. Evaluated on eight datasets spanning smartphones, computers, and web browsers, our method achieves an average accuracy improvement of 23.70\% over the best-performing baseline. Analysis verifies the generalization ability of our method through experiments on nine different backbones. The codes are available at https://github.com/Wuzheng02/GEM-OODforGUIagents.
Authors: Kangwen Zhao, Jianfeng Cai, Jinhua Zhu, Ruopei Sun, Dongyun Xue, Wengang Zhou, Li Li, Houqiang Li
Abstract: Reinforcement Learning from Human Feedback relies on reward models to align large language models with human preferences. However, RLHF often suffers from reward hacking, wherein policy learning exploits flaws in the trained reward model to maximize reward scores without genuinely aligning with human preferences. A significant example of such reward hacking is length bias, where reward models usually favor longer responses irrespective of actual response quality. Previous works on length bias have notable limitations, these approaches either mitigate bias without characterizing the bias form, or simply assume a linear length-reward relation. To accurately model the intricate nature of length bias and facilitate more effective bias mitigation, we propose FiMi-RM (Bias Fitting to Mitigate Length Bias of Reward Model in RLHF), a framework that autonomously learns and corrects underlying bias patterns. Our approach consists of three stages: First, we train a standard reward model which inherently contains length bias. Next, we deploy a lightweight fitting model to explicitly capture the non-linear relation between length and reward. Finally, we incorporate this learned relation into the reward model to debias. Experimental results demonstrate that FiMi-RM achieves a more balanced length-reward distribution. Furthermore, when applied to alignment algorithms, our debiased reward model improves length-controlled win rate and reduces verbosity without compromising its performance.
Authors: Zi Liang, Haibo Hu, Qingqing Ye, Yaxin Xiao, Ronghua Li
Abstract: Low rank adaptation (LoRA) has emerged as a prominent technique for fine-tuning large language models (LLMs) thanks to its superb efficiency gains over previous methods. While extensive studies have examined the performance and structural properties of LoRA, its behavior upon training-time attacks remain underexplored, posing significant security risks. In this paper, we theoretically investigate the security implications of LoRA's low-rank structure during fine-tuning, in the context of its robustness against data poisoning and backdoor attacks. We propose an analytical framework that models LoRA's training dynamics, employs the neural tangent kernel to simplify the analysis of the training process, and applies information theory to establish connections between LoRA's low rank structure and its vulnerability against training-time attacks. Our analysis indicates that LoRA exhibits better robustness to backdoor attacks than full fine-tuning, while becomes more vulnerable to untargeted data poisoning due to its over-simplified information geometry. Extensive experimental evaluations have corroborated our theoretical findings.
Authors: Maksim Zhdanov, Nabil Iqbal, Erik Bekkers, Patrick Forr\'e
Abstract: Conformal symmetries, i.e.\ coordinate transformations that preserve angles, play a key role in many fields, including physics, mathematics, computer vision and (geometric) machine learning. Here we build a neural network that is equivariant under general conformal transformations. To achieve this, we lift data from flat Euclidean space to Anti de Sitter (AdS) space. This allows us to exploit a known correspondence between conformal transformations of flat space and isometric transformations on the AdS space. We then build upon the fact that such isometric transformations have been extensively studied on general geometries in the geometric deep learning literature. We employ message-passing layers conditioned on the proper distance, yielding a computationally efficient framework. We validate our model on tasks from computer vision and statistical physics, demonstrating strong performance, improved generalization capacities, and the ability to extract conformal data such as scaling dimensions from the trained network.
Authors: Hao Wang, Jindong Han, Wei Fan, Weijia Zhang, Hao Liu
Abstract: Data Assimilation (DA) plays a critical role in atmospheric science by reconstructing spatially continous estimates of the system state, which serves as initial conditions for scientific analysis. While recent advances in diffusion models have shown great potential for DA tasks, most existing approaches remain purely data-driven and often overlook the physical laws that govern complex atmospheric dynamics. As a result, they may yield physically inconsistent reconstructions that impair downstream applications. To overcome this limitation, we propose PhyDA, a physics-guided diffusion framework designed to ensure physical coherence in atmospheric data assimilation. PhyDA introduces two key components: (1) a Physically Regularized Diffusion Objective that integrates physical constraints into the training process by penalizing deviations from known physical laws expressed as partial differential equations, and (2) a Virtual Reconstruction Encoder that bridges observational sparsity for structured latent representations, further enhancing the model's ability to infer complete and physically coherent states. Experiments on the ERA5 reanalysis dataset demonstrate that PhyDA achieves superior accuracy and better physical plausibility compared to state-of-the-art baselines. Our results emphasize the importance of combining generative modeling with domain-specific physical knowledge and show that PhyDA offers a promising direction for improving real-world data assimilation systems.
Authors: Yuanze Hu, Zhaoxin Fan, Xinyu Wang, Gen Li, Ye Qiu, Zhichao Yang, Wenjun Wu, Kejian Wu, Yifan Sun, Xiaotie Deng, Jin Dong
Abstract: Lightweight Vision-Language Models (VLMs) are indispensable for resource-constrained applications. The prevailing approach to aligning vision and language models involves freezing both the vision encoder and the language model while training small connector modules. However, this strategy heavily depends on the intrinsic capabilities of the language model, which can be suboptimal for lightweight models with limited representational capacity. In this work, we investigate this alignment bottleneck through the lens of mutual information, demonstrating that the constrained capacity of the language model inherently limits the Effective Mutual Information (EMI) between multimodal inputs and outputs, thereby compromising alignment quality. To address this challenge, we propose TinyAlign, a novel framework inspired by Retrieval-Augmented Generation, which strategically retrieves relevant context from a memory bank to enrich multimodal inputs and enhance their alignment. Extensive empirical evaluations reveal that TinyAlign significantly reduces training loss, accelerates convergence, and enhances task performance. Remarkably, it allows models to achieve baseline-level performance with only 40\% of the fine-tuning data, highlighting exceptional data efficiency. Our work thus offers a practical pathway for developing more capable lightweight VLMs while introducing a fresh theoretical lens to better understand and address alignment bottlenecks in constrained multimodal systems.
Authors: Zhiwei Yang, Zeyang Fan, Yihang Lai, Qi Chen, Tian Zhang, Jian Dai, Kun Xu
Abstract: MZI-based block optical neural networks (BONNs), which can achieve large-scale network models, have increasingly drawn attentions. However, the robustness of the current training algorithm is not high enough. Moreover, large-scale BONNs usually contain numerous trainable parameters, resulting in expensive computation and power consumption. In this article, by pruning matrix blocks and directly optimizing the individuals in population, we propose an on-chip covariance matrix adaptation evolution strategy and attention-based pruning (CAP) algorithm for large-scale BONNs. The calculated results demonstrate that the CAP algorithm can prune 60% and 80% of the parameters for MNIST and Fashion-MNIST datasets, respectively, while only degrades the performance by 3.289% and 4.693%. Considering the influence of dynamic noise in phase shifters, our proposed CAP algorithm (performance degradation of 22.327% for MNIST dataset and 24.019% for Fashion-MNIST dataset utilizing a poor fabricated chip and electrical control with a standard deviation of 0.5) exhibits strongest robustness compared with both our previously reported block adjoint training algorithm (43.963% and 41.074%) and the covariance matrix adaptation evolution strategy (25.757% and 32.871%), respectively. Moreover, when 60% of the parameters are pruned, the CAP algorithm realizes 88.5% accuracy in experiment for the simplified MNIST dataset, which is similar to the simulation result without noise (92.1%). Additionally, we simulationally and experimentally demonstrate that using MZIs with only internal phase shifters to construct BONNs is an efficient way to reduce both the system area and the required trainable parameters. Notably, our proposed CAP algorithm show excellent potential for larger-scale network models and more complex tasks.
Authors: Alberto Fern\'andez-Hern\'andez, Jose I. Mestre, Manuel F. Dolz, Jose Duato, Enrique S. Quintana-Ort\'i
Abstract: Initialization plays a critical role in Deep Neural Network training, directly influencing convergence, stability, and generalization. Common approaches such as Glorot and He initializations rely on randomness, which can produce uneven weight distributions across layer connections. In this paper, we introduce the Sinusoidal initialization, a novel deterministic method that employs sinusoidal functions to construct structured weight matrices expressly to improve the spread and balance of weights throughout the network while simultaneously fostering a more uniform, well-conditioned distribution of neuron activation states from the very first forward pass. Because Sinusoidal initialization begins with weights and activations that are already evenly and efficiently utilized, it delivers consistently faster convergence, greater training stability, and higher final accuracy across a wide range of models, including convolutional neural networks, vision transformers, and large language models. On average, our experiments show an increase of 4.8 % in final validation accuracy and 20.9 % in convergence speed. By replacing randomness with structure, this initialization provides a stronger and more reliable foundation for Deep Learning systems.
Authors: Tom George Grigg, Mason Burlage, Oliver Brook Scott, Adam Taouil, Dominique Sydow, Liam Wilbraham
Abstract: Exhaustive virtual screening is highly informative but often intractable against the expensive objective functions involved in modern drug discovery. This problem is exacerbated in combinatorial contexts such as multi-vector expansion, where molecular spaces can quickly become ultra-large. Here, we introduce Scalable Active Learning via Synthon Acquisition (SALSA): a simple algorithm applicable to multi-vector expansion which extends pool-based active learning to non-enumerable spaces by factoring modeling and acquisition over synthon or fragment choices. Through experiments on ligand- and structure-based objectives, we highlight SALSA's sample efficiency, and its ability to scale to spaces of trillions of compounds. Further, we demonstrate application toward multi-parameter objective design tasks on three protein targets - finding SALSA-generated molecules have comparable chemical property profiles to known bioactives, and exhibit greater diversity and higher scores over an industry-leading generative approach.
Authors: Shengsheng Lin, Haojun Chen, Haijie Wu, Chunyun Qiu, Weiwei Lin
Abstract: Sufficiently modeling the correlations among variables (aka channels) is crucial for achieving accurate multivariate time series forecasting (MTSF). In this paper, we propose a novel technique called Temporal Query (TQ) to more effectively capture multivariate correlations, thereby improving model performance in MTSF tasks. Technically, the TQ technique employs periodically shifted learnable vectors as queries in the attention mechanism to capture global inter-variable patterns, while the keys and values are derived from the raw input data to encode local, sample-level correlations. Building upon the TQ technique, we develop a simple yet efficient model named Temporal Query Network (TQNet), which employs only a single-layer attention mechanism and a lightweight multi-layer perceptron (MLP). Extensive experiments demonstrate that TQNet learns more robust multivariate correlations, achieving state-of-the-art forecasting accuracy across 12 challenging real-world datasets. Furthermore, TQNet achieves high efficiency comparable to linear-based methods even on high-dimensional datasets, balancing performance and computational cost. The code is available at: https://github.com/ACAT-SCUT/TQNet.
Authors: Eilon Vaknin Laufer, Boaz Nadler
Abstract: Recovering a low rank matrix from a subset of its entries, some of which may be corrupted, is known as the robust matrix completion (RMC) problem. Existing RMC methods have several limitations: they require a relatively large number of observed entries; they may fail under overparametrization, when their assumed rank is higher than the correct one; and many of them fail to recover even mildly ill-conditioned matrices. In this paper we propose a novel RMC method, denoted $\texttt{RGNMR}$, which overcomes these limitations. $\texttt{RGNMR}$ is a simple factorization-based iterative algorithm, which combines a Gauss-Newton linearization with removal of entries suspected to be outliers. On the theoretical front, we prove that under suitable assumptions, $\texttt{RGNMR}$ is guaranteed exact recovery of the underlying low rank matrix. Our theoretical results improve upon the best currently known for factorization-based methods. On the empirical front, we show via several simulations the advantages of $\texttt{RGNMR}$ over existing RMC methods, and in particular its ability to handle a small number of observed entries, overparameterization of the rank and ill-conditioned matrices.
Authors: Uri Dalal, Meirav Segal, Zvika Ben-Haim, Dan Lahav, Omer Nevo
Abstract: Large language models (LLMs) achieve impressive abilities in numerous domains, but exhibit inconsistent performance in response to minor input changes. Rather than view this as a drawback, in this paper we introduce a novel method for leveraging models' inconsistency to boost Pass@k performance. Specifically, we present a "Variator" agent that generates k variants of a given task and submits one candidate solution for each one. Our variant generation approach is applicable to a wide range of domains as it is task agnostic and compatible with free-form inputs. We demonstrate the efficacy of our agent theoretically using a probabilistic model of the inconsistency effect, and show empirically that it outperforms the baseline on the APPS dataset. Furthermore, we establish that inconsistency persists even in frontier reasoning models across coding and cybersecurity domains, suggesting our method is likely to remain relevant for future model generations.
Authors: James Rowbottom, Stefania Fresca, Pietro Lio, Carola-Bibiane Sch\"onlieb, Nicolas Boull\'e
Abstract: Operator learning is a rapidly growing field that aims to approximate nonlinear operators related to partial differential equations (PDEs) using neural operators. These rely on discretization of input and output functions and are, usually, expensive to train for large-scale problems at high-resolution. Motivated by this, we present a Multi-Level Monte Carlo (MLMC) approach to train neural operators by leveraging a hierarchy of resolutions of function dicretization. Our framework relies on using gradient corrections from fewer samples of fine-resolution data to decrease the computational cost of training while maintaining a high level accuracy. The proposed MLMC training procedure can be applied to any architecture accepting multi-resolution data. Our numerical experiments on a range of state-of-the-art models and test-cases demonstrate improved computational efficiency compared to traditional single-resolution training approaches, and highlight the existence of a Pareto curve between accuracy and computational time, related to the number of samples per resolution.
Authors: Jan Hagnberger, Daniel Musekamp, Mathias Niepert
Abstract: Solving time-dependent Partial Differential Equations (PDEs) using a densely discretized spatial domain is a fundamental problem in various scientific and engineering disciplines, including modeling climate phenomena and fluid dynamics. However, performing these computations directly in the physical space often incurs significant computational costs. To address this issue, several neural surrogate models have been developed that operate in a compressed latent space to solve the PDE. While these approaches reduce computational complexity, they often use Transformer-based attention mechanisms to handle irregularly sampled domains, resulting in increased memory consumption. In contrast, convolutional neural networks allow memory-efficient encoding and decoding but are limited to regular discretizations. Motivated by these considerations, we propose CALM-PDE, a model class that efficiently solves arbitrarily discretized PDEs in a compressed latent space. We introduce a novel continuous convolution-based encoder-decoder architecture that uses an epsilon-neighborhood-constrained kernel and learns to apply the convolution operator to adaptive and optimized query points. We demonstrate the effectiveness of CALM-PDE on a diverse set of PDEs with both regularly and irregularly sampled spatial domains. CALM-PDE is competitive with or outperforms existing baseline methods while offering significant improvements in memory and inference time efficiency compared to Transformer-based methods.
Authors: Xuerui Su, Liya Guo, Yue Wang, Yi Zhu, Zhiming Ma, Zun Wang, Yuting Liu
Abstract: Inference scaling further accelerates Large Language Models (LLMs) toward Artificial General Intelligence (AGI), with large-scale Reinforcement Learning (RL) to unleash long Chain-of-Thought reasoning. Most contemporary reasoning approaches usually rely on handcrafted rule-based reward functions. However, the tarde-offs of exploration and exploitation in RL algorithms involves multiple complex considerations, and the theoretical and empirical impacts of manually designed reward functions remain insufficiently explored. In this paper, we propose Decoupled Group Reward Optimization (DGRO), a general RL algorithm for LLM reasoning. On the one hand, DGRO decouples the traditional regularization coefficient into two independent hyperparameters: one scales the policy gradient term, and the other regulates the distance from the sampling policy. This decoupling not only enables precise control over balancing exploration and exploitation, but also can be seamlessly extended to Online Policy Mirror Descent (OPMD) algorithms in Kimi k1.5 and Direct Reward Optimization. On the other hand, we observe that reward variance significantly affects both convergence speed and final model performance. We conduct both theoretical analysis and extensive empirical validation to assess DGRO, including a detailed ablation study that investigates its performance and optimization dynamics. Experimental results show that DGRO achieves state-of-the-art performance on the Logic dataset with an average accuracy of 96.9\%, and demonstrates strong generalization across mathematical benchmarks.
Authors: Chuanxing Geng, Qifei Li, Xinrui Wang, Dong Liang, Songcan Chen, Pong C. Yuen
Abstract: Using unlabeled wild data containing both in-distribution (ID) and out-of-distribution (OOD) data to improve the safety and reliability of models has recently received increasing attention. Existing methods either design customized losses for labeled ID and unlabeled wild data then perform joint optimization, or first filter out OOD data from the latter then learn an OOD detector. While achieving varying degrees of success, two potential issues remain: (i) Labeled ID data typically dominates the learning of models, inevitably making models tend to fit OOD data as IDs; (ii) The selection of thresholds for identifying OOD data in unlabeled wild data usually faces dilemma due to the unavailability of pure OOD samples. To address these issues, we propose a novel loss-difference OOD detection framework (LoD) by \textit{intentionally label-noisifying} unlabeled wild data. Such operations not only enable labeled ID data and OOD data in unlabeled wild data to jointly dominate the models' learning but also ensure the distinguishability of the losses between ID and OOD samples in unlabeled wild data, allowing the classic clustering technique (e.g., K-means) to filter these OOD samples without requiring thresholds any longer. We also provide theoretical foundation for LoD's viability, and extensive experiments verify its superiority.
Authors: Chengping He, Mingrui Jiang, Keyi Shan, Szu-Hao Yang, Zefan Li, Shengbo Wang, Giacomo Pedretti, Jim Ignowski, Can Li
Abstract: Brain-inspired computing aims to mimic cognitive functions like associative memory, the ability to recall complete patterns from partial cues. Memristor technology offers promising hardware for such neuromorphic systems due to its potential for efficient in-memory analog computing. Hopfield Neural Networks (HNNs) are a classic model for associative memory, but implementations on conventional hardware suffer from efficiency bottlenecks, while prior memristor-based HNNs faced challenges with vulnerability to hardware defects due to offline training, limited storage capacity, and difficulty processing analog patterns. Here we introduce and experimentally demonstrate on integrated memristor hardware a new hardware-adaptive learning algorithm for associative memories that significantly improves defect tolerance and capacity, and naturally extends to scalable multilayer architectures capable of handling both binary and continuous patterns. Our approach achieves 3x effective capacity under 50% device faults compared to state-of-the-art methods. Furthermore, its extension to multilayer architectures enables superlinear capacity scaling (\(\propto N^{1.49}\ for binary patterns) and effective recalling of continuous patterns (\propto N^{1.74}\ scaling), as compared to linear capacity scaling for previous HNNs. It also provides flexibility to adjust capacity by tuning hidden neurons for the same-sized patterns. By leveraging the massive parallelism of the hardware enabled by synchronous updates, it reduces energy by 8.8x and latency by 99.7% for 64-dimensional patterns over asynchronous schemes, with greater improvements at scale. This promises the development of more reliable memristor-based associative memory systems and enables new applications research due to the significantly improved capacity, efficiency, and flexibility.
Authors: Akhila Henry, Nithin Nagaraj
Abstract: This study presents novel Augmented Regression Models using Neurochaos Learning (NL), where Tracemean features derived from the Neurochaos Learning framework are integrated with traditional regression algorithms : Linear Regression, Ridge Regression, Lasso Regression, and Support Vector Regression (SVR). Our approach was evaluated using ten diverse real-life datasets and a synthetically generated dataset of the form $y = mx + c + \epsilon$. Results show that incorporating the Tracemean feature (mean of the chaotic neural traces of the neurons in the NL architecture) significantly enhances regression performance, particularly in Augmented Lasso Regression and Augmented SVR, where six out of ten real-life datasets exhibited improved predictive accuracy. Among the models, Augmented Chaotic Ridge Regression achieved the highest average performance boost (11.35 %). Additionally, experiments on the simulated dataset demonstrated that the Mean Squared Error (MSE) of the augmented models consistently decreased and converged towards the Minimum Mean Squared Error (MMSE) as the sample size increased. This work demonstrates the potential of chaos-inspired features in regression tasks, offering a pathway to more accurate and computationally efficient prediction models.
Authors: Tai Nguyen, Phong Le, Carola Doerr, Nguyen Dang
Abstract: It is well known that evolutionary algorithms can benefit from dynamic choices of the key parameters that control their behavior, to adjust their search strategy to the different stages of the optimization process. A prominent example where dynamic parameter choices have shown a provable super-constant speed-up is the $(1+(\lambda,\lambda))$ Genetic Algorithm optimizing the OneMax function. While optimal parameter control policies result in linear expected running times, this is not possible with static parameter choices. This result has spurred a lot of interest in parameter control policies. However, many works, in particular theoretical running time analyses, focus on controlling one single parameter. Deriving policies for controlling multiple parameters remains very challenging. In this work we reconsider the problem of the $(1+(\lambda,\lambda))$ Genetic Algorithm optimizing OneMax. We decouple its four main parameters and investigate how well state-of-the-art deep reinforcement learning techniques can approximate good control policies. We show that although making deep reinforcement learning learn effectively is a challenging task, once it works, it is very powerful and is able to find policies that outperform all previously known control policies on the same benchmark. Based on the results found through reinforcement learning, we derive a simple control policy that consistently outperforms the default theory-recommended setting by $27\%$ and the irace-tuned policy, the strongest existing control policy on this benchmark, by $13\%$, for all tested problem sizes up to $40{,}000$.
Authors: Douglas Orr, Luka Ribar, Carlo Luschi
Abstract: Weight quantisation is an essential technique for enabling efficient training and deployment of modern deep learning models. However, the recipe book of quantisation formats is large and the formats are often chosen empirically. In this paper, we propose a framework for systematic design and analysis of quantisation formats. By connecting the question of format design with the classical quantisation theory, we show that the strong practical performance of popular formats comes from their ability to represent values using variable-length codes. Framing the optimisation problem as minimising the KL divergence between the original and quantised model outputs, the objective is aligned with minimising the squared quantisation error of the model parameters. We therefore develop and evaluate squared-error-optimal formats for known distributions, observing significant improvement of variable-length codes over fixed-length codes. Uniform quantisation followed by lossless compression with a variable-length code is shown to be optimal. However, we find that commonly used block formats and sparse outlier formats also outperform fixed-length codes, implying they also exploit variable-length encoding. Finally, by using the relationship between the Fisher information and KL divergence, we derive the optimal allocation of bit-widths to individual parameter tensors across the model's layers, saving up to 0.25 bits per parameter when tested with direct-cast quantisation of language models.
Authors: Baohao Liao, Hanze Dong, Yuhui Xu, Doyen Sahoo, Christof Monz, Junnan Li, Caiming Xiong
Abstract: Inference-time scaling techniques have significantly bolstered the reasoning capabilities of large language models (LLMs) by harnessing additional computational effort at inference without retraining. Similarly, Chain-of-Thought (CoT) prompting and its extension, Long CoT, improve accuracy by generating rich intermediate reasoning trajectories, but these approaches incur substantial token costs that impede their deployment in latency-sensitive settings. In this work, we first show that truncated CoT, which stops reasoning before completion and directly generates the final answer, often matches full CoT sampling while using dramatically fewer tokens. Building on this insight, we introduce Fractured Sampling, a unified inference-time strategy that interpolates between full CoT and solution-only sampling along three orthogonal axes: (1) the number of reasoning trajectories, (2) the number of final solutions per trajectory, and (3) the depth at which reasoning traces are truncated. Through extensive experiments on five diverse reasoning benchmarks and several model scales, we demonstrate that Fractured Sampling consistently achieves superior accuracy-cost trade-offs, yielding steep log-linear scaling gains in Pass@k versus token budget. Our analysis reveals how to allocate computation across these dimensions to maximize performance, paving the way for more efficient and scalable LLM reasoning.
Authors: James E. Warner, Tristan A. Shah, Patrick E. Leser, Geoffrey F. Bomarito, Joshua D. Pribe, Michael C. Stanley
Abstract: Deep generative models are promising tools for science and engineering, but their reliance on abundant, high-quality data limits applicability. We present a novel framework for generative modeling of random fields (probability distributions over continuous functions) that incorporates domain knowledge to supplement limited, sparse, and indirect data. The foundation of the approach is latent flow matching, where generative modeling occurs on compressed function representations in the latent space of a pre-trained variational autoencoder (VAE). Innovations include the adoption of a function decoder within the VAE and integration of physical/statistical constraints into the VAE training process. In this way, a latent function representation is learned that yields continuous random field samples satisfying domain-specific constraints when decoded, even in data-limited regimes. Efficacy is demonstrated on two challenging applications: wind velocity field reconstruction from sparse sensors and material property inference from a limited number of indirect measurements. Results show that the proposed framework achieves significant improvements in reconstruction accuracy compared to unconstrained methods and enables effective inference with relatively small training datasets that is intractable without constraints.
Authors: Jiyuan Pei, Yi Mei, Jialin Liu, Mengjie Zhang
Abstract: Meta-Black-Box Optimization (MetaBBO) garners attention due to its success in automating the configuration and generation of black-box optimizers, significantly reducing the human effort required for optimizer design and discovering optimizers with higher performance than classic human-designed optimizers. However, existing MetaBBO methods conduct one-off training under the assumption that a stationary problem distribution with extensive and representative training problem samples is pre-available. This assumption is often impractical in real-world scenarios, where diverse problems following shifting distribution continually arise. Consequently, there is a pressing need for methods that can continuously learn from new problems encountered on-the-fly and progressively enhance their capabilities. In this work, we explore a novel paradigm of lifelong learning in MetaBBO and introduce LiBOG, a novel approach designed to learn from sequentially encountered problems and generate high-performance optimizers for Black-Box Optimization (BBO). LiBOG consolidates knowledge both across tasks and within tasks to mitigate catastrophic forgetting. Extensive experiments demonstrate LiBOG's effectiveness in learning to generate high-performance optimizers in a lifelong learning manner, addressing catastrophic forgetting while maintaining plasticity to learn new tasks.
Authors: Jack Chen, Fazhong Liu, Naruto Liu, Yuhan Luo, Erqu Qin, Harry Zheng, Tian Dong, Haojin Zhu, Yan Meng, Xiao Wang
Abstract: Large language models (LLMs) excel at mathematical reasoning and logical problem-solving. The current popular training paradigms primarily use supervised fine-tuning (SFT) and reinforcement learning (RL) to enhance the models' reasoning abilities. However, when using SFT or RL alone, there are respective challenges: SFT may suffer from overfitting, while RL is prone to mode collapse. The state-of-the-art methods have proposed hybrid training schemes. However, static switching faces challenges such as poor generalization across different tasks and high dependence on data quality. In response to these challenges, inspired by the curriculum learning-quiz mechanism in human reasoning cultivation, We propose SASR, a step-wise adaptive hybrid training framework that theoretically unifies SFT and RL and dynamically balances the two throughout optimization. SASR uses SFT for initial warm-up to establish basic reasoning skills, and then uses an adaptive dynamic adjustment algorithm based on gradient norm and divergence relative to the original distribution to seamlessly integrate SFT with the online RL method GRPO. By monitoring the training status of LLMs and adjusting the training process in sequence, SASR ensures a smooth transition between training schemes, maintaining core reasoning abilities while exploring different paths. Experimental results demonstrate that SASR outperforms SFT, RL, and static hybrid training methods.
Authors: Zihan Gu, Han Zhang, Ruoyu Chen, Yue Hu, Hua Zhang
Abstract: Positional encoding (PE) is essential for enabling Transformers to model sequential structure. However, the mechanisms by which different PE schemes couple token content and positional information-and how these mechanisms influence model dynamics-remain theoretically underexplored. In this work, we present a unified framework that analyzes PE through the spectral properties of Toeplitz and related matrices derived from attention logits. We show that multiplicative content-position coupling-exemplified by Rotary Positional Encoding (RoPE) via a Hadamard product with a Toeplitz matrix-induces spectral contraction, which theoretically improves optimization stability and efficiency. Guided by this theory, we construct synthetic tasks that contrast content-position dependent and content-position independent settings, and evaluate a range of PE methods. Our experiments reveal strong alignment with theory: RoPE consistently outperforms other methods on position-sensitive tasks and induces "single-head deposit" patterns in early layers, indicating localized positional processing. Further analyses show that modifying the method and timing of PE coupling, such as MLA in Deepseek-V3, can effectively mitigate this concentration. These results establish explicit content-relative mixing with relative-position Toeplitz signals as a key principle for effective PE design and provide new insight into how positional structure is integrated in Transformer architectures.
Authors: Vijay Ekambaram, Subodh Kumar, Arindam Jati, Sumanta Mukherjee, Tomoya Sakai, Pankaj Dayama, Wesley M. Gifford, Jayant Kalagnanam
Abstract: The rise of time-series pre-trained models has advanced temporal representation learning, but current state-of-the-art models are often large-scale, requiring substantial compute. We introduce TSPulse, ultra-compact time-series pre-trained models with only 1M parameters, specialized to perform strongly across classification, anomaly detection, imputation, and retrieval tasks. TSPulse introduces innovations at both the architecture and task levels. At the architecture level, it employs a dual-space masked reconstruction, learning from both time and frequency domains to capture complementary signals. This is further enhanced by a dual-embedding disentanglement, generating both detailed embeddings for fine-grained analysis and high-level semantic embeddings for broader task understanding. Notably, TSPulse's semantic embeddings are robust to shifts in time, magnitude, and noise, which is important for robust retrieval. At the task level, TSPulse incorporates TSLens, a fine-tuning component enabling task-specific feature attention. It also introduces a multi-head triangulation technique that correlates deviations from multiple prediction heads, enhancing anomaly detection by fusing complementary model outputs. Additionally, a hybrid mask pretraining is proposed to improves zero-shot imputation by reducing pre-training bias. These architecture and task innovations collectively contribute to TSPulse's significant performance gains: 5-16% on the UEA classification benchmarks, +20% on the TSB-AD anomaly detection leaderboard, +50% in zero-shot imputation, and +25% in time-series retrieval. Remarkably, these results are achieved with just 1M parameters, making TSPulse 10-100X smaller than existing pre-trained models. Its efficiency enables GPU-free inference and rapid pre-training, setting a new standard for efficient time-series pre-trained models. Models will be open-sourced soon.
Authors: Hongrui Kou, Jingkai Li, Ziyu Wang, Zhouhang Lv, Yuxin Zhang, Cheng Wang
Abstract: Accurate prediction of traffic flow parameters and real time identification of congestion states are essential for the efficient operation of intelligent transportation systems. This paper proposes a Periodic Pattern Transformer Network (PPTNet) for traffic flow prediction, integrating periodic pattern extraction with the Transformer architecture, coupled with a fuzzy inference method for real-time congestion identification. Firstly, a high-precision traffic flow dataset (Traffic Flow Dataset for China's Congested Highways and Expressways, TF4CHE) suitable for congested highway scenarios in China is constructed based on drone aerial imagery data. Subsequently, the proposed PPTNet employs Fast Fourier Transform to capture multi-scale periodic patterns and utilizes two-dimensional Inception convolutions to efficiently extract intra and inter periodic features. A Transformer decoder dynamically models temporal dependencies, enabling accurate predictions of traffic density and speed. Finally, congestion probabilities are calculated in real-time using the predicted outcomes via a Mamdani fuzzy inference-based congestion identification module. Experimental results demonstrate that the proposed PPTNet significantly outperforms mainstream traffic prediction methods in prediction accuracy, and the congestion identification module effectively identifies real-time road congestion states, verifying the superiority and practicality of the proposed method in real-world traffic scenarios. Project page: https://github.com/ADSafetyJointLab/PPTNet.
Authors: Gabriel B\'ena, Maxence Faldor, Dan F. M. Goodman, Antoine Cully
Abstract: Cellular automata have long been celebrated for their ability to generate complex behaviors from simple, local rules, with well-known discrete models like Conway's Game of Life proven capable of universal computation. Recent advancements have extended cellular automata into continuous domains, raising the question of whether these systems retain the capacity for universal computation. In parallel, neural cellular automata have emerged as a powerful paradigm where rules are learned via gradient descent rather than manually designed. This work explores the potential of neural cellular automata to develop a continuous Universal Cellular Automaton through training by gradient descent. We introduce a cellular automaton model, objective functions and training strategies to guide neural cellular automata toward universal computation in a continuous setting. Our experiments demonstrate the successful training of fundamental computational primitives - such as matrix multiplication and transposition - culminating in the emulation of a neural network solving the MNIST digit classification task directly within the cellular automata state. These results represent a foundational step toward realizing analog general-purpose computers, with implications for understanding universal computation in continuous dynamics and advancing the automated discovery of complex cellular automata behaviors via machine learning.
Authors: Shmulik Markovich-Golan, Daniel Ohayon, Itay Niv, Yair Hanani
Abstract: Quantization is essential for Neural Network (NN) compression, reducing model size and computational demands by using lower bit-width data types, though aggressive reduction often hampers accuracy. Mixed Precision (MP) mitigates this tradeoff by varying the numerical precision across network layers. This study focuses on automatically selecting an optimal MP configuration within Post-Training Quantization (PTQ) for inference. The first key contribution is a novel sensitivity metric derived from a first-order Taylor series expansion of the loss function as a function of quantization errors in weights and activations. This metric, based on the Mean Square Error (MSE) of the loss, is efficiently calculated per layer using high-precision forward and backward passes over a small calibration dataset. The metric is additive across layers, with low calibration memory overhead as weight optimization is unnecessary. The second contribution is an accurate hardware-aware method for predicting MP time gain by modeling it as additive for sequential sub-graphs. An algorithm partitions the model graph into sequential subgraphs, measuring time gain for each configuration using a few samples. After calibrating per-layer sensitivity and time gain, an Integer Programming (IP) problem is formulated to maximize time gain while keeping loss MSE below a set threshold. Memory gain and theoretical time gain based on Multiply and Accumulate (MAC) operations are also considered. Rigorous experiments on the Intel Gaudi 2 accelerator validate the approach on several Large Language Models (LLMs).
Authors: Jie Yan, Xin Liu, Zhong-Yuan Zhang
Abstract: Federated clustering (FC) aims to discover global cluster structures across decentralized clients without sharing raw data, making privacy preservation a fundamental requirement. There are two critical challenges: (1) privacy leakage during collaboration, and (2) robustness degradation due to aggregation of proxy information from non-independent and identically distributed (Non-IID) local data, leading to inaccurate or inconsistent global clustering. Existing solutions typically rely on model-specific local proxies, which are sensitive to data heterogeneity and inherit inductive biases from their centralized counterparts, thus limiting robustness and generality. We propose Omni Federated Clustering (OmniFC), a unified and model-agnostic framework. Leveraging Lagrange coded computing, our method enables clients to share only encoded data, allowing exact reconstruction of the global distance matrix--a fundamental representation of sample relationships--without leaking private information, even under client collusion. This construction is naturally resilient to Non-IID data distributions. This approach decouples FC from model-specific proxies, providing a unified extension mechanism applicable to diverse centralized clustering methods. Theoretical analysis confirms both reconstruction fidelity and privacy guarantees, while comprehensive experiments demonstrate OmniFC's superior robustness, effectiveness, and generality across various benchmarks compared to state-of-the-art methods. Code will be released.
Authors: Dennis Frauen, Maresa Schr\"oder, Konstantin Hess, Stefan Feuerriegel
Abstract: Estimating heterogeneous treatment effects (HTEs) is crucial for personalized decision-making. However, this task is challenging in survival analysis, which includes time-to-event data with censored outcomes (e.g., due to study dropout). In this paper, we propose a toolbox of novel orthogonal survival learners to estimate HTEs from time-to-event data under censoring. Our learners have three main advantages: (i) we show that learners from our toolbox are guaranteed to be orthogonal and thus come with favorable theoretical properties; (ii) our toolbox allows for incorporating a custom weighting function, which can lead to robustness against different types of low overlap, and (iii) our learners are model-agnostic (i.e., they can be combined with arbitrary machine learning models). We instantiate the learners from our toolbox using several weighting functions and, as a result, propose various neural orthogonal survival learners. Some of these coincide with existing survival learners (including survival versions of the DR- and R-learner), while others are novel and further robust w.r.t. low overlap regimes specific to the survival setting (i.e., survival overlap and censoring overlap). We then empirically verify the effectiveness of our learners for HTE estimation in different low-overlap regimes through numerical experiments. In sum, we provide practitioners with a large toolbox of learners that can be used for randomized and observational studies with censored time-to-event data.
Authors: Xiaoyu Yang, Jie Lu, En Yu
Abstract: This paper uncovers a critical yet overlooked phenomenon in multi-modal large language models (MLLMs): detrimental concept drift within chain-of-thought (CoT) reasoning during non-stationary reinforcement fine-tuning (RFT), where reasoning token distributions evolve unpredictably, thereby introducing significant biases in final predictions. To address this, we are pioneers in establishing the theoretical bridge between concept drift theory and RFT processes by formalizing CoT's autoregressive token streams as non-stationary distributions undergoing arbitrary temporal shifts. Leveraging this framework, we propose a novel counterfact-aware RFT that systematically decouples beneficial distribution adaptation from harmful concept drift through concept graph-empowered LLM experts generating counterfactual reasoning trajectories. Our solution, Counterfactual Preference Optimization (CPO), enables stable RFT in non-stationary environments, particularly within the medical domain, through custom-tuning of counterfactual-aware preference alignment. Extensive experiments demonstrate our superior performance of robustness, generalization and coordination within RFT. Besides, we also contributed a large-scale dataset CXR-CounterFact (CCF), comprising 320,416 meticulously curated counterfactual reasoning trajectories derived from MIMIC-CXR. Our code and data are public.
Authors: Adrien Lagesse, Marc Lelarge
Abstract: We propose a novel benchmarking methodology for graph neural networks (GNNs) based on the graph alignment problem, a combinatorial optimization task that generalizes graph isomorphism by aligning two unlabeled graphs to maximize overlapping edges. We frame this problem as a self-supervised learning task and present several methods to generate graph alignment datasets using synthetic random graphs and real-world graph datasets from multiple domains. For a given graph dataset, we generate a family of graph alignment datasets with increasing difficulty, allowing us to rank the performance of various architectures. Our experiments indicate that anisotropic graph neural networks outperform standard convolutional architectures. To further demonstrate the utility of the graph alignment task, we show its effectiveness for unsupervised GNN pre-training, where the learned node embeddings outperform other positional encodings on three molecular regression tasks and achieve state-of-the-art results on the PCQM4Mv2 dataset with significantly fewer parameters. To support reproducibility and further research, we provide an open-source Python package to generate graph alignment datasets and benchmark new GNN architectures.
Authors: Dennis Frauen, Valentyn Melnychuk, Jonas Schweisthal, Stefan Feuerriegel
Abstract: Decision-making across various fields, such as medicine, heavily relies on conditional average treatment effects (CATEs). Practitioners commonly make decisions by checking whether the estimated CATE is positive, even though the decision-making performance of modern CATE estimators is poorly understood from a theoretical perspective. In this paper, we study optimal decision-making based on two-stage CATE estimators (e.g., DR-learner), which are considered state-of-the-art and widely used in practice. We prove that, while such estimators may be optimal for estimating CATE, they can be suboptimal when used for decision-making. Intuitively, this occurs because such estimators prioritize CATE accuracy in regions far away from the decision boundary, which is ultimately irrelevant to decision-making. As a remedy, we propose a novel two-stage learning objective that retargets the CATE to balance CATE estimation error and decision performance. We then propose a neural method that optimizes an adaptively-smoothed approximation of our learning objective. Finally, we confirm the effectiveness of our method both empirically and theoretically. In sum, our work is the first to show how two-stage CATE estimators can be adapted for optimal decision-making.
Authors: Christodoulos Kechris, Jonathan Dan, David Atienza
Abstract: Traditional saliency map methods, popularized in computer vision, highlight individual points (pixels) of the input that contribute the most to the model's output. However, in time-series they offer limited insights as semantically meaningful features are often found in other domains. We introduce Cross-domain Integrated Gradients, a generalization of Integrated Gradients. Our method enables feature attributions on any domain that can be formulated as an invertible, differentiable transformation of the time domain. Crucially, our derivation extends the original Integrated Gradients into the complex domain, enabling frequency-based attributions. We provide the necessary theoretical guarantees, namely, path independence and completeness. Our approach reveals interpretable, problem-specific attributions that time-domain methods cannot capture, on three real-world tasks: wearable sensor heart rate extraction, electroencephalography-based seizure detection, and zero-shot time-series forecasting. We release an open-source Tensorflow/PyTorch library to enable plug-and-play cross-domain explainability for time-series models. These results demonstrate the ability of cross-domain integrated gradients to provide semantically meaningful insights in time-series models that are impossible with traditional time-domain saliency.
Authors: Ji Qi, Tam Thuc Do, Mingxiao Liu, Zhuoshi Pan, Yuzhe Li, Gene Cheung, H. Vicky Zhao
Abstract: To forecast traffic with both spatial and temporal dimensions, we unroll a mixed-graph-based optimization algorithm into a lightweight and interpretable transformer-like neural net. Specifically, we construct two graphs: an undirected graph $\mathcal{G}^u$ capturing spatial correlations across geography, and a directed graph $\mathcal{G}^d$ capturing sequential relationships over time. We formulate a prediction problem for the future samples of signal $\mathbf{x}$, assuming it is "smooth" with respect to both $\mathcal{G}^u$ and $\mathcal{G}^d$, where we design new $\ell_2$ and $\ell_1$-norm variational terms to quantify and promote signal smoothness (low-frequency reconstruction) on a directed graph. We construct an iterative algorithm based on alternating direction method of multipliers (ADMM), and unroll it into a feed-forward network for data-driven parameter learning. We insert graph learning modules for $\mathcal{G}^u$ and $\mathcal{G}^d$, which are akin to the self-attention mechanism in classical transformers. Experiments show that our unrolled networks achieve competitive traffic forecast performance as state-of-the-art prediction schemes, while reducing parameter counts drastically. Our code is available in https://github.com/SingularityUndefined/Unrolling-GSP-STForecast.
URLs: https://github.com/SingularityUndefined/Unrolling-GSP-STForecast.
Authors: Guangda Liu, Chengwei Li, Zhenyu Ning, Jing Lin, Yiwu Yao, Danning Ke, Minyi Guo, Jieru Zhao
Abstract: Large language models (LLMs) have been widely deployed with rapidly expanding context windows to support increasingly demanding applications. However, long contexts pose significant deployment challenges, primarily due to the KV cache whose size grows proportionally with context length. While KV cache compression methods are proposed to address this issue, KV dropping methods incur considerable accuracy loss, and KV retrieval methods suffer from significant efficiency bottlenecks. We propose FreeKV, an algorithm-system co-optimization framework to enhance KV retrieval efficiency while preserving accuracy. On the algorithm side, FreeKV introduces speculative retrieval to shift the KV selection and recall processes out of the critical path, combined with fine-grained correction to ensure accuracy. On the system side, FreeKV employs hybrid KV layouts across CPU and GPU memory to eliminate fragmented data transfers, and leverages double-buffered streamed recall to further improve efficiency. Experiments demonstrate that FreeKV achieves near-lossless accuracy across various scenarios and models, delivering up to 13$\times$ speedup compared to SOTA KV retrieval methods.
Authors: Sungmin Cha, Kyunghyun Cho
Abstract: Knowledge distillation (KD) is a core component in the training and deployment of modern generative models, particularly large language models (LLMs). While its empirical benefits are well documented--enabling smaller student models to emulate the performance of much larger teachers--the underlying mechanisms by which KD improves generative quality remain poorly understood. In this work, we present a minimal working explanation of KD in generative modeling. Using a controlled simulation with mixtures of Gaussians, we demonstrate that distillation induces a trade-off between precision and recall in the student model. As the teacher distribution becomes more selective, the student concentrates more probability mass on high-likelihood regions at the expense of coverage--a behavior modulated by a single entropy-controlling parameter. We then validate this effect in a large-scale language modeling setup using the SmolLM2 family of models. Empirical results reveal the same precision-recall dynamics observed in simulation, where precision corresponds to sample quality and recall to distributional coverage. This precision-recall trade-off proves especially beneficial in scenarios where sample quality outweighs diversity, such as instruction tuning or downstream generation. Our analysis provides a simple and general explanation for the effectiveness of KD in generative modeling.
Authors: Kathrin Lammers, Valerie Vaquet, Barbara Hammer
Abstract: As machine learning is increasingly applied in an online fashion to deal with evolving data streams, the fairness of these algorithms is a matter of growing ethical and legal concern. In many use cases, class imbalance in the data also needs to be dealt with to ensure predictive performance. Current fairness-aware stream learners typically attempt to solve these issues through in- or post-processing by focusing on optimizing one specific discrimination metric, addressing class imbalance in a separate processing step. While C-SMOTE is a highly effective model-agnostic pre-processing approach to mitigate class imbalance, as a side effect of this method, algorithmic bias is often introduced. Therefore, we propose CFSMOTE - a fairness-aware, continuous SMOTE variant - as a pre-processing approach to simultaneously address the class imbalance and fairness concerns by employing situation testing and balancing fairness-relevant groups during oversampling. Unlike other fairness-aware stream learners, CFSMOTE is not optimizing for only one specific fairness metric, therefore avoiding potentially problematic trade-offs. Our experiments show significant improvement on several common group fairness metrics in comparison to vanilla C-SMOTE while maintaining competitive performance, also in comparison to other fairness-aware algorithms.
Authors: Fran\c{c}ois Bachoc (IMT), J\'er\^ome Bolte (TSE-R), Ryan Boustany (TSE-R), Jean-Michel Loubes (IMT)
Abstract: Despite growing empirical evidence of bias amplification in machine learning, its theoretical foundations remain poorly understood. We develop a formal framework for majority-minority learning tasks, showing how standard training can favor majority groups and produce stereotypical predictors that neglect minority-specific features. Assuming population and variance imbalance, our analysis reveals three key findings: (i) the close proximity between ``full-data'' and stereotypical predictors, (ii) the dominance of a region where training the entire model tends to merely learn the majority traits, and (iii) a lower bound on the additional training required. Our results are illustrated through experiments in deep learning for tabular and image classification tasks.
Authors: Francesco Innocenti, El Mehdi Achour, Christopher L. Buckley
Abstract: The biological implausibility of backpropagation (BP) has motivated many alternative, brain-inspired algorithms that attempt to rely only on local information, such as predictive coding (PC) and equilibrium propagation. However, these algorithms have notoriously struggled to train very deep networks, preventing them from competing with BP in large-scale settings. Indeed, scaling PC networks (PCNs) has recently been posed as a challenge for the community (Pinchetti et al., 2024). Here, we show that 100+ layer PCNs can be trained reliably using a Depth-$\mu$P parameterisation (Yang et al., 2023; Bordelon et al., 2023) which we call "$\mu$PC". Through an extensive analysis of the scaling behaviour of PCNs, we reveal several pathologies that make standard PCNs difficult to train at large depths. We then show that, despite addressing only some of these instabilities, $\mu$PC allows stable training of very deep (up to 128-layer) residual networks on simple classification tasks with competitive performance and little tuning compared to current benchmarks. Moreover, $\mu$PC enables zero-shot transfer of both weight and activity learning rates across widths and depths. Our results have implications for other local algorithms and could be extended to convolutional and transformer architectures. Code for $\mu$PC is made available as part of a JAX library for PCNs at https://github.com/thebuckleylab/jpc (Innocenti et al., 2024).
Authors: Emile van Krieken, Pasquale Minervini, Edoardo Ponti, Antonio Vergari
Abstract: Neurosymbolic (NeSy) predictors combine neural perception with symbolic reasoning to solve tasks like visual reasoning. However, standard NeSy predictors assume conditional independence between the symbols they extract, thus limiting their ability to model interactions and uncertainty - often leading to overconfident predictions and poor out-of-distribution generalisation. To overcome the limitations of the independence assumption, we introduce neurosymbolic diffusion models (NeSyDMs), a new class of NeSy predictors that use discrete diffusion to model dependencies between symbols. Our approach reuses the independence assumption from NeSy predictors at each step of the diffusion process, enabling scalable learning while capturing symbol dependencies and uncertainty quantification. Across both synthetic and real-world benchmarks - including high-dimensional visual path planning and rule-based autonomous driving - NeSyDMs achieve state-of-the-art accuracy among NeSy predictors and demonstrate strong calibration.
Authors: Yunhao Ni, Yuhe Liu, Wenxin Sun, Yitong Tang, Yuxin Guo, Peilin Feng, Wenjun Wu, Lei Huang
Abstract: Universal approximation theorem (UAT) is a fundamental theory for deep neural networks (DNNs), demonstrating their powerful representation capacity to represent and approximate any function. The analyses and proofs of UAT are based on traditional network with only linear and nonlinear activation functions, but omitting normalization layers, which are commonly employed to enhance the training of modern networks. This paper conducts research on UAT of DNNs with normalization layers for the first time. We theoretically prove that an infinitely wide network -- composed solely of parallel layer normalization (PLN) and linear layers -- has universal approximation capacity. Additionally, we investigate the minimum number of neurons required to approximate $L$-Lipchitz continuous functions, with a single hidden-layer network. We compare the approximation capacity of PLN with traditional activation functions in theory. Different from the traditional activation functions, we identify that PLN can act as both activation function and normalization in deep neural networks at the same time. We also find that PLN can improve the performance when replacing LN in transformer architectures, which reveals the potential of PLN used in neural architectures.
Authors: Dongsu Lee, Minhae Kwon
Abstract: The goal of offline reinforcement learning (RL) is to extract a high-performance policy from the fixed datasets, minimizing performance degradation due to out-of-distribution (OOD) samples. Offline model-based RL (MBRL) is a promising approach that ameliorates OOD issues by enriching state-action transitions with augmentations synthesized via a learned dynamics model. Unfortunately, seminal offline MBRL methods often struggle in sparse-reward, long-horizon tasks. In this work, we introduce a novel MBRL framework, dubbed Temporal Distance-Aware Transition Augmentation (TempDATA), that generates augmented transitions in a temporally structured latent space rather than in raw state space. To model long-horizon behavior, TempDATA learns a latent abstraction that captures a temporal distance from both trajectory and transition levels of state space. Our experiments confirm that TempDATA outperforms previous offline MBRL methods and achieves matching or surpassing the performance of diffusion-based trajectory augmentation and goal-conditioned RL on the D4RL AntMaze, FrankaKitchen, CALVIN, and pixel-based FrankaKitchen.
Authors: Maksim Bobrin, Ilya Zisman, Alexander Nikulin, Vladislav Kurenkov, Dmitry Dylov
Abstract: Behavioral Foundation Models (BFMs) proved successful in producing policies for arbitrary tasks in a zero-shot manner, requiring no test-time training or task-specific fine-tuning. Among the most promising BFMs are the ones that estimate the successor measure learned in an unsupervised way from task-agnostic offline data. However, these methods fail to react to changes in the dynamics, making them inefficient under partial observability or when the transition function changes. This hinders the applicability of BFMs in a real-world setting, e.g., in robotics, where the dynamics can unexpectedly change at test time. In this work, we demonstrate that Forward-Backward (FB) representation, one of the methods from the BFM family, cannot distinguish between distinct dynamics, leading to an interference among the latent directions, which parametrize different policies. To address this, we propose a FB model with a transformer-based belief estimator, which greatly facilitates zero-shot adaptation. We also show that partitioning the policy encoding space into dynamics-specific clusters, aligned with the context-embedding directions, yields additional gain in performance. These traits allow our method to respond to the dynamics observed during training and to generalize to unseen ones. Empirically, in the changing dynamics setting, our approach achieves up to a 2x higher zero-shot returns compared to the baselines for both discrete and continuous tasks.
Authors: Sara Alosaime (University of Warwick), Arshad Jhumka (University of Leeds)
Abstract: Federated Learning (FL) is a privacy-preserving machine learning technique that allows decentralized collaborative model training across a set of distributed clients, by avoiding raw data exchange. A fundamental component of FL is the selection of a subset of clients in each round for model training by a central server. Current selection strategies are myopic in nature in that they are based on past or current interactions, often leading to inefficiency issues such as straggling clients. In this paper, we address this serious shortcoming by proposing the RIFLES approach that builds a novel availability forecasting layer to support the client selection process. We make the following contributions: (i) we formalise the sequential selection problem and reduce it to a scheduling problem and show that the problem is NP-complete, (ii) leveraging heartbeat messages from clients, RIFLES build an availability prediction layer to support (long term) selection decisions, (iii) we propose a novel adaptive selection strategy to support efficient learning and resource usage. To circumvent the inherent exponential complexity, we present RIFLES, a heuristic that leverages clients' historical availability data by using a CNN-LSTM time series forecasting model, allowing the server to predict the optimal participation times of clients, thereby enabling informed selection decisions. By comparing against other FL techniques, we show that RIFLES provide significant improvement by between 10%-50% on a variety of metrics such as accuracy and test loss. To the best of our knowledge, it is the first work to investigate FL as a scheduling problem.
Authors: Juntian Zhu, Miguel de Carvalho, Zhouwang Yang, Fengxiang He
Abstract: An AI agent might surprisingly find she has reached an unknown state which she has never been aware of -- an unknown unknown. We mathematically ground this scenario in reinforcement learning: an agent, after taking an action calculated from value functions $Q$ and $V$ defined on the {\it {aware domain}}, reaches a state out of the domain. To enable the agent to handle this scenario, we propose an {\it episodic Markov decision {process} with growing awareness} (EMDP-GA) model, taking a new {\it noninformative value expansion} (NIVE) approach to expand value functions to newly aware areas: when an agent arrives at an unknown unknown, value functions $Q$ and $V$ whereon are initialised by noninformative beliefs -- the averaged values on the aware domain. This design is out of respect for the complete absence of knowledge in the newly discovered state. The upper confidence bound momentum Q-learning is then adapted to the growing awareness for training the EMDP-GA model. We prove that (1) the regret of our approach is asymptotically consistent with the state of the art (SOTA) without exposure to unknown unknowns in an extremely uncertain environment, and (2) our computational complexity and space complexity are comparable with the SOTA -- these collectively suggest that though an unknown unknown is surprising, it will be asymptotically properly discovered with decent speed and an affordable cost.
Authors: Christoph J\"urgen Hemmer, Daniel Durstewitz
Abstract: Complex, temporally evolving phenomena, from climate to brain activity, are governed by dynamical systems (DS). DS reconstruction (DSR) seeks to infer generative surrogate models of these from observed data, reproducing their long-term behavior. Existing DSR approaches require purpose-training for any new system observed, lacking the zero-shot and in-context inference capabilities known from LLMs. Here we introduce DynaMix, a novel multivariate ALRNN-based mixture-of-experts architecture pre-trained for DSR, the first DSR model able to generalize zero-shot to out-of-domain DS. Just from a provided context signal, without any re-training, DynaMix faithfully forecasts the long-term evolution of novel DS where existing time series (TS) foundation models, like Chronos, fail -- at a fraction of the number of parameters and orders of magnitude faster inference times. DynaMix outperforms TS foundation models in terms of long-term statistics, and often also short-term forecasts, even on real-world time series, like traffic or weather data, typically used for training and evaluating TS models, but not at all part of DynaMix' training corpus. We illustrate some of the failure modes of TS models for DSR problems, and conclude that models built on DS principles may bear a huge potential also for advancing the TS prediction field.
Authors: Pranav Vaidhyanathan, Lucas Schorling, Natalia Ares, Michael A. Osborne
Abstract: We introduce Velocity-Regularized Adam (VRAdam), a physics-inspired optimizer for training deep neural networks that draws on ideas from quartic terms for kinetic energy with its stabilizing effects on various system dynamics. Previous algorithms, including the ubiquitous Adam, operate at the so called adaptive edge of stability regime during training leading to rapid oscillations and slowed convergence of loss. However, VRAdam adds a higher order penalty on the learning rate based on the velocity such that the algorithm automatically slows down whenever weight updates become large. In practice, we observe that the effective dynamic learning rate shrinks in high-velocity regimes, damping oscillations and allowing for a more aggressive base step size when necessary without divergence. By combining this velocity-based regularizer for global damping with per-parameter scaling of Adam to create a hybrid optimizer, we demonstrate that VRAdam consistently exceeds the performance against standard optimizers including AdamW. We benchmark various tasks such as image classification, language modeling, image generation and generative modeling using diverse architectures and training methodologies including Convolutional Neural Networks (CNNs), Transformers, and GFlowNets.
Authors: Stephen Zhang, Suryanarayana Maddu, Xiaoje Qiu, Victor Chard\`es
Abstract: Time-resolved single-cell omics data offers high-throughput, genome-wide measurements of cellular states, which are instrumental to reverse-engineer the processes underpinning cell fate. Such technologies are inherently destructive, allowing only cross-sectional measurements of the underlying stochastic dynamical system. Furthermore, cells may divide or die in addition to changing their molecular state. Collectively these present a major challenge to inferring realistic biophysical models. We present a novel approach, \emph{unbalanced} probability flow inference, that addresses this challenge for biological processes modelled as stochastic dynamics with growth. By leveraging a Lagrangian formulation of the Fokker-Planck equation, our method accurately disentangles drift from intrinsic noise and growth. We showcase the applicability of our approach through evaluation on a range of simulated and real single-cell RNA-seq datasets. Comparing to several existing methods, we find our method achieves higher accuracy while enjoying a simple two-step training scheme.
Authors: Francesco D'Amico, Dario Bocchi, Matteo Negri
Abstract: Scaling laws in deep learning - empirical power-law relationships linking model performance to resource growth - have emerged as simple yet striking regularities across architectures, datasets, and tasks. These laws are particularly impactful in guiding the design of state-of-the-art models, since they quantify the benefits of increasing data or model size, and hint at the foundations of interpretability in machine learning. However, most studies focus on asymptotic behavior at the end of training or on the optimal training time given the model size. In this work, we uncover a richer picture by analyzing the entire training dynamics through the lens of spectral complexity norms. We identify two novel dynamical scaling laws that govern how performance evolves during training. These laws together recover the well-known test error scaling at convergence, offering a mechanistic explanation of generalization emergence. Our findings are consistent across CNNs, ResNets, and Vision Transformers trained on MNIST, CIFAR-10 and CIFAR-100. Furthermore, we provide analytical support using a solvable model: a single-layer perceptron trained with binary cross-entropy. In this setting, we show that the growth of spectral complexity driven by the implicit bias mirrors the generalization behavior observed at fixed norm, allowing us to connect the performance dynamics to classical learning rules in the perceptron.
Authors: Yuan-Zheng Lei, Yaobang Gong, Dianwei Chen, Yao Cheng, Xianfeng Terry Yang
Abstract: Physics-informed machine learning (PIML) is crucial in modern traffic flow modeling because it combines the benefits of both physics-based and data-driven approaches. In conventional PIML, physical information is typically incorporated by constructing a hybrid loss function that combines data-driven loss and physics loss through linear scalarization. The goal is to find a trade-off between these two objectives to improve the accuracy of model predictions. However, from a mathematical perspective, linear scalarization is limited to identifying only the convex region of the Pareto front, as it treats data-driven and physics losses as separate objectives. Given that most PIML loss functions are non-convex, linear scalarization restricts the achievable trade-off solutions. Moreover, tuning the weighting coefficients for the two loss components can be both time-consuming and computationally challenging. To address these limitations, this paper introduces a paradigm shift in PIML by reformulating the training process as a multi-objective optimization problem, treating data-driven loss and physics loss independently. We apply several multi-gradient descent algorithms (MGDAs), including traditional multi-gradient descent (TMGD) and dual cone gradient descent (DCGD), to explore the Pareto front in this multi-objective setting. These methods are evaluated on both macroscopic and microscopic traffic flow models. In the macroscopic case, MGDAs achieved comparable performance to traditional linear scalarization methods. Notably, in the microscopic case, MGDAs significantly outperformed their scalarization-based counterparts, demonstrating the advantages of a multi-objective optimization approach in complex PIML scenarios.
Authors: Le Vu Anh, Dinh Duc Nha Nguyen, Phi Long Nguyen
Abstract: Large Language Models (LLMs) have become foundational in modern artificial intelligence, powering a wide range of applications from code generation and virtual assistants to scientific research and enterprise automation. However, concerns about data contamination--where test data overlaps with training data--have raised serious questions about the reliability of these applications. Despite awareness of this issue, existing methods fall short in effectively identifying or mitigating contamination. In this paper, we propose Residual-Noise Fingerprinting (RN-F), a novel framework for detecting contaminated data in LLMs. RN-F is a single-pass, gradient-free detection method that leverages residual signal patterns without introducing additional floating-point operations. Our approach is lightweight, model-agnostic, and efficient. We evaluate RN-F on multiple LLMs across various contaminated datasets and show that it consistently outperforms existing state-of-the-art methods, achieving performance improvements of up to 10.5% in contamination detection metrics.
Authors: Carlos Rodriguez-Pardo, Louis Daumas, Leonardo Chiani, Massimo Tavoni
Abstract: Climate-economic modeling under uncertainty presents significant computational challenges that may limit policymakers' ability to address climate change effectively. This paper explores neural network-based approaches for solving high-dimensional optimal control problems arising from models that incorporate ambiguity aversion in climate mitigation decisions. We develop a continuous-time endogenous-growth economic model that accounts for multiple mitigation pathways, including emission-free capital and carbon intensity reductions. Given the inherent complexity and high dimensionality of these models, traditional numerical methods become computationally intractable. We benchmark several neural network architectures against finite-difference generated solutions, evaluating their ability to capture the dynamic interactions between uncertainty, technology transitions, and optimal climate policy. Our findings demonstrate that appropriate neural architecture selection significantly impacts both solution accuracy and computational efficiency when modeling climate-economic systems under uncertainty. These methodological advances enable more sophisticated modeling of climate policy decisions, allowing for better representation of technology transitions and uncertainty-critical elements for developing effective mitigation strategies in the face of climate change.
Authors: Anthony Zhou, Amir Barati Farimani
Abstract: Many architectures for neural PDE surrogates have been proposed in recent years, largely based on neural networks or operator learning. In this work, we derive and propose a new architecture, the Neural Functional, which learns function to scalar mappings. Its implementation leverages insights from operator learning and neural fields, and we show the ability of neural functionals to implicitly learn functional derivatives. For the first time, this allows for an extension of Hamiltonian mechanics to neural PDE surrogates by learning the Hamiltonian functional and optimizing its functional derivatives. We demonstrate that the Hamiltonian Neural Functional can be an effective surrogate model through improved stability and conserving energy-like quantities on 1D and 2D PDEs. Beyond PDEs, functionals are prevalent in physics; functional approximation and learning with its gradients may find other uses, such as in molecular dynamics or design optimization.
Authors: Elias Collaert, Abel Rodr\'iguez, Sander Joos, Lieven Desmet, Vera Rimmer
Abstract: Despite significant advancements in the area, adversarial robustness remains a critical challenge in systems employing machine learning models. The removal of adversarial perturbations at inference time, known as adversarial purification, has emerged as a promising defense strategy. To achieve this, state-of-the-art methods leverage diffusion models that inject Gaussian noise during a forward process to dilute adversarial perturbations, followed by a denoising step to restore clean samples before classification. In this work, we propose FlowPure, a novel purification method based on Continuous Normalizing Flows (CNFs) trained with Conditional Flow Matching (CFM) to learn mappings from adversarial examples to their clean counterparts. Unlike prior diffusion-based approaches that rely on fixed noise processes, FlowPure can leverage specific attack knowledge to improve robustness under known threats, while also supporting a more general stochastic variant trained on Gaussian perturbations for settings where such knowledge is unavailable. Experiments on CIFAR-10 and CIFAR-100 demonstrate that our method outperforms state-of-the-art purification-based defenses in preprocessor-blind and white-box scenarios, and can do so while fully preserving benign accuracy in the former. Moreover, our results show that not only is FlowPure a highly effective purifier but it also holds a strong potential for adversarial detection, identifying preprocessor-blind PGD samples with near-perfect accuracy.
Authors: Alonso Urbano, David W. Romero, Max Zimmer, Sebastian Pokutta
Abstract: Real-world data often exhibits unknown or approximate symmetries, yet existing equivariant networks must commit to a fixed transformation group prior to training, e.g., continuous $SO(2)$ rotations. This mismatch degrades performance when the actual data symmetries differ from those in the transformation group. We introduce RECON, a framework to discover each input's intrinsic symmetry distribution from unlabeled data. RECON leverages class-pose decompositions and applies a data-driven normalization to align arbitrary reference frames into a common natural pose, yielding directly comparable and interpretable symmetry descriptors. We demonstrate effective symmetry discovery on 2D image benchmarks and -- for the first time -- extend it to 3D transformation groups, paving the way towards more flexible equivariant modeling.
Authors: Yifu Cai, Xinyu Li, Mononito Goswami, Micha{\l} Wili\'nski, Gus Welter, Artur Dubrawski
Abstract: We introduce TimeSeriesGym, a scalable benchmarking framework for evaluating Artificial Intelligence (AI) agents on time series machine learning engineering challenges. Existing benchmarks lack scalability, focus narrowly on model building in well-defined settings, and evaluate only a limited set of research artifacts (e.g., CSV submission files). To make AI agent benchmarking more relevant to the practice of machine learning engineering, our framework scales along two critical dimensions. First, recognizing that effective ML engineering requires a range of diverse skills, TimeSeriesGym incorporates challenges from diverse sources spanning multiple domains and tasks. We design challenges to evaluate both isolated capabilities (including data handling, understanding research repositories, and code translation) and their combinations, and rather than addressing each challenge independently, we develop tools that support designing multiple challenges at scale. Second, we implement evaluation mechanisms for multiple research artifacts, including submission files, code, and models, using both precise numeric measures and more flexible LLM-based evaluation approaches. This dual strategy balances objective assessment with contextual judgment. Although our initial focus is on time series applications, our framework can be readily extended to other data modalities, broadly enhancing the comprehensiveness and practical utility of agentic AI evaluation. We open-source our benchmarking framework to facilitate future research on the ML engineering capabilities of AI agents.
Authors: Hengli Li, Chenxi Li, Tong Wu, Xuekai Zhu, Yuxuan Wang, Zhaoxin Yu, Eric Hanchen Jiang, Song-Chun Zhu, Zixia Jia, Ying Nian Wu, Zilong Zheng
Abstract: Reasoning ability, a core component of human intelligence, continues to pose a significant challenge for Large Language Models (LLMs) in the pursuit of AGI. Although model performance has improved under the training scaling law, significant challenges remain, particularly with respect to training algorithms, such as catastrophic forgetting, and the limited availability of novel training data. As an alternative, test-time scaling enhances reasoning performance by increasing test-time computation without parameter updating. Unlike prior methods in this paradigm focused on token space, we propose leveraging latent space for more effective reasoning and better adherence to the test-time scaling law. We introduce LatentSeek, a novel framework that enhances LLM reasoning through Test-Time Instance-level Adaptation (TTIA) within the model's latent space. Specifically, LatentSeek leverages policy gradient to iteratively update latent representations, guided by self-generated reward signals. LatentSeek is evaluated on a range of reasoning benchmarks, including GSM8K, MATH-500, and AIME2024, across multiple LLM architectures. Results show that LatentSeek consistently outperforms strong baselines, such as Chain-of-Thought prompting and fine-tuning-based methods. Furthermore, our analysis demonstrates that LatentSeek is highly efficient, typically converging within a few iterations for problems of average complexity, while also benefiting from additional iterations, thereby highlighting the potential of test-time scaling in the latent space. These findings position LatentSeek as a lightweight, scalable, and effective solution for enhancing the reasoning capabilities of LLMs.
Authors: Reza T. Batley, Sourav Saha
Abstract: Contemporary models of high dimensional physical systems are constrained by the curse of dimensionality and a reliance on dense data. We introduce KHRONOS (Kernel Expansion Hierarchy for Reduced Order, Neural Optimized Surrogates), an AI framework for model based, model free and model inversion tasks. KHRONOS constructs continuously differentiable target fields with a hierarchical composition of per-dimension kernel expansions, which are tensorized into modes and then superposed. We evaluate KHRONOS on a canonical 2D, Poisson equation benchmark: across 16 to 512 degrees of freedom (DoFs), it obtained L2 square errors of 5e-4 down to 6e-10. This represents a 100 time gain over Kolmogorov Arnold Networks (which itself reports a 100 times improvement on MLPs/PINNs with 100 times fewer parameters) when controlling for the number of parameters. This also represents a 1e4 times improvement in L2 square error compared to standard linear FEM at comparable DoFs. Inference complexity is dominated by inner products, yielding sub-millisecond full-field predictions that scale to an arbitrary resolution. For inverse problems, KHRONOS facilitates rapid, iterative level set recovery in only a few forward evaluations, with sub-microsecond per sample latency. KHRONOS scalability, expressivity, and interpretability open new avenues in constrained edge computing, online control, computer vision, and beyond.
Authors: Song-Lin Li, Rui Zhu, Yu-Feng Li, Lan-Zhe Guo
Abstract: Semi-supervised learning (SSL) alleviates the cost of data labeling process by exploiting unlabeled data, and has achieved promising results on various tasks such as image classification. Meanwhile, the Pretrain-Finetuning paradigm has garnered significant attention in recent years, and exploiting pre-trained models could also reduce the requirement of labeled data in downstream tasks. Therefore, a question naturally occurs: \emph{When the labeled data is scarce in the target tasks, should we exploit unlabeled data or pre-trained models?} To answer this question, we select pre-trained Vision-Language Models (VLMs) as representative pretrain-finetuning instances and propose \textit{Few-shot SSL} -- a framework that enables fair comparison between these two paradigms by controlling the amount of labeled data used. Extensive experiments across various settings demonstrate that pre-trained VLMs generally outperform SSL methods in nearly all cases, except when the data has low resolution or lacks clear semantic structure. Therefore, we encourage future SSL research to compare with pre-trained models and explore deeper integration, such as using pre-trained knowledge to enhance pseudo-labeling. To support future research, we release our unified reproduction and evaluation framework. Codes are available at https://anonymous.4open.science/r/Rethinking-SSL-and-Pretrain-Finetuning-5566
URLs: https://anonymous.4open.science/r/Rethinking-SSL-and-Pretrain-Finetuning-5566
Authors: Yuhang Wang, Youhe Jiang, Bin Cui, Fangcheng Fu
Abstract: Recent advances in test-time scaling suggest that Large Language Models (LLMs) can gain better capabilities by generating Chain-of-Thought reasoning (analogous to human thinking) to respond a given request, and meanwhile exploring more reasoning branches (i.e., generating multiple responses and ensembling them) can improve the final output quality. However, when incorporating the two scaling dimensions, we find that the system efficiency is dampened significantly for two reasons. Firstly, the time cost to generate the final output increases substantially as many reasoning branches would be trapped in the over-thinking dilemma, producing excessively long responses. Secondly, generating multiple reasoning branches for each request increases memory consumption, which is unsuitable for LLM serving since we can only batch a limited number of requests to process simultaneously. To address this, we present SART, a serving framework for efficient and accurate LLM reasoning. The essential idea is to manage the thinking to be short and right, rather than long. For one thing, we devise a redundant sampling with early stopping approach based on empirical observations and theoretic analysis, which increases the likelihood of obtaining short-thinking responses when sampling reasoning branches. For another, we propose to dynamically prune low-quality branches so that only right-thinking branches are maintained, reducing the memory consumption and allowing us to batch more requests. Experimental results demonstrate that SART not only improves the accuracy of LLM reasoning but also enhances the serving efficiency, outperforming existing methods by up to 28.2 times and on average 15.7 times in terms of efficiency when achieving the same level of accuracy.
Authors: Yuval Grinberg, Nimrod Harel, Jacob Goldberger, Ofir Lindenbaum
Abstract: Falsely annotated samples, also known as noisy labels, can significantly harm the performance of deep learning models. Two main approaches for learning with noisy labels are global noise estimation and data filtering. Global noise estimation approximates the noise across the entire dataset using a noise transition matrix, but it can unnecessarily adjust correct labels, leaving room for local improvements. Data filtering, on the other hand, discards potentially noisy samples but risks losing valuable data. Our method identifies potentially noisy samples based on their loss distribution. We then apply a selection process to separate noisy and clean samples and learn a noise transition matrix to correct the loss for noisy samples while leaving the clean data unaffected, thereby improving the training process. Our approach ensures robust learning and enhanced model performance by preserving valuable information from noisy samples and refining the correction process. We applied our method to standard image datasets (MNIST, CIFAR-10, and CIFAR-100) and a biological scRNA-seq cell-type annotation dataset. We observed a significant improvement in model accuracy and robustness compared to traditional methods.
Authors: Andrej \v{C}op, Bla\v{z} Bertalani\v{c}, Marko Grobelnik, Carolina Fortuna
Abstract: As the complexity and number of machine learning (ML) models grows, well-documented ML models are essential for developers and companies to use or adapt them to their specific use cases. Model metadata, already present in unstructured format as model cards in online repositories such as Hugging Face, could be more structured and machine readable while also incorporating environmental impact metrics such as energy consumption and carbon footprint. Our work extends the existing State of the Art by defining a structured schema for ML model metadata focusing on machine-readable format and support for integration into a knowledge graph (KG) for better organization and querying, enabling a wider set of use cases. Furthermore, we present an example wireless localization model metadata dataset consisting of 22 models trained on 4 datasets, integrated into a Neo4j-based KG with 113 nodes and 199 relations.
Authors: Shuqing Luo (Katie), Pingzhi Li (Katie), Jie Peng (Katie), Hanrui Wang (Katie), Yang (Katie), Zhao (Kevin), Yu (Kevin), Cao, Yu Cheng, Tianlong Chen
Abstract: Mixture-of-experts (MoE) architectures could achieve impressive computational efficiency with expert parallelism, which relies heavily on all-to-all communication across devices. Unfortunately, such communication overhead typically constitutes a significant portion of the total runtime, hampering the scalability of distributed training and inference for modern MoE models (consuming over $40\%$ runtime in large-scale training). In this paper, we first define collaborative communication to illustrate this intrinsic limitation, and then propose system- and algorithm-level innovations to reduce communication costs. Specifically, given a pair of experts co-activated by one token, we call them "collaborated", which comprises $2$ cases as intra- and inter-collaboration, depending on whether they are kept on the same device. Our pilot investigations reveal that augmenting the proportion of intra-collaboration can accelerate expert parallelism at scale. It motivates us to strategically optimize collaborative communication for accelerated MoE training and inference, dubbed Occult. Our designs are capable of either delivering exact results with reduced communication cost or controllably minimizing the cost with collaboration pruning, materialized by modified fine-tuning. Comprehensive experiments on various MoE-LLMs demonstrate that Occult can be faster than popular state-of-the-art inference or training frameworks (more than $1.5\times$ speed up across multiple tasks and models) with comparable or superior quality compared to the standard fine-tuning. Code is available at $\href{https://github.com/UNITES-Lab/Occult}{https://github.com/UNITES-Lab/Occult}$.
URLs: https://github.com/UNITES-Lab/Occult, https://github.com/UNITES-Lab/Occult
Authors: Nimrod Berman, Ilan Naiman, Moshe Eliasof, Hedi Zisling, Omri Azencot
Abstract: Diffusion-based generative models have demonstrated exceptional performance, yet their iterative sampling procedures remain computationally expensive. A prominent strategy to mitigate this cost is distillation, with offline distillation offering particular advantages in terms of efficiency, modularity, and flexibility. In this work, we identify two key observations that motivate a principled distillation framework: (1) while diffusion models have been viewed through the lens of dynamical systems theory, powerful and underexplored tools can be further leveraged; and (2) diffusion models inherently impose structured, semantically coherent trajectories in latent space. Building on these observations, we introduce the Koopman Distillation Model KDM, a novel offline distillation approach grounded in Koopman theory-a classical framework for representing nonlinear dynamics linearly in a transformed space. KDM encodes noisy inputs into an embedded space where a learned linear operator propagates them forward, followed by a decoder that reconstructs clean samples. This enables single-step generation while preserving semantic fidelity. We provide theoretical justification for our approach: (1) under mild assumptions, the learned diffusion dynamics admit a finite-dimensional Koopman representation; and (2) proximity in the Koopman latent space correlates with semantic similarity in the generated outputs, allowing for effective trajectory alignment. Empirically, KDM achieves state-of-the-art performance across standard offline distillation benchmarks, improving FID scores by up to 40% in a single generation step. All implementation details and code for the experimental setups are provided in our GitHub - https://github.com/azencot-group/KDM, or in our project page - https://sites.google.com/view/koopman-distillation-model.
URLs: https://github.com/azencot-group/KDM,, https://sites.google.com/view/koopman-distillation-model.
Authors: Yasi Zhang, Tianyu Chen, Zhendong Wang, Ying Nian Wu, Mingyuan Zhou, Oscar Leong
Abstract: Learning generative models from corrupted data is a fundamental yet persistently challenging task across scientific disciplines, particularly when access to clean data is limited or expensive. Denoising Score Distillation (DSD) \cite{chen2025denoising} recently introduced a novel and surprisingly effective strategy that leverages score distillation to train high-fidelity generative models directly from noisy observations. Building upon this foundation, we propose \textit{Restoration Score Distillation} (RSD), a principled generalization of DSD that accommodates a broader range of corruption types, such as blurred, incomplete, or low-resolution images. RSD operates by first pretraining a teacher diffusion model solely on corrupted data and subsequently distilling it into a single-step generator that produces high-quality reconstructions. Empirically, RSD consistently surpasses its teacher model across diverse restoration tasks on both natural and scientific datasets. Moreover, beyond standard diffusion objectives, the RSD framework is compatible with several corruption-aware training techniques such as Ambient Tweedie, Ambient Diffusion, and its Fourier-space variant, enabling flexible integration with recent advances in diffusion modeling. Theoretically, we demonstrate that in a linear regime, RSD recovers the eigenspace of the clean data covariance matrix from linear measurements, thereby serving as an implicit regularizer. This interpretation recasts score distillation not only as a sampling acceleration technique but as a principled approach to enhancing generative performance in severely degraded data regimes.
Authors: Benoit Dherin, Michael Munn, Hanna Mazzawi, Michael Wunder, Sourabh Medapati, Javier Gonzalvo
Abstract: Modern deep learning algorithms use variations of gradient descent as their main learning methods. Gradient descent can be understood as the simplest Ordinary Differential Equation (ODE) solver; namely, the Euler method applied to the gradient flow differential equation. Since Euler, many ODE solvers have been devised that follow the gradient flow equation more precisely and more stably. Runge-Kutta (RK) methods provide a family of very powerful explicit and implicit high-order ODE solvers. However, these higher-order solvers have not found wide application in deep learning so far. In this work, we evaluate the performance of higher-order RK solvers when applied in deep learning, study their limitations, and propose ways to overcome these drawbacks. In particular, we explore how to improve their performance by naturally incorporating key ingredients of modern neural network optimizers such as preconditioning, adaptive learning rates, and momentum.
Authors: Matan Abudy, Orr Well, Emmanuel Chemla, Roni Katzir, Nur Lan
Abstract: State-of-the-art neural networks can be trained to become remarkable solutions to many problems. But while these architectures can express symbolic, perfect solutions, trained models often arrive at approximations instead. We show that the choice of regularization method plays a crucial role: when trained on formal languages with standard regularization ($L_1$, $L_2$, or none), expressive architectures not only fail to converge to correct solutions but are actively pushed away from perfect initializations. In contrast, applying the Minimum Description Length (MDL) principle to balance model complexity with data fit provides a theoretically grounded regularization method. Using MDL, perfect solutions are selected over approximations, independently of the optimization algorithm. We propose that unlike existing regularization techniques, MDL introduces the appropriate inductive bias to effectively counteract overfitting and promote generalization.
Authors: Gabriel Malikal, Ismail Alkhouri, Alvaro Velasquez, Adam M Alessio, Saiprasad Ravishankar
Abstract: The Maximum Cut (MaxCut) problem is NP-Complete, and obtaining its optimal solution is NP-hard in the worst case. As a result, heuristic-based algorithms are commonly used, though their design often requires significant domain expertise. More recently, learning-based methods trained on large (un)labeled datasets have been proposed; however, these approaches often struggle with generalizability and scalability. A well-known approximation algorithm for MaxCut is the Goemans-Williamson (GW) algorithm, which relaxes the Quadratic Unconstrained Binary Optimization (QUBO) formulation into a semidefinite program (SDP). The GW algorithm then applies hyperplane rounding by uniformly sampling a random hyperplane to convert the SDP solution into binary node assignments. In this paper, we propose a training-data-free approach based on a non-episodic reinforcement learning formulation, in which an agent learns to select improved rounding hyperplanes that yield better cuts than those produced by the GW algorithm. By optimizing over a Markov Decision Process (MDP), our method consistently achieves better cuts across large-scale graphs with varying densities and degree distributions.
Authors: Dongyi Wang, Yuanwei Jiang, Zhenyi Zhang, Xiang Gu, Peijie Zhou, Jian Sun
Abstract: Learning the underlying dynamics of single cells from snapshot data has gained increasing attention in scientific and machine learning research. The destructive measurement technique and cell proliferation/death result in unpaired and unbalanced data between snapshots, making the learning of the underlying dynamics challenging. In this paper, we propose joint Velocity-Growth Flow Matching (VGFM), a novel paradigm that jointly learns state transition and mass growth of single-cell populations via flow matching. VGFM builds an ideal single-cell dynamics containing velocity of state and growth of mass, driven by a presented two-period dynamic understanding of the static semi-relaxed optimal transport, a mathematical tool that seeks the coupling between unpaired and unbalanced data. To enable practical usage, we approximate the ideal dynamics using neural networks, forming our joint velocity and growth matching framework. A distribution fitting loss is also employed in VGFM to further improve the fitting performance for snapshot data. Extensive experimental results on both synthetic and real datasets demonstrate that VGFM can capture the underlying biological dynamics accounting for mass and state variations over time, outperforming existing approaches for single-cell dynamics modeling.
Authors: Artem Riabinin, Egor Shulgin, Kaja Gruntkowska, Peter Richt\'arik
Abstract: Recent developments in deep learning optimization have brought about radically new algorithms based on the Linear Minimization Oracle (LMO) framework, such as $\sf Muon$ and $\sf Scion$. After over a decade of $\sf Adam$'s dominance, these LMO-based methods are emerging as viable replacements, offering several practical advantages such as improved memory efficiency, better hyperparameter transferability, and most importantly, superior empirical performance on large-scale tasks, including LLM training. However, a significant gap remains between their practical use and our current theoretical understanding: prior analyses (1) overlook the layer-wise LMO application of these optimizers in practice, and (2) rely on an unrealistic smoothness assumption, leading to impractically small stepsizes. To address both, we propose a new LMO-based method called $\sf Gluon$, capturing prior theoretically analyzed methods as special cases, and introduce a new refined generalized smoothness model that captures the layer-wise geometry of neural networks, matches the layer-wise practical implementation of $\sf Muon$ and $\sf Scion$, and leads to convergence guarantees with strong practical predictive power. Unlike prior results, our theoretical stepsizes closely match the fine-tuned values reported by Pethick et al. (2025). Our experiments with NanoGPT and CNN confirm that our assumption holds along the optimization trajectory, ultimately closing the gap between theory and practice.
Authors: Si-Yang Liu, Qile Zhou, Han-Jia Ye
Abstract: Tabular data, a fundamental data format in machine learning, is predominantly utilized in competitions and real-world applications. The performance of tabular models--such as gradient boosted decision trees and neural networks--can vary significantly across datasets due to differences in feature distributions and task characteristics. Achieving top performance on each dataset often requires specialized expert knowledge. To address this variability, practitioners often aggregate the predictions of multiple models. However, conventional aggregation strategies typically rely on static combination rules and lack instance-level adaptability. In this work, we propose an in-context ensemble framework for tabular prediction that leverages large language models (LLMs) to perform dynamic, instance-specific integration of external model predictions. Without access to raw tabular features or semantic information, our method constructs a context around each test instance using its nearest neighbors and the predictions from a pool of external models. Within this enriched context, we introduce Chain of Tabular Thoughts (CoT$^2$), a prompting strategy that guides LLMs through multi-step, interpretable reasoning, making still further progress toward expert-level decision-making. Experimental results show that our method outperforms well-tuned baselines and standard ensemble techniques across a wide range of tabular datasets.
Authors: Zhi-Hao Tan, Zi-Chen Zhao, Hao-Yu Shi, Xin-Yu Zhang, Peng Tan, Yang Yu, Zhi-Hua Zhou
Abstract: The learnware paradigm offers a novel approach to machine learning by enabling users to reuse a set of well-trained models for tasks beyond the models' original purposes. It eliminates the need to build models from scratch, instead relying on specifications (representations of a model's capabilities) to identify and leverage the most suitable models for new tasks. While learnware has proven effective in many scenarios, its application to language models has remained largely unexplored. At the same time, large language models (LLMs) have demonstrated remarkable universal question-answering abilities, yet they face challenges in specialized scenarios due to data scarcity, privacy concerns, and high computational costs, thus more and more specialized small language models (SLMs) are being trained for specific domains. To address these limitations systematically, the learnware paradigm provides a promising solution by enabling maximum utilization of specialized SLMs, and allowing users to identify and reuse them in a collaborative and privacy-preserving manner. This paper presents a preliminary attempt to apply the learnware paradigm to language models. We simulated a learnware system comprising approximately 100 learnwares of specialized SLMs with 8B parameters, fine-tuned across finance, healthcare, and mathematics domains. Each learnware contains an SLM and a specification, which enables users to identify the most relevant models without exposing their own data. Experimental results demonstrate promising performance: by selecting one suitable learnware for each task-specific inference, the system outperforms the base SLMs on all benchmarks. Compared to LLMs, the system outperforms Qwen1.5-110B, Qwen2.5-72B, and Llama3.1-70B-Instruct by at least 14% in finance domain tasks, and surpasses Flan-PaLM-540B (ranked 7th on the Open Medical LLM Leaderboard) in medical domain tasks.
Authors: Sifeng Shang, Jiayi Zhou, Chenyu Lin, Minxian Li, Kaiyang Zhou
Abstract: As the size of large language models grows exponentially, GPU memory has become a bottleneck for adapting these models to downstream tasks. In this paper, we aim to push the limits of memory-efficient training by minimizing memory usage on model weights, gradients, and optimizer states, within a unified framework. Our idea is to eliminate both gradients and optimizer states using zeroth-order optimization, which approximates gradients by perturbing weights during forward passes to identify gradient directions. To minimize memory usage on weights, we employ model quantization, e.g., converting from bfloat16 to int4. However, directly applying zeroth-order optimization to quantized weights is infeasible due to the precision gap between discrete weights and continuous gradients, which would otherwise require de-quantization and re-quantization. To overcome this challenge, we propose Quantized Zeroth-order Optimization (QZO), a novel approach that perturbs the continuous quantization scale for gradient estimation and uses a directional derivative clipping method to stabilize training. QZO is orthogonal to both scalar-based and codebook-based post-training quantization methods. Compared to full-parameter fine-tuning in bfloat16, QZO can reduce the total memory cost by more than 18$\times$ for 4-bit LLMs, and enables fine-tuning Llama-2-13B and Stable Diffusion 3.5 Large within a single 24GB GPU.
Authors: Meshi Bashari, Roy Maor Lotan, Yonghoon Lee, Edgar Dobriban, Yaniv Romano
Abstract: Conformal prediction is a framework for predictive inference with a distribution-free, finite-sample guarantee. However, it tends to provide uninformative prediction sets when calibration data are scarce. This paper introduces Synthetic-powered predictive inference (SPPI), a novel framework that incorporates synthetic data -- e.g., from a generative model -- to improve sample efficiency. At the core of our method is a score transporter: an empirical quantile mapping that aligns nonconformity scores from trusted, real data with those from synthetic data. By carefully integrating the score transporter into the calibration process, SPPI provably achieves finite-sample coverage guarantees without making any assumptions about the real and synthetic data distributions. When the score distributions are well aligned, SPPI yields substantially tighter and more informative prediction sets than standard conformal prediction. Experiments on image classification and tabular regression demonstrate notable improvements in predictive efficiency in data-scarce settings.
Authors: Penghui Qi, Zichen Liu, Tianyu Pang, Chao Du, Wee Sun Lee, Min Lin
Abstract: Scaling test-time compute is crucial for enhancing the reasoning capabilities of large language models (LLMs). Existing approaches typically employ reinforcement learning (RL) to maximize a verifiable reward obtained at the end of reasoning traces. However, such methods optimize only the final performance under a large and fixed token budget, which hinders efficiency in both training and deployment. In this work, we present a novel framework, AnytimeReasoner, to optimize anytime reasoning performance, which aims to improve token efficiency and the flexibility of reasoning under varying token budget constraints. To achieve this, we truncate the complete thinking process to fit within sampled token budgets from a prior distribution, compelling the model to summarize the optimal answer for each truncated thinking for verification. This introduces verifiable dense rewards into the reasoning process, facilitating more effective credit assignment in RL optimization. We then optimize the thinking and summary policies in a decoupled manner to maximize the cumulative reward. Additionally, we introduce a novel variance reduction technique, Budget Relative Policy Optimization (BRPO), to enhance the robustness and efficiency of the learning process when reinforcing the thinking policy. Empirical results in mathematical reasoning tasks demonstrate that our method consistently outperforms GRPO across all thinking budgets under various prior distributions, enhancing both training and token efficiency.
Authors: Dulhan Jayalath, Gilad Landau, Oiwi Parker Jones
Abstract: Despite major advances in surgical brain-to-text (B2T), i.e. transcribing speech from invasive brain recordings, non-invasive alternatives have yet to surpass even chance on standard metrics. This remains a barrier to building a non-invasive brain-computer interface (BCI) capable of restoring communication in paralysed individuals without surgery. Here, we present the first non-invasive B2T result that significantly exceeds these critical baselines, raising BLEU by $1.4\mathrm{-}2.6\times$ over prior work. This result is driven by three contributions: (1) we extend recent word-classification models with LLM-based rescoring, transforming single-word predictors into closed-vocabulary B2T systems; (2) we introduce a predictive in-filling approach to handle out-of-vocabulary (OOV) words, substantially expanding the effective vocabulary; and (3) we demonstrate, for the first time, how to scale non-invasive B2T models across datasets, unlocking deep learning at scale and improving accuracy by $2.1\mathrm{-}2.3\times$. Through these contributions, we offer new insights into the roles of data quality and vocabulary size. Together, our results remove a major obstacle to realising practical non-invasive B2T systems.
Authors: Zhengyang Geng, Mingyang Deng, Xingjian Bai, J. Zico Kolter, Kaiming He
Abstract: We propose a principled and effective framework for one-step generative modeling. We introduce the notion of average velocity to characterize flow fields, in contrast to instantaneous velocity modeled by Flow Matching methods. A well-defined identity between average and instantaneous velocities is derived and used to guide neural network training. Our method, termed the MeanFlow model, is self-contained and requires no pre-training, distillation, or curriculum learning. MeanFlow demonstrates strong empirical performance: it achieves an FID of 3.43 with a single function evaluation (1-NFE) on ImageNet 256x256 trained from scratch, significantly outperforming previous state-of-the-art one-step diffusion/flow models. Our study substantially narrows the gap between one-step diffusion/flow models and their multi-step predecessors, and we hope it will motivate future research to revisit the foundations of these powerful models.
Authors: Botian Xu, Feng Gao, Chao Yu, Ruize Zhang, Yi Wu, Yu Wang
Abstract: In this work, we introduce OmniDrones, an efficient and flexible platform tailored for reinforcement learning in drone control, built on Nvidia's Omniverse Isaac Sim. It employs a bottom-up design approach that allows users to easily design and experiment with various application scenarios on top of GPU-parallelized simulations. It also offers a range of benchmark tasks, presenting challenges ranging from single-drone hovering to over-actuated system tracking. In summary, we propose an open-sourced drone simulation platform, equipped with an extensive suite of tools for drone learning. It includes 4 drone models, 5 sensor modalities, 4 control modes, over 10 benchmark tasks, and a selection of widely used RL baselines. To showcase the capabilities of OmniDrones and to support future research, we also provide preliminary results on these benchmark tasks. We hope this platform will encourage further studies on applying RL to practical drone systems.
Authors: Yuhang Huang, JIazhao Zhang, Shilong Zou, XInwang Liu, Ruizhen Hu, Kai Xu
Abstract: Predictive manipulation has recently gained considerable attention in the Embodied AI community due to its potential to improve robot policy performance by leveraging predicted states. However, generating accurate future visual states of robot-object interactions from world models remains a well-known challenge, particularly in achieving high-quality pixel-level representations. To this end, we propose LaDi-WM, a world model that predicts the latent space of future states using diffusion modeling. Specifically, LaDi-WM leverages the well-established latent space aligned with pre-trained Visual Foundation Models (VFMs), which comprises both geometric features (DINO-based) and semantic features (CLIP-based). We find that predicting the evolution of the latent space is easier to learn and more generalizable than directly predicting pixel-level images. Building on LaDi-WM, we design a diffusion policy that iteratively refines output actions by incorporating forecasted states, thereby generating more consistent and accurate results. Extensive experiments on both synthetic and real-world benchmarks demonstrate that LaDi-WM significantly enhances policy performance by 27.9\% on the LIBERO-LONG benchmark and 20\% on the real-world scenario. Furthermore, our world model and policies achieve impressive generalizability in real-world experiments.
Authors: Dan Luo, Jinyu Zhou, Le Xu, Sisi Yuan, Xuan Lin
Abstract: Predicting drug-target binding affinity (DTA) is essential for identifying potential therapeutic candidates in drug discovery. However, most existing models rely heavily on static protein structures, often overlooking the dynamic nature of proteins, which is crucial for capturing conformational flexibility that will be beneficial for protein binding interactions. We introduce DynamicDTA, an innovative deep learning framework that incorporates static and dynamic protein features to enhance DTA prediction. The proposed DynamicDTA takes three types of inputs, including drug sequence, protein sequence, and dynamic descriptors. A molecular graph representation of the drug sequence is generated and subsequently processed through graph convolutional network, while the protein sequence is encoded using dilated convolutions. Dynamic descriptors, such as root mean square fluctuation, are processed through a multi-layer perceptron. These embedding features are fused with static protein features using cross-attention, and a tensor fusion network integrates all three modalities for DTA prediction. Extensive experiments on three datasets demonstrate that DynamicDTA achieves by at least 3.4% improvement in RMSE score with comparison to seven state-of-the-art baseline methods. Additionally, predicting novel drugs for Human Immunodeficiency Virus Type 1 and visualizing the docking complexes further demonstrates the reliability and biological relevance of DynamicDTA.
Authors: Yuhang Wang, Abdulaziz Alhuraish, Shuyi Wang, Hao Zhou
Abstract: Leveraging a newly released open dataset of Lane Keeping Assist (LKA) systems from production vehicles, this paper presents the first comprehensive empirical analysis of real-world LKA performance. Our study yields three key findings: (i) LKA failures can be systematically categorized into perception, planning, and control errors. We present representative examples of each failure mode through in-depth analysis of LKA-related CAN signals, enabling both justification of the failure mechanisms and diagnosis of when and where each module begins to degrade; (ii) LKA systems tend to follow a fixed lane-centering strategy, often resulting in outward drift that increases linearly with road curvature, whereas human drivers proactively steer slightly inward on similar curved segments; (iii) We provide the first statistical summary and distribution analysis of environmental and road conditions under LKA failures, identifying with statistical significance that faded lane markings, low pavement laneline contrast, and sharp curvature are the most dominant individual factors, along with critical combinations that substantially increase failure likelihood. Building on these insights, we propose a theoretical model that integrates road geometry, speed limits, and LKA steering capability to inform infrastructure design. Additionally, we develop a machine learning-based model to assess roadway readiness for LKA deployment, offering practical tools for safer infrastructure planning, especially in rural areas. This work highlights key limitations of current LKA systems and supports the advancement of safer and more reliable autonomous driving technologies.
Authors: Yuhang Wang, Hao Zhou
Abstract: Lane Keeping Assist systems, while increasingly prevalent, often suffer from unpredictable real-world failures, largely due to their opaque, black-box nature, which limits driver anticipation and trust. To bridge the gap between automated assistance and effective human oversight, we present LKAlert, a novel supervisory alert system that leverages VLM to forecast potential LKA risk 1-3 seconds in advance. LKAlert processes dash-cam video and CAN data, integrating surrogate lane segmentation features from a parallel interpretable model as automated guiding attention. Unlike traditional binary classifiers, LKAlert issues both predictive alert and concise natural language explanation, enhancing driver situational awareness and trust. To support the development and evaluation of such systems, we introduce OpenLKA-Alert, the first benchmark dataset designed for predictive and explainable LKA failure warnings. It contains synchronized multimodal inputs and human-authored justifications across annotated temporal windows. We further contribute a generalizable methodological framework for VLM-based black-box behavior prediction, combining surrogate feature guidance with LoRA. This framework enables VLM to reason over structured visual context without altering its vision backbone, making it broadly applicable to other complex, opaque systems requiring interpretable oversight. Empirical results correctly predicts upcoming LKA failures with 69.8% accuracy and a 58.6\% F1-score. The system also generates high-quality textual explanations for drivers (71.7 ROUGE-L) and operates efficiently at approximately 2 Hz, confirming its suitability for real-time, in-vehicle use. Our findings establish LKAlert as a practical solution for enhancing the safety and usability of current ADAS and offer a scalable paradigm for applying VLMs to human-centered supervision of black-box automation.
Authors: Jose Fuentes, Ines Ortega-Fernandez, Nora M. Villanueva, Marta Sestelo
Abstract: User and Entity Behaviour Analytics (UEBA) is a broad branch of data analytics that attempts to build a normal behavioural profile in order to detect anomalous events. Among the techniques used to detect anomalies, Deep Autoencoders constitute one of the most promising deep learning models on UEBA tasks, allowing explainable detection of security incidents that could lead to the leak of personal data, hijacking of systems, or access to sensitive business information. In this study, we introduce the first implementation of an explainable UEBA-based anomaly detection framework that leverages Deep Autoencoders in combination with Doc2Vec to process both numerical and textual features. Additionally, based on the theoretical foundations of neural networks, we offer a novel proof demonstrating the equivalence of two widely used definitions for fully-connected neural networks. The experimental results demonstrate the proposed framework capability to detect real and synthetic anomalies effectively generated from real attack data, showing that the models provide not only correct identification of anomalies but also explainable results that enable the reconstruction of the possible origin of the anomaly. Our findings suggest that the proposed UEBA framework can be seamlessly integrated into enterprise environments, complementing existing security systems for explainable threat detection.
Authors: Muzun Althunayyan, Amir Javed, Omer Rana
Abstract: Connected and Autonomous Vehicles (CAVs) enhance mobility but face cybersecurity threats, particularly through the insecure Controller Area Network (CAN) bus. Cyberattacks can have devastating consequences in connected vehicles, including the loss of control over critical systems, necessitating robust security solutions. In-vehicle Intrusion Detection Systems (IDSs) offer a promising approach by detecting malicious activities in real time. This survey provides a comprehensive review of state-of-the-art research on learning-based in-vehicle IDSs, focusing on Machine Learning (ML), Deep Learning (DL), and Federated Learning (FL) approaches. Based on the reviewed studies, we critically examine existing IDS approaches, categorising them by the types of attacks they detect - known, unknown, and combined known-unknown attacks - while identifying their limitations. We also review the evaluation metrics used in research, emphasising the need to consider multiple criteria to meet the requirements of safety-critical systems. Additionally, we analyse FL-based IDSs and highlight their limitations. By doing so, this survey helps identify effective security measures, address existing limitations, and guide future research toward more resilient and adaptive protection mechanisms, ensuring the safety and reliability of CAVs.
Authors: Stylianos Stasinos, Martino Mensio, Elena Lazovik, Athanasios Trantas
Abstract: Biodiversity research requires complete and detailed information to study ecosystem dynamics at different scales. Employing data-driven methods like Machine Learning is getting traction in ecology and more specific biodiversity, offering alternative modelling pathways. For these methods to deliver accurate results there is the need for large, curated and multimodal datasets that offer granular spatial and temporal resolutions. In this work, we introduce BioCube, a multimodal, fine-grained global dataset for ecology and biodiversity research. BioCube incorporates species observations through images, audio recordings and descriptions, environmental DNA, vegetation indices, agricultural, forest, land indicators, and high-resolution climate variables. All observations are geospatially aligned under the WGS84 geodetic system, spanning from 2000 to 2020. The dataset will become available at https://huggingface.co/datasets/BioDT/BioCube while the acquisition and processing code base at https://github.com/BioDT/bfm-data.
URLs: https://huggingface.co/datasets/BioDT/BioCube, https://github.com/BioDT/bfm-data.
Authors: Zeynep Engin, David Hand
Abstract: As AI systems evolve from static tools to dynamic agents, traditional categorical governance frameworks -- based on fixed risk tiers, levels of autonomy, or human oversight models -- are increasingly insufficient on their own. Systems built on foundation models, self-supervised learning, and multi-agent architectures increasingly blur the boundaries that categories were designed to police. In this Perspective, we make the case for dimensional governance: a framework that tracks how decision authority, process autonomy, and accountability (the 3As) distribute dynamically across human-AI relationships. A critical advantage of this approach is its ability to explicitly monitor system movement toward and across key governance thresholds, enabling preemptive adjustments before risks materialize. This dimensional approach provides the necessary foundation for more adaptive categorization, enabling thresholds and classifications that can evolve with emerging capabilities. While categories remain essential for decision-making, building them upon dimensional foundations allows for context-specific adaptability and stakeholder-responsive governance that static approaches cannot achieve. We outline key dimensions, critical trust thresholds, and practical examples illustrating where rigid categorical frameworks fail -- and where a dimensional mindset could offer a more resilient and future-proof path forward for both governance and innovation at the frontier of artificial intelligence.
Authors: Akarsh Kumar, Jeff Clune, Joel Lehman, Kenneth O. Stanley
Abstract: Much of the excitement in modern AI is driven by the observation that scaling up existing systems leads to better performance. But does better performance necessarily imply better internal representations? While the representational optimist assumes it must, this position paper challenges that view. We compare neural networks evolved through an open-ended search process to networks trained via conventional stochastic gradient descent (SGD) on the simple task of generating a single image. This minimal setup offers a unique advantage: each hidden neuron's full functional behavior can be easily visualized as an image, thus revealing how the network's output behavior is internally constructed neuron by neuron. The result is striking: while both networks produce the same output behavior, their internal representations differ dramatically. The SGD-trained networks exhibit a form of disorganization that we term fractured entangled representation (FER). Interestingly, the evolved networks largely lack FER, even approaching a unified factored representation (UFR). In large models, FER may be degrading core model capacities like generalization, creativity, and (continual) learning. Therefore, understanding and mitigating FER could be critical to the future of representation learning.
Authors: Asher Moldwin, Amarda Shehu
Abstract: This paper surveys foundation models for AI-enabled biological design, focusing on recent developments in applying large-scale, self-supervised models to tasks such as protein engineering, small molecule design, and genomic sequence design. Though this domain is evolving rapidly, this survey presents and discusses a taxonomy of current models and methods. The focus is on challenges and solutions in adapting these models for biological applications, including biological sequence modeling architectures, controllability in generation, and multi-modal integration. The survey concludes with a discussion of open problems and future directions, offering concrete next-steps to improve the quality of biological sequence generation.
Authors: Pengrui Quan, Brian Wang, Kang Yang, Liying Han, Mani Srivastava
Abstract: Spatiotemporal reasoning plays a key role in Cyber-Physical Systems (CPS). Despite advances in Large Language Models (LLMs) and Large Reasoning Models (LRMs), their capacity to reason about complex spatiotemporal signals remains underexplored. This paper proposes a hierarchical SpatioTemporal reAsoning benchmaRK, STARK, to systematically evaluate LLMs across three levels of reasoning complexity: state estimation (e.g., predicting field variables, localizing and tracking events in space and time), spatiotemporal reasoning over states (e.g., inferring spatial-temporal relationships), and world-knowledge-aware reasoning that integrates contextual and domain knowledge (e.g., intent prediction, landmark-aware navigation). We curate 26 distinct spatiotemporal tasks with diverse sensor modalities, comprising 14,552 challenges where models answer directly or by Python Code Interpreter. Evaluating 3 LRMs and 8 LLMs, we find LLMs achieve limited success in tasks requiring geometric reasoning (e.g., multilateration or triangulation), particularly as complexity increases. Surprisingly, LRMs show robust performance across tasks with various levels of difficulty, often competing or surpassing traditional first-principle-based methods. Our results show that in reasoning tasks requiring world knowledge, the performance gap between LLMs and LRMs narrows, with some LLMs even surpassing LRMs. However, the LRM o3 model continues to achieve leading performance across all evaluated tasks, a result attributed primarily to the larger size of the reasoning models. STARK motivates future innovations in model architectures and reasoning paradigms for intelligent CPS by providing a structured framework to identify limitations in the spatiotemporal reasoning of LLMs and LRMs.
Authors: Michael L. Wells, Kamel Lahouel, Bruno Jedynak
Abstract: We present a novel kernel-based method for learning multivariate stochastic differential equations (SDEs). The method follows a two-step procedure: we first estimate the drift term function, then the (matrix-valued) diffusion function given the drift. Occupation kernels are integral functionals on a reproducing kernel Hilbert space (RKHS) that aggregate information over a trajectory. Our approach leverages vector-valued occupation kernels for estimating the drift component of the stochastic process. For diffusion estimation, we extend this framework by introducing operator-valued occupation kernels, enabling the estimation of an auxiliary matrix-valued function as a positive semi-definite operator, from which we readily derive the diffusion estimate. This enables us to avoid common challenges in SDE learning, such as intractable likelihoods, by optimizing a reconstruction-error-based objective. We propose a simple learning procedure that retains strong predictive accuracy while using Fenchel duality to promote efficiency. We validate the method on simulated benchmarks and a real-world dataset of Amyloid imaging in healthy and Alzheimer's disease (AD) subjects.
Authors: Berkcan Kapusuzoglu, Supriyo Chakraborty, Chia-Hsuan Lee, Sambit Sahu
Abstract: Supervised fine-tuning (SFT) using expert demonstrations often suffer from the imitation problem, where the model learns to reproduce the correct responses without \emph{understanding} the underlying rationale. To address this limitation, we propose \textsc{Critique-Guided Distillation (CGD)}, a novel multi-stage framework that integrates teacher model generated \emph{explanatory critiques} and \emph{refined responses} into the SFT process. A student model is then trained to map the triplet of prompt, teacher critique, and its own initial response to the corresponding refined teacher response, thereby learning both \emph{what} to imitate and \emph{why}. Using entropy-based analysis, we show that \textsc{CGD} reduces refinement uncertainty and can be interpreted as a Bayesian posterior update. We perform extensive empirical evaluation of \textsc{CGD}, on variety of benchmark tasks, and demonstrate significant gains on both math (AMC23 +17.5%) and language understanding tasks (MMLU-Pro +6.3%), while successfully mitigating the format drift issues observed in previous critique fine-tuning (CFT) techniques.
Authors: Ivan Bioli, Carlo Marcati, Giancarlo Sangalli
Abstract: Natural Gradient Descent (NGD) has emerged as a promising optimization algorithm for training neural network-based solvers for partial differential equations (PDEs), such as Physics-Informed Neural Networks (PINNs). However, its practical use is often limited by the high computational cost of solving linear systems involving the Gramian matrix. While matrix-free NGD methods based on the conjugate gradient (CG) method avoid explicit matrix inversion, the ill-conditioning of the Gramian significantly slows the convergence of CG. In this work, we extend matrix-free NGD to broader classes of problems than previously considered and propose the use of Randomized Nystr\"om preconditioning to accelerate convergence of the inner CG solver. The resulting algorithm demonstrates substantial performance improvements over existing NGD-based methods on a range of PDE problems discretized using neural networks.
Authors: Falong Fan, Xi Li
Abstract: Multi-agent systems leverage advanced AI models as autonomous agents that interact, cooperate, or compete to complete complex tasks across applications such as robotics and traffic management. Despite their growing importance, safety in multi-agent systems remains largely underexplored, with most research focusing on single AI models rather than interacting agents. This work investigates backdoor vulnerabilities in multi-agent systems and proposes a defense mechanism based on agent interactions. By leveraging reasoning abilities, each agent evaluates responses from others to detect illogical reasoning processes, which indicate poisoned agents. Experiments on LLM-based multi-agent systems, including ChatGPT series and Llama 3, demonstrate the effectiveness of the proposed method, achieving high accuracy in identifying poisoned agents while minimizing false positives on clean agents. We believe this work provides insights into multi-agent system safety and contributes to the development of robust, trustworthy AI interactions.
Authors: Shubham Vatsal, Harsh Dubey, Aditi Singh
Abstract: Large language models (LLMs) have demonstrated impressive performance across a wide range of Natural Language Processing (NLP) tasks. However, ensuring their effectiveness across multiple languages presents unique challenges. Multilingual prompt engineering has emerged as a key approach to enhance LLMs' capabilities in diverse linguistic settings without requiring extensive parameter re-training or fine-tuning. With growing interest in multilingual prompt engineering over the past two to three years, researchers have explored various strategies to improve LLMs' performance across languages and NLP tasks. By crafting structured natural language prompts, researchers have successfully extracted knowledge from LLMs across different languages, making these techniques an accessible pathway for a broader audience, including those without deep expertise in machine learning, to harness the capabilities of LLMs. In this paper, we survey and categorize different multilingual prompting techniques based on the NLP tasks they address across a diverse set of datasets that collectively span around 250 languages. We further highlight the LLMs employed, present a taxonomy of approaches and discuss potential state-of-the-art (SoTA) methods for specific multilingual datasets. Additionally, we derive a range of insights across language families and resource levels (high-resource vs. low-resource), including analyses such as the distribution of NLP tasks by language resource type and the frequency of prompting methods across different language families. Our survey reviews 36 research papers covering 39 prompting techniques applied to 30 multilingual NLP tasks, with the majority of these studies published in the last two years.
Authors: Andrew Millard, Zheng Zhao, Joshua Murphy, Simon Maskell
Abstract: Sequential Monte Carlo (SMC) methods offer a principled approach to Bayesian uncertainty quantification but are traditionally limited by the need for full-batch gradient evaluations. We introduce a scalable variant by incorporating Stochastic Gradient Hamiltonian Monte Carlo (SGHMC) proposals into SMC, enabling efficient mini-batch based sampling. Our resulting SMCSGHMC algorithm outperforms standard stochastic gradient descent (SGD) and deep ensembles across image classification, out-of-distribution (OOD) detection, and transfer learning tasks. We further show that SMCSGHMC mitigates overfitting and improves calibration, providing a flexible, scalable pathway for converting pretrained neural networks into well-calibrated Bayesian models.
Authors: Hansen Chang, Christian DeLozier
Abstract: Programmers have long ignored warnings, especially those generated by static analysis tools, due to the potential for false-positives. In some cases, warnings may be indicative of larger issues, but programmers may not understand how a seemingly unimportant warning can grow into a vulnerability. Because these messages tend to be long and confusing, programmers tend to ignore them if they do not cause readily identifiable issues. Large language models can simplify these warnings, explain the gravity of important warnings, and suggest potential fixes to increase developer compliance with fixing warnings.
Authors: Zhibo Hu, Chen Wang, Yanfeng Shu, Hye-Young Paik, Liming Zhu
Abstract: Ambiguity in natural language is a significant obstacle for achieving accurate text to structured data mapping through large language models (LLMs), which affects the performance of tasks such as mapping text to agentic tool calling and text-to-SQL queries. Existing methods of ambiguity handling either exploit ReACT framework to produce the correct mapping through trial and error, or supervised fine tuning to guide models to produce a biased mapping to improve certain tasks. In this paper, we adopt a different approach that characterizes the representation difference of ambiguous text in the latent space and leverage the difference to identify ambiguity before mapping them to structured data. To detect ambiguity of a sentence, we focused on the relationship between ambiguous questions and their interpretations and what cause the LLM ignore multiple interpretations. Different to the distance calculated by dense embedding vectors, we utilize the observation that ambiguity is caused by concept missing in latent space of LLM to design a new distance measurement, computed through the path kernel by the integral of gradient values for each concepts from sparse-autoencoder (SAE) under each state. We identify patterns to distinguish ambiguous questions with this measurement. Based on our observation, We propose a new framework to improve the performance of LLMs on ambiguous agentic tool calling through missing concepts prediction.
Authors: Diksha Goel, Kristen Moore, Jeff Wang, Minjune Kim, Thanh Thi Nguyen
Abstract: Reinforcement Learning (RL) agents are increasingly used to simulate sophisticated cyberattacks, but their decision-making processes remain opaque, hindering trust, debugging, and defensive preparedness. In high-stakes cybersecurity contexts, explainability is essential for understanding how adversarial strategies are formed and evolve over time. In this paper, we propose a unified, multi-layer explainability framework for RL-based attacker agents that reveals both strategic (MDP-level) and tactical (policy-level) reasoning. At the MDP level, we model cyberattacks as a Partially Observable Markov Decision Processes (POMDPs) to expose exploration-exploitation dynamics and phase-aware behavioural shifts. At the policy level, we analyse the temporal evolution of Q-values and use Prioritised Experience Replay (PER) to surface critical learning transitions and evolving action preferences. Evaluated across CyberBattleSim environments of increasing complexity, our framework offers interpretable insights into agent behaviour at scale. Unlike previous explainable RL methods, which are often post-hoc, domain-specific, or limited in depth, our approach is both agent- and environment-agnostic, supporting use cases ranging from red-team simulation to RL policy debugging. By transforming black-box learning into actionable behavioural intelligence, our framework enables both defenders and developers to better anticipate, analyse, and respond to autonomous cyber threats.
Authors: Ryan Hoque, Peide Huang, David J. Yoon, Mouli Sivapurapu, Jian Zhang
Abstract: Imitation learning for manipulation has a well-known data scarcity problem. Unlike natural language and 2D computer vision, there is no Internet-scale corpus of data for dexterous manipulation. One appealing option is egocentric human video, a passively scalable data source. However, existing large-scale datasets such as Ego4D do not have native hand pose annotations and do not focus on object manipulation. To this end, we use Apple Vision Pro to collect EgoDex: the largest and most diverse dataset of dexterous human manipulation to date. EgoDex has 829 hours of egocentric video with paired 3D hand and finger tracking data collected at the time of recording, where multiple calibrated cameras and on-device SLAM can be used to precisely track the pose of every joint of each hand. The dataset covers a wide range of diverse manipulation behaviors with everyday household objects in 194 different tabletop tasks ranging from tying shoelaces to folding laundry. Furthermore, we train and systematically evaluate imitation learning policies for hand trajectory prediction on the dataset, introducing metrics and benchmarks for measuring progress in this increasingly important area. By releasing this large-scale dataset, we hope to push the frontier of robotics, computer vision, and foundation models.
Authors: Sumeet Batra, Gaurav Sukhatme
Abstract: Training vision-based manipulation policies that are robust across diverse visual environments remains an important and unresolved challenge in robot learning. Current approaches often sidestep the problem by relying on invariant representations such as point clouds and depth, or by brute-forcing generalization through visual domain randomization and/or large, visually diverse datasets. Disentangled representation learning - especially when combined with principles of associative memory - has recently shown promise in enabling vision-based reinforcement learning policies to be robust to visual distribution shifts. However, these techniques have largely been constrained to simpler benchmarks and toy environments. In this work, we scale disentangled representation learning and associative memory to more visually and dynamically complex manipulation tasks and demonstrate zero-shot adaptability to visual perturbations in both simulation and on real hardware. We further extend this approach to imitation learning, specifically Diffusion Policy, and empirically show significant gains in visual generalization compared to state-of-the-art imitation learning methods. Finally, we introduce a novel technique adapted from the model equivariance literature that transforms any trained neural network policy into one invariant to 2D planar rotations, making our policy not only visually robust but also resilient to certain camera perturbations. We believe that this work marks a significant step towards manipulation policies that are not only adaptable out of the box, but also robust to the complexities and dynamical nature of real-world deployment. Supplementary videos are available at https://sites.google.com/view/vis-gen-robotics/home.
Authors: Shijun Liang, Ismail R. Alkhouri, Siddhant Gautam, Qing Qu, Saiprasad Ravishankar
Abstract: Recent advances in data-centric deep generative models have led to significant progress in solving inverse imaging problems. However, these models (e.g., diffusion models (DMs)) typically require large amounts of fully sampled (clean) training data, which is often impractical in medical and scientific settings such as dynamic imaging. On the other hand, training-data-free approaches like the Deep Image Prior (DIP) do not require clean ground-truth images but suffer from noise overfitting and can be computationally expensive as the network parameters need to be optimized for each measurement set independently. Moreover, DIP-based methods often overlook the potential of learning a prior using a small number of sub-sampled measurements (or degraded images) available during training. In this paper, we propose UGoDIT, an Unsupervised Group DIP via Transferable weights, designed for the low-data regime where only a very small number, M, of sub-sampled measurement vectors are available during training. Our method learns a set of transferable weights by optimizing a shared encoder and M disentangled decoders. At test time, we reconstruct the unseen degraded image using a DIP network, where part of the parameters are fixed to the learned weights, while the remaining are optimized to enforce measurement consistency. We evaluate UGoDIT on both medical (multi-coil MRI) and natural (super resolution and non-linear deblurring) image recovery tasks under various settings. Compared to recent standalone DIP methods, UGoDIT provides accelerated convergence and notable improvement in reconstruction quality. Furthermore, our method achieves performance competitive with SOTA DM-based and supervised approaches, despite not requiring large amounts of clean training data.
Authors: Grace M. Lu (Department of Materials Science,Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, USA), Dallas R. Trinkle (Department of Materials Science,Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, USA)
Abstract: Explainable machine learning can help to discover new physical relationships for material properties. To understand the material properties that govern the activation energy for oxygen diffusion in perovskites and pyrochlores, we build a database of experimental activation energies and apply a grouping algorithm to the material property features. These features are then used to fit seven different machine learning models. An ensemble consensus determines that the most important features for predicting the activation energy are the ionicity of the A-site bond and the partial pressure of oxygen for perovskites. For pyrochlores, the two most important features are the A-site $s$ valence electron count and the B-site electronegativity. The most important features are all constructed using the weighted averages of elemental metal properties, despite weighted averages of the constituent binary oxides being included in our feature set. This is surprising because the material properties of the constituent oxides are more similar to the experimentally measured properties of perovskites and pyrochlores than the features of the metals that are chosen. The easy-to-measure features identified in this work enable rapid screening for new materials with fast oxide-ion diffusivity.
Authors: Pedro Figueiredo, Qihao He, Steve Bako, Nima Khademi Kalantari
Abstract: We propose a neural approach for estimating spatially varying light selection distributions to improve importance sampling in Monte Carlo rendering, particularly for complex scenes with many light sources. Our method uses a neural network to predict the light selection distribution at each shading point based on local information, trained by minimizing the KL-divergence between the learned and target distributions in an online manner. To efficiently manage hundreds or thousands of lights, we integrate our neural approach with light hierarchy techniques, where the network predicts cluster-level distributions and existing methods sample lights within clusters. Additionally, we introduce a residual learning strategy that leverages initial distributions from existing techniques, accelerating convergence during training. Our method achieves superior performance across diverse and challenging scenes.
Authors: Hao Mark Chen, Guanxi Lu, Yasuyuki Okoshi, Zhiwen Mo, Masato Motomura, Hongxiang Fan
Abstract: Test-time scaling (TTS) has proven effective in enhancing the reasoning capabilities of large language models (LLMs). Verification plays a key role in TTS, simultaneously influencing (1) reasoning performance and (2) compute efficiency, due to the quality and computational cost of verification. In this work, we challenge the conventional paradigms of verification, and make the first attempt toward systematically investigating the impact of verification granularity-that is, how frequently the verifier is invoked during generation, beyond verifying only the final output or individual generation steps. To this end, we introduce Variable Granularity Search (VG-Search), a unified algorithm that generalizes beam search and Best-of-N sampling via a tunable granularity parameter g. Extensive experiments with VG-Search under varying compute budgets, generator-verifier configurations, and task attributes reveal that dynamically selecting g can improve the compute efficiency and scaling behavior. Building on these findings, we propose adaptive VG-Search strategies that achieve accuracy gains of up to 3.1\% over Beam Search and 3.6\% over Best-of-N, while reducing FLOPs by over 52\%. We will open-source the code to support future research.
Authors: Jiahao Yu, Qizhen Ying, Leyang Wang, Ziyue Jiang, Song Liu
Abstract: This paper introduces a novel iterative method for missing data imputation that sequentially reduces the mutual information between data and their corresponding missing mask. Inspired by GAN-based approaches, which train generators to decrease the predictability of missingness patterns, our method explicitly targets the reduction of mutual information. Specifically, our algorithm iteratively minimizes the KL divergence between the joint distribution of the imputed data and missing mask, and the product of their marginals from the previous iteration. We show that the optimal imputation under this framework corresponds to solving an ODE, whose velocity field minimizes a rectified flow training objective. We further illustrate that some existing imputation techniques can be interpreted as approximate special cases of our mutual-information-reducing framework. Comprehensive experiments on synthetic and real-world datasets validate the efficacy of our proposed approach, demonstrating superior imputation performance.
Authors: Zhanxiang Hua, Ryan Sobash, David John Gagne II, Yingkai Sha, Alexandra Anderson-Frey
Abstract: Improving the skill of medium-range (1-8 day) severe weather prediction is crucial for mitigating societal impacts. This study introduces a novel approach leveraging decoder-only transformer networks to post-process AI-based weather forecasts, specifically from the Pangu-Weather model, for improved severe weather guidance. Unlike traditional post-processing methods that use a dense neural network to predict the probability of severe weather using discrete forecast samples, our method treats forecast lead times as sequential ``tokens'', enabling the transformer to learn complex temporal relationships within the evolving atmospheric state. We compare this approach against post-processing of the Global Forecast System (GFS) using both a traditional dense neural network and our transformer, as well as configurations that exclude convective parameters to fairly evaluate the impact of using the Pangu-Weather AI model. Results demonstrate that the transformer-based post-processing significantly enhances forecast skill compared to dense neural networks. Furthermore, AI-driven forecasts, particularly Pangu-Weather initialized from high resolution analysis, exhibit superior performance to GFS in the medium-range, even without explicit convective parameters. Our approach offers improved accuracy, and reliability, which also provides interpretability through feature attribution analysis, advancing medium-range severe weather prediction capabilities.
Authors: Shijun Li, Hilaf Hasson, Joydeep Ghosh
Abstract: Agents powered by advanced large language models (LLMs) have demonstrated impressive capabilities across diverse complex applications. Recently, Multi-Agent Systems (MAS), wherein multiple agents collaborate and communicate with each other, have exhibited enhanced capabilities in complex tasks, such as high-quality code generation and arithmetic reasoning. However, the development of such systems often relies on handcrafted methods, and the literature on systematic design and optimization of LLM-based MAS remains limited. In this work, we introduce OMAC, a general framework designed for holistic optimization of LLM-based MAS. Specifically, we identify five key optimization dimensions for MAS, encompassing both agent functionality and collaboration structure. Building upon these dimensions, we first propose a general algorithm, utilizing two actors termed the Semantic Initializer and the Contrastive Comparator, to optimize any single dimension. Then, we present an algorithm for joint optimization across multiple dimensions. Extensive experiments demonstrate the superior performance of OMAC on code generation, arithmetic reasoning, and general reasoning tasks against state-of-the-art approaches.
Authors: Seungeun Oh, Jinhyuk Kim, Jihong Park, Seung-Woo Ko, Jinho Choi, Tony Q. S. Quek, Seong-Lyun Kim
Abstract: To support emerging language-based applications using dispersed and heterogeneous computing resources, the hybrid language model (HLM) offers a promising architecture, where an on-device small language model (SLM) generates draft tokens that are validated and corrected by a remote large language model (LLM). However, the original HLM suffers from substantial communication overhead, as the LLM requires the SLM to upload the full vocabulary distribution for each token. Moreover, both communication and computation resources are wasted when the LLM validates tokens that are highly likely to be accepted. To overcome these limitations, we propose communication-efficient and uncertainty-aware HLM (CU-HLM). In CU-HLM, the SLM transmits truncated vocabulary distributions only when its output uncertainty is high. We validate the feasibility of this opportunistic transmission by discovering a strong correlation between SLM's uncertainty and LLM's rejection probability. Furthermore, we theoretically derive optimal uncertainty thresholds and optimal vocabulary truncation strategies. Simulation results show that, compared to standard HLM, CU-HLM achieves up to 206$\times$ higher token throughput by skipping 74.8% transmissions with 97.4% vocabulary compression, while maintaining 97.4% accuracy.
Authors: Quan Cheng, Yuanyu Wan, Lingyu Wu, Chenping Hou, Lijun Zhang
Abstract: Continual learning aims to learn multiple tasks sequentially while preserving prior knowledge, but faces the challenge of catastrophic forgetting when acquiring new knowledge. Recently, approaches leveraging pre-trained models have gained increasing popularity to mitigate this issue, due to the strong generalization ability of foundation models. To adjust pre-trained models for new tasks, existing methods usually employ low-rank adaptation, which restricts parameter updates to a fixed low-rank subspace. However, constraining the optimization space inherently compromises the model's learning capacity, resulting in inferior performance. To address the limitation, we propose Continuous Subspace Optimization for Continual Learning (CoSO) to fine-tune the model in a series of subspaces rather than a single one. These sequential subspaces are dynamically determined through the singular value decomposition of gradients. CoSO updates the model by projecting gradients into these subspaces, ensuring memory-efficient optimization. To mitigate forgetting, the optimization subspaces of each task are set to be orthogonal to the historical task subspace. During task learning, CoSO maintains a task-specific component that captures the critical update directions associated with the current task. Upon completing a task, this component is used to update the historical task subspace, laying the groundwork for subsequent learning. Extensive experiments on multiple datasets demonstrate that CoSO significantly outperforms state-of-the-art methods, especially in challenging scenarios with long task sequences.
Authors: Yang Xiao, Tianyi Peng, Rohan Kumar Das, Yuchen Hu, Huiping Zhuang
Abstract: Keyword spotting (KWS) offers a vital mechanism to identify spoken commands in voice-enabled systems, where user demands often shift, requiring models to learn new keywords continually over time. However, a major problem is catastrophic forgetting, where models lose their ability to recognize earlier keywords. Although several continual learning methods have proven their usefulness for reducing forgetting, most existing approaches depend on storing and revisiting old data to combat catastrophic forgetting. Though effective, these methods face two practical challenges: 1) privacy risks from keeping user data and 2) large memory and time consumption that limit deployment on small devices. To address these issues, we propose an exemplar-free Analytic Continual Learning (AnalyticKWS) method that updates model parameters without revisiting earlier data. Inspired by efficient learning principles, AnalyticKWS computes a closed-form analytical solution for model updates and requires only a single epoch of adaptation for incoming keywords. AnalyticKWS demands fewer computational resources by avoiding gradient-based updates and does not store old data. By eliminating the need for back-propagation during incremental learning, the model remains lightweight and efficient. As a result, AnalyticKWS meets the challenges mentioned earlier and suits resource-limited settings well. Extensive experiments on various datasets and settings show that AnalyticKWS consistently outperforms existing continual learning methods.
Authors: Junlang Huang, Hao Chen, Li Luo, Yong Cai, Lexin Zhang, Tianhao Ma, Yitian Zhang, Zhong Guan
Abstract: Simulation of high-order nonlinear system requires extensive computational resources, especially in modern VLSI backend design where bifurcation-induced instability and chaos-like transient behaviors pose challenges. We present S-Crescendo - a nested transformer weaving framework that synergizes S-domain with neural operators for scalable time-domain prediction in high-order nonlinear networks, alleviating the computational bottlenecks of conventional solvers via Newton-Raphson method. By leveraging the partial-fraction decomposition of an n-th order transfer function into first-order modal terms with repeated poles and residues, our method bypasses the conventional Jacobian matrix-based iterations and efficiently reduces computational complexity from cubic $O(n^3)$ to linear $O(n)$.The proposed architecture seamlessly integrates an S-domain encoder with an attention-based correction operator to simultaneously isolate dominant response and adaptively capture higher-order non-linearities. Validated on order-1 to order-10 networks, our method achieves up to 0.99 test-set ($R^2$) accuracy against HSPICE golden waveforms and accelerates simulation by up to 18(X), providing a scalable, physics-aware framework for high-dimensional nonlinear modeling.
Authors: Yiting Wang, Guoheng Sun, Wanghao Ye, Gang Qu, Ang Li
Abstract: Automating Register Transfer Level (RTL) code generation using Large Language Models (LLMs) offers substantial promise for streamlining digital circuit design and reducing human effort. However, current LLM-based approaches face significant challenges with training data scarcity, poor specification-code alignment, lack of verification mechanisms, and balancing generalization with specialization. Inspired by DeepSeek-R1, we introduce VeriReason, a framework integrating supervised fine-tuning with Guided Reward Proximal Optimization (GRPO) reinforcement learning for RTL generation. Using curated training examples and a feedback-driven reward model, VeriReason combines testbench evaluations with structural heuristics while embedding self-checking capabilities for autonomous error correction. On the VerilogEval Benchmark, VeriReason delivers significant improvements: achieving 83.1% functional correctness on the VerilogEval Machine benchmark, substantially outperforming both comparable-sized models and much larger commercial systems like GPT-4 Turbo. Additionally, our approach demonstrates up to a 2.8X increase in first-attempt functional correctness compared to baseline methods and exhibits robust generalization to unseen designs. To our knowledge, VeriReason represents the first system to successfully integrate explicit reasoning capabilities with reinforcement learning for Verilog generation, establishing a new state-of-the-art for automated RTL synthesis. The models and datasets are available at: https://huggingface.co/collections/AI4EDA-CASE Code is Available at: https://github.com/NellyW8/VeriReason
URLs: https://huggingface.co/collections/AI4EDA-CASE, https://github.com/NellyW8/VeriReason
Authors: Chicago Y. Park, Shirin Shoushtari, Hongyu An, Ulugbek S. Kamilov
Abstract: Diffusion models are widely used in applications ranging from image generation to inverse problems. However, training diffusion models typically requires clean ground-truth images, which are unavailable in many applications. We introduce the Measurement Score-based diffusion Model (MSM), a novel framework that learns partial measurement scores using only noisy and subsampled measurements. MSM models the distribution of full measurements as an expectation over partial scores induced by randomized subsampling. To make the MSM representation computationally efficient, we also develop a stochastic sampling algorithm that generates full images by using a randomly selected subset of partial scores at each step. We additionally propose a new posterior sampling method for solving inverse problems that reconstructs images using these partial scores. We provide a theoretical analysis that bounds the Kullback-Leibler divergence between the distributions induced by full and stochastic sampling, establishing the accuracy of the proposed algorithm. We demonstrate the effectiveness of MSM on natural images and multi-coil MRI, showing that it can generate high-quality images and solve inverse problems -- all without access to clean training data. Code is available at https://github.com/wustl-cig/MSM.
Authors: Shiming Chen, Dingjie Fu, Salman Khan, Fahad Shahbaz Khan
Abstract: Remarkable progress in zero-shot learning (ZSL) has been achieved using generative models. However, existing generative ZSL methods merely generate (imagine) the visual features from scratch guided by the strong class semantic vectors annotated by experts, resulting in suboptimal generative performance and limited scene generalization. To address these and advance ZSL, we propose an inductive variational autoencoder for generative zero-shot learning, dubbed GenZSL. Mimicking human-level concept learning, GenZSL operates by inducting new class samples from similar seen classes using weak class semantic vectors derived from target class names (i.e., CLIP text embedding). To ensure the generation of informative samples for training an effective ZSL classifier, our GenZSL incorporates two key strategies. Firstly, it employs class diversity promotion to enhance the diversity of class semantic vectors. Secondly, it utilizes target class-guided information boosting criteria to optimize the model. Extensive experiments conducted on three popular benchmark datasets showcase the superiority and potential of our GenZSL with significant efficacy and efficiency over f-VAEGAN, e.g., 24.7% performance gains and more than $60\times$ faster training speed on AWA2. Codes are available at https://github.com/shiming-chen/GenZSL.
Authors: Peter Vere\v{s}, Richard Cloete, Matthew J. Payne, Abraham Loeb
Abstract: We present a comprehensive analysis of the digest2 parameters for candidates of the Near-Earth Object Confirmation Page (NEOCP) that were reported between 2019 and 2024. Our study proposes methods for significantly reducing the inclusion of non-NEO objects on the NEOCP. Despite the substantial increase in near-Earth object (NEO) discoveries in recent years, only about half of the NEOCP candidates are ultimately confirmed as NEOs. Therefore, much observing time is spent following up on non-NEOs. Furthermore, approximately 11% of the candidates remain unconfirmed because the follow-up observations are insufficient. These are nearly 600 cases per year. To reduce false positives and minimize wasted resources on non-NEOs, we refine the posting criteria for NEOCP based on a detailed analysis of all digest2 scores. We investigated 30 distinct digest2 parameter categories for candidates that were confirmed as NEOs and non-NEOs. From this analysis, we derived a filtering mechanism based on selected digest2 parameters that were able to exclude 20% of the non-NEOs from the NEOCP while maintaining a minimal loss of true NEOs. We also investigated the application of four machine-learning (ML) techniques, that is, the gradient-boosting machine (GBM), the random forest (RF) classifier, the stochastic gradient descent (SGD) classifier, and neural networks (NN) to classify NEOCP candidates as NEOs or non-NEOs. Based on digest2 parameters as input, our ML models achieved a precision of approximately 95% in distinguishing between NEOs and non-NEOs. Results. Combining the digest2 parameter filter with an ML-based classification model, we demonstrate a significant reduction in non-NEOs on the NEOCP that exceeds 80%, while limiting the loss of NEO discovery tracklets to 5.5%. Importantly, we show that most follow-up tracklets of initially misclassified NEOs are later correctly identified as NEOs.
Authors: Yu-Ting Lee, Hui-Ying Shih, Fu-Chieh Chang, Pei-Yuan Wu
Abstract: We provide an explanation for the performance gains of intrinsic self-correction, a process where a language model iteratively refines its outputs without external feedback. More precisely, we investigate how prompting induces interpretable changes in hidden states and thus affects the output distributions. We hypothesize that each prompt-induced shift lies in a linear span of some linear representation vectors, naturally separating tokens based on individual concept alignment. Building around this idea, we give a mathematical formulation of self-correction and derive a concentration result for output tokens based on alignment magnitudes. Our experiments on text detoxification with zephyr-7b-sft reveal a substantial gap in the inner products of the prompt-induced shifts and the unembeddings of the top-100 most toxic tokens vs. those of the unembeddings of the bottom-100 least toxic tokens, under toxic instructions. This suggests that self-correction prompts enhance a language model's capability of latent concept recognition. Our analysis offers insights into the underlying mechanism of self-correction by characterizing how prompting works explainably. For reproducibility, our code is available.
Authors: Haitao Li, Che Liu, Zhengyao Ding, Ziyi Liu, Zhengxing Huang
Abstract: Electrocardiograms (ECGs) are essential for diagnosing cardiovascular diseases. While previous ECG-text contrastive learning methods have shown promising results, they often overlook the incompleteness of the reports. Given an ECG, the report is generated by first identifying key waveform features and then inferring the final diagnosis through these features. Despite their importance, these waveform features are often not recorded in the report as intermediate results. Aligning ECGs with such incomplete reports impedes the model's ability to capture the ECG's waveform features and limits its understanding of diagnostic reasoning based on those features. To address this, we propose FG-CLEP (Fine-Grained Contrastive Language ECG Pre-training), which aims to recover these waveform features from incomplete reports with the help of large language models (LLMs), under the challenges of hallucinations and the non-bijective relationship between waveform features and diagnoses. Additionally, considering the frequent false negatives due to the prevalence of common diagnoses in ECGs, we introduce a semantic similarity matrix to guide contrastive learning. Furthermore, we adopt a sigmoid-based loss function to accommodate the multi-label nature of ECG-related tasks. Experiments on six datasets demonstrate that FG-CLEP outperforms state-of-the-art methods in both zero-shot prediction and linear probing across these datasets.
Authors: Adam Kovari, Yasin Ghafourian, Csaba Hegedus, Belal Abu Naim, Kitti Mezei, Pal Varga, Markus Tauber
Abstract: As artificial intelligence (AI) regulations evolve and the regulatory landscape develops and becomes more complex, ensuring compliance with ethical guidelines and legal frameworks remains a challenge for AI developers. This paper introduces an AI-driven self-assessment chatbot designed to assist users in navigating the European Union AI Act and related standards. Leveraging a Retrieval-Augmented Generation (RAG) framework, the chatbot enables real-time, context-aware compliance verification by retrieving relevant regulatory texts and providing tailored guidance. By integrating both public and proprietary standards, it streamlines regulatory adherence, reduces complexity, and fosters responsible AI development. The paper explores the chatbot's architecture, comparing naive and graph-based RAG models, and discusses its potential impact on AI governance.
Authors: Jitendra K Tugnait
Abstract: We consider the problem of inferring the conditional independence graph (CIG) of high-dimensional Gaussian vectors from multi-attribute data. Most existing methods for graph estimation are based on single-attribute models where one associates a scalar random variable with each node. In multi-attribute graphical models, each node represents a random vector. In this paper we provide a unified theoretical analysis of multi-attribute graph learning using a penalized log-likelihood objective function. We consider both convex (sparse-group lasso) and sparse-group non-convex (log-sum and smoothly clipped absolute deviation (SCAD) penalties) penalty/regularization functions. An alternating direction method of multipliers (ADMM) approach coupled with local linear approximation to non-convex penalties is presented for optimization of the objective function. For non-convex penalties, theoretical analysis establishing local consistency in support recovery, local convexity and precision matrix estimation in high-dimensional settings is provided under two sets of sufficient conditions: with and without some irrepresentability conditions. We illustrate our approaches using both synthetic and real-data numerical examples. In the synthetic data examples the sparse-group log-sum penalized objective function significantly outperformed the lasso penalized as well as SCAD penalized objective functions with $F_1$-score and Hamming distance as performance metrics.
Authors: Drashthi Doshi, Aditya Vema Reddy Kesari, Swaprava Nath, Avishek Ghosh, Suhas S Kowshik
Abstract: Classical federated learning (FL) assumes that the clients have a limited amount of noisy data with which they voluntarily participate and contribute towards learning a global, more accurate model in a principled manner. The learning happens in a distributed fashion without sharing the data with the center. However, these methods do not consider the incentive of an agent for participating and contributing to the process, given that data collection and running a distributed algorithm is costly for the clients. The question of rationality of contribution has been asked recently in the literature and some results exist that consider this problem. This paper addresses the question of simultaneous parameter learning and incentivizing contribution, which distinguishes it from the extant literature. Our first mechanism incentivizes each client to contribute to the FL process at a Nash equilibrium and simultaneously learn the model parameters. However, this equilibrium outcome can be away from the optimal, where clients contribute with their full data and the algorithm learns the optimal parameters. We propose a second mechanism with monetary transfers that is budget balanced and enables the full data contribution along with optimal parameter learning. Large scale experiments with real (federated) datasets (CIFAR-10, FeMNIST, and Twitter) show that these algorithms converge quite fast in practice, yield good welfare guarantees, and better model performance for all agents.
Authors: Jianyi Zhang, Ziyin Zhou, Yilong Li, Qichao Jin
Abstract: Federated learning (FL) is gaining increasing attention as an emerging collaborative machine learning approach, particularly in the context of large-scale computing and data systems. However, the fundamental algorithm of FL, Federated Averaging (FedAvg), is susceptible to backdoor attacks. Although researchers have proposed numerous defense algorithms, two significant challenges remain. The attack is becoming more stealthy and harder to detect, and current defense methods are unable to handle 50\% or more malicious users or assume an auxiliary server dataset. To address these challenges, we propose a novel defense algorithm, FL-PLAS, \textbf{F}ederated \textbf{L}earning based on \textbf{P}artial\textbf{ L}ayer \textbf{A}ggregation \textbf{S}trategy. In particular, we divide the local model into a feature extractor and a classifier. In each iteration, the clients only upload the parameters of a feature extractor after local training. The server then aggregates these local parameters and returns the results to the clients. Each client retains its own classifier layer, ensuring that the backdoor labels do not impact other clients. We assess the effectiveness of FL-PLAS against state-of-the-art (SOTA) backdoor attacks on three image datasets and compare our approach to six defense strategies. The results of the experiment demonstrate that our methods can effectively protect local models from backdoor attacks. Without requiring any auxiliary dataset for the server, our method achieves a high main-task accuracy with a lower backdoor accuracy even under the condition of 90\% malicious users with the attacks of trigger, semantic and edge-case.
Authors: Vinod Raman, Hilal Asi, Satyen Kale
Abstract: Recent advances in test-time alignment methods, such as Best-of-N sampling, offer a simple and effective way to steer language models (LMs) toward preferred behaviors using reward models (RM). However, these approaches can be computationally expensive, especially when applied uniformly across prompts without accounting for differences in alignment difficulty. In this work, we propose a prompt-adaptive strategy for Best-of-N alignment that allocates inference-time compute more efficiently. Motivated by latency concerns, we develop a two-stage algorithm: an initial exploratory phase estimates the reward distribution for each prompt using a small exploration budget, and a second stage adaptively allocates the remaining budget using these estimates. Our method is simple, practical, and compatible with any LM/RM combination. Empirical results on the AlpacaEval dataset for 12 LM/RM pairs and 50 different batches of prompts show that our adaptive strategy consistently outperforms the uniform allocation with the same inference budget. Moreover, our experiments show that our adaptive strategy remains competitive against uniform allocations with 20% larger inference budgets and even improves in performance as the batch size grows.
Authors: Tiannuo Yang, Zebin Yao, Bowen Jin, Lixiao Cui, Yusen Li, Gang Wang, Xiaoguang Liu
Abstract: Large Language Model (LLM)-based search agents have shown remarkable capabilities in solving complex tasks by dynamically decomposing problems and addressing them through interleaved reasoning and retrieval. However, this interleaved paradigm introduces substantial efficiency bottlenecks. First, we observe that both highly accurate and overly approximate retrieval methods degrade system efficiency: exact search incurs significant retrieval overhead, while coarse retrieval requires additional reasoning steps during generation. Second, we identify inefficiencies in system design, including improper scheduling and frequent retrieval stalls, which lead to cascading latency -- where even minor delays in retrieval amplify end-to-end inference time. To address these challenges, we introduce SearchAgent-X, a high-efficiency inference framework for LLM-based search agents. SearchAgent-X leverages high-recall approximate retrieval and incorporates two key techniques: priority-aware scheduling and non-stall retrieval. Extensive experiments demonstrate that SearchAgent-X consistently outperforms state-of-the-art systems such as vLLM and HNSW-based retrieval across diverse tasks, achieving up to 3.4$\times$ higher throughput and 5$\times$ lower latency, without compromising generation quality. SearchAgent-X is available at https://github.com/tiannuo-yang/SearchAgent-X.
Authors: Guy Davidson, Todd M. Gureckis, Brenden M. Lake, Adina Williams
Abstract: Demonstrations and instructions are two primary approaches for prompting language models to perform in-context learning (ICL) tasks. Do identical tasks elicited in different ways result in similar representations of the task? An improved understanding of task representation mechanisms would offer interpretability insights and may aid in steering models. We study this through function vectors, recently proposed as a mechanism to extract few-shot ICL task representations. We generalize function vectors to alternative task presentations, focusing on short textual instruction prompts, and successfully extract instruction function vectors that promote zero-shot task accuracy. We find evidence that demonstration- and instruction-based function vectors leverage different model components, and offer several controls to dissociate their contributions to task performance. Our results suggest that different task presentations do not induce a common task representation but elicit different, partly overlapping mechanisms. Our findings offer principled support to the practice of combining textual instructions and task demonstrations, imply challenges in universally monitoring task inference across presentation forms, and encourage further examinations of LLM task inference mechanisms.
Authors: Yunshui Li, Yiyuan Ma, Shen Yan, Chaoyi Zhang, Jing Liu, Jianqiao Lu, Ziwen Xu, Mengzhao Chen, Minrui Wang, Shiyi Zhan, Jin Ma, Xunhao Lai, Yao Luo, Xingyan Bin, Hongbin Ren, Mingji Han, Wenhao Hao, Bairen Yi, LingJun Liu, Bole Ma, Xiaoying Jia, Zhou Xun, Liang Xiang, Yonghui Wu
Abstract: Model merging has emerged as a promising technique for enhancing large language models, though its application in large-scale pre-training remains relatively unexplored. In this paper, we present a comprehensive investigation of model merging techniques during the pre-training process. Through extensive experiments with both dense and Mixture-of-Experts (MoE) architectures ranging from millions to over 100 billion parameters, we demonstrate that merging checkpoints trained with constant learning rates not only achieves significant performance improvements but also enables accurate prediction of annealing behavior. These improvements lead to both more efficient model development and significantly lower training costs. Our detailed ablation studies on merging strategies and hyperparameters provide new insights into the underlying mechanisms while uncovering novel applications. Through comprehensive experimental analysis, we offer the open-source community practical pre-training guidelines for effective model merging.
Authors: Marco Fiandri, Alberto Maria Metelli, Francesco Trov\`o
Abstract: Stochastic rising rested bandit (SRRB) is a setting where the arms' expected rewards increase as they are pulled. It models scenarios in which the performances of the different options grow as an effect of an underlying learning process (e.g., online model selection). Even if the bandit literature provides specifically crafted algorithms based on upper-confidence bounds for such a setting, no study about Thompson sampling TS-like algorithms has been performed so far. The strong regularity of the expected rewards in the SRRB setting suggests that specific instances may be tackled effectively using adapted and sliding-window TS approaches. This work provides novel regret analyses for such algorithms in SRRBs, highlighting the challenges and providing new technical tools of independent interest. Our results allow us to identify under which assumptions TS-like algorithms succeed in achieving sublinear regret and which properties of the environment govern the complexity of the regret minimization problem when approached with TS. Furthermore, we provide a regret lower bound based on a complexity index we introduce. Finally, we conduct numerical simulations comparing TS-like algorithms with state-of-the-art approaches for SRRBs in synthetic and real-world settings.
Authors: Daniel Cederberg
Abstract: Over the past decades, there has been a surge of interest in studying low-dimensional structures within high-dimensional data. Statistical factor models $-$ i.e., low-rank plus diagonal covariance structures $-$ offer a powerful framework for modeling such structures. However, traditional methods for fitting statistical factor models, such as principal component analysis (PCA) or maximum likelihood estimation assuming the data is Gaussian, are highly sensitive to heavy tails and outliers in the observed data. In this paper, we propose a novel expectation-maximization (EM) algorithm for robustly fitting statistical factor models. Our approach is based on Tyler's M-estimator of the scatter matrix for an elliptical distribution, and consists of solving Tyler's maximum likelihood estimation problem while imposing a structural constraint that enforces the low-rank plus diagonal covariance structure. We present numerical experiments on both synthetic and real examples, demonstrating the robustness of our method for direction-of-arrival estimation in nonuniform noise and subspace recovery.
Authors: Nikita P. Kalinin, Ryan McKenna, Jalaj Upadhyay, Christoph H. Lampert
Abstract: Matrix factorization mechanisms for differentially private training have emerged as a promising approach to improve model utility under privacy constraints. In practical settings, models are typically trained over multiple epochs, requiring matrix factorizations that account for repeated participation. Existing theoretical upper and lower bounds on multi-epoch factorization error leave a significant gap. In this work, we introduce a new explicit factorization method, Banded Inverse Square Root (BISR), which imposes a banded structure on the inverse correlation matrix. This factorization enables us to derive an explicit and tight characterization of the multi-epoch error. We further prove that BISR achieves asymptotically optimal error by matching the upper and lower bounds. Empirically, BISR performs on par with state-of-the-art factorization methods, while being simpler to implement, computationally efficient, and easier to analyze.
Authors: Rodrigo Moreira, Tereza C. M. Carvalho, Fl\'avio de Oliveira Silva, Nazim Agoulmine, Joberto S. B. Martins
Abstract: The 6G mobile network is the next evolutionary step after 5G, with a prediction of an explosive surge in mobile traffic. It provides ultra-low latency, higher data rates, high device density, and ubiquitous coverage, positively impacting services in various areas. Energy saving is a major concern for new systems in the telecommunications sector because all players are expected to reduce their carbon footprints to contribute to mitigating climate change. Network slicing is a fundamental enabler for 6G/5G mobile networks and various other new systems, such as the Internet of Things (IoT), Internet of Vehicles (IoV), and Industrial IoT (IIoT). However, energy-saving methods embedded in network slicing architectures are still a research gap. This paper discusses how to embed energy-saving methods in network-slicing architectures that are a fundamental enabler for nearly all new innovative systems being deployed worldwide. This paper's main contribution is a proposal to save energy in network slicing. That is achieved by deploying ML-native agents in NS architectures to dynamically orchestrate and optimize resources based on user demands. The SFI2 network slicing reference architecture is the concrete use case scenario in which contrastive learning improves energy saving for resource allocation.
Authors: Ranit Karmakar, Simon F. N{\o}rrelykke
Abstract: Segmentation evaluation metrics traditionally rely on binary decision logic: predictions are either correct or incorrect, based on rigid IoU thresholds. Detection--based metrics such as F1 and mAP determine correctness at the object level using fixed overlap cutoffs, while overlap--based metrics like Intersection over Union (IoU) and Dice operate at the pixel level, often overlooking instance--level structure. Panoptic Quality (PQ) attempts to unify detection and segmentation assessment, but it remains dependent on hard-threshold matching--treating predictions below the threshold as entirely incorrect. This binary framing obscures important distinctions between qualitatively different errors and fails to reward gradual model improvements. We propose SoftPQ, a flexible and interpretable instance segmentation metric that redefines evaluation as a graded continuum rather than a binary classification. SoftPQ introduces tunable upper and lower IoU thresholds to define a partial matching region and applies a sublinear penalty function to ambiguous or fragmented predictions. These extensions allow SoftPQ to exhibit smoother score behavior, greater robustness to structural segmentation errors, and more informative feedback for model development and evaluation. Through controlled perturbation experiments, we show that SoftPQ captures meaningful differences in segmentation quality that existing metrics overlook, making it a practical and principled alternative for both benchmarking and iterative model refinement.
Authors: Hossein Babaei, Mel White, Sina Alemohammad, Richard G. Baraniuk
Abstract: State-Space Models (SSMs) have proven to be powerful tools for modeling long-range dependencies in sequential data. While the recent method known as HiPPO has demonstrated strong performance, and formed the basis for machine learning models S4 and Mamba, it remains limited by its reliance on closed-form solutions for a few specific, well-behaved bases. The SaFARi framework generalized this approach, enabling the construction of SSMs from arbitrary frames, including non-orthogonal and redundant ones, thus allowing an infinite diversity of possible "species" within the SSM family. In this paper, we introduce WaLRUS (Wavelets for Long-range Representation Using SSMs), a new implementation of SaFARi built from Daubechies wavelets.
Authors: Sen Fang, Weiyuan Ding, Bowen Xu
Abstract: Assessing the programming capabilities of Large Language Models (LLMs) is crucial for their effective use in software engineering. Current evaluations, however, predominantly measure the accuracy of generated code on static benchmarks, neglecting the critical aspect of model robustness during programming tasks. While adversarial attacks offer insights on model robustness, their effectiveness is limited and evaluation could be constrained. Current adversarial attack methods for robustness evaluation yield inconsistent results, struggling to provide a unified evaluation across different LLMs. We introduce EVALOOP, a novel assessment framework that evaluate the robustness from a self-consistency perspective, i.e., leveraging the natural duality inherent in popular software engineering tasks, e.g., code generation and code summarization. EVALOOP initiates a self-contained feedback loop: an LLM generates output (e.g., code) from an input (e.g., natural language specification), and then use the generated output as the input to produce a new output (e.g., summarizes that code into a new specification). EVALOOP repeats the process to assess the effectiveness of EVALOOP in each loop. This cyclical strategy intrinsically evaluates robustness without rely on any external attack setups, providing a unified metric to evaluate LLMs' robustness in programming. We evaluate 16 prominent LLMs (e.g., GPT-4.1, O4-mini) on EVALOOP and found that EVALOOP typically induces a 5.01%-19.31% absolute drop in pass@1 performance within ten loops. Intriguingly, robustness does not always align with initial performance (i.e., one-time query); for instance, GPT-3.5-Turbo, despite superior initial code generation compared to DeepSeek-V2, demonstrated lower robustness over repeated evaluation loop.
Authors: Wenquan Lu, Jiaqi Zhang, Hugues Van Assel, Randall Balestriero
Abstract: Self-Supervised Learning (SSL) has become a powerful solution to extract rich representations from unlabeled data. Yet, SSL research is mostly focused on clean, curated and high-quality datasets. As a result, applying SSL on noisy data remains a challenge, despite being crucial to applications such as astrophysics, medical imaging, geophysics or finance. In this work, we present a fully self-supervised framework that enables noise-robust representation learning without requiring a denoiser at inference or downstream fine-tuning. Our method first trains an SSL denoiser on noisy data, then uses it to construct a denoised-to-noisy data curriculum (i.e., training first on denoised, then noisy samples) for pretraining a SSL backbone (e.g., DINOv2), combined with a teacher-guided regularization that anchors noisy embeddings to their denoised counterparts. This process encourages the model to internalize noise robustness. Notably, the denoiser can be discarded after pretraining, simplifying deployment. On ImageNet-1k with ViT-B under extreme Gaussian noise ($\sigma=255$, SNR = 0.72 dB), our method improves linear probing accuracy by 4.8% over DINOv2, demonstrating that denoiser-free robustness can emerge from noise-aware pretraining. The code is available at https://github.com/wenquanlu/noisy_dinov2.
Authors: Mathanesh Vellingiri Ramasamy, Dimas Rizky Kurniasalim
Abstract: Accurate road segmentation is essential for autonomous driving and ADAS, enabling effective navigation in complex environments. This study examines how model architecture and dataset choice affect segmentation by training a modified VGG-16 on the Comma10k dataset and a modified U-Net on the KITTI Road dataset. Both models achieved high accuracy, with cross-dataset testing showing VGG-16 outperforming U-Net despite U-Net being trained for more epochs. We analyze model performance using metrics such as F1-score, mean intersection over union, and precision, discussing how architecture and dataset impact results.
Authors: Karthik Gopinath, Annabel Sorby-Adams, Jonathan W. Ramirez, Dina Zemlyanker, Jennifer Guo, David Hunt, Christine L. Mac Donald, C. Dirk Keene, Timothy Coalson, Matthew F. Glasser, David Van Essen, Matthew S. Rosen, Oula Puonti, W. Taylor Kimberly, Juan Eugenio Iglesias
Abstract: Three-dimensional reconstruction of cortical surfaces from MRI for morphometric analysis is fundamental for understanding brain structure. While high-field MRI (HF-MRI) is standard in research and clinical settings, its limited availability hinders widespread use. Low-field MRI (LF-MRI), particularly portable systems, offers a cost-effective and accessible alternative. However, existing cortical surface analysis tools are optimized for high-resolution HF-MRI and struggle with the lower signal-to-noise ratio and resolution of LF-MRI. In this work, we present a machine learning method for 3D reconstruction and analysis of portable LF-MRI across a range of contrasts and resolutions. Our method works "out of the box" without retraining. It uses a 3D U-Net trained on synthetic LF-MRI to predict signed distance functions of cortical surfaces, followed by geometric processing to ensure topological accuracy. We evaluate our method using paired HF/LF-MRI scans of the same subjects, showing that LF-MRI surface reconstruction accuracy depends on acquisition parameters, including contrast type (T1 vs T2), orientation (axial vs isotropic), and resolution. A 3mm isotropic T2-weighted scan acquired in under 4 minutes, yields strong agreement with HF-derived surfaces: surface area correlates at r=0.96, cortical parcellations reach Dice=0.98, and gray matter volume achieves r=0.93. Cortical thickness remains more challenging with correlations up to r=0.70, reflecting the difficulty of sub-mm precision with 3mm voxels. We further validate our method on challenging postmortem LF-MRI, demonstrating its robustness. Our method represents a step toward enabling cortical surface analysis on portable LF-MRI. Code is available at https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAny
Authors: Quanjiang Guo, Jinchuan Zhang, Sijie Wang, Ling Tian, Zhao Kang, Bin Yan, Weidong Xiao
Abstract: Few-Shot Relation Extraction (FSRE) remains a challenging task due to the scarcity of annotated data and the limited generalization capabilities of existing models. Although large language models (LLMs) have demonstrated potential in FSRE through in-context learning (ICL), their general-purpose training objectives often result in suboptimal performance for task-specific relation extraction. To overcome these challenges, we propose TKRE (Two-Stage Knowledge-Guided Pre-training for Relation Extraction), a novel framework that synergistically integrates LLMs with traditional relation extraction models, bridging generative and discriminative learning paradigms. TKRE introduces two key innovations: (1) leveraging LLMs to generate explanation-driven knowledge and schema-constrained synthetic data, addressing the issue of data scarcity; and (2) a two-stage pre-training strategy combining Masked Span Language Modeling (MSLM) and Span-Level Contrastive Learning (SCL) to enhance relational reasoning and generalization. Together, these components enable TKRE to effectively tackle FSRE tasks. Comprehensive experiments on benchmark datasets demonstrate the efficacy of TKRE, achieving new state-of-the-art performance in FSRE and underscoring its potential for broader application in low-resource scenarios. \footnote{The code and data are released on https://github.com/UESTC-GQJ/TKRE.
Authors: Tingfeng Lan, Yusen Wu, Bin Ma, Zhaoyuan Su, Rui Yang, Tekin Bicer, Dong Li, Yue Cheng
Abstract: Fine-tuning large language models (LLMs) often exceeds GPU memory limits, prompting systems to offload model states to CPU memory. However, existing offloaded training frameworks like ZeRO-Offload treat all parameters equally and update the full model on the CPU, causing severe GPU stalls, where fast, expensive GPUs sit idle waiting for slow CPU updates and limited-bandwidth PCIe transfers. We present ZenFlow, a new offloading framework that prioritizes important parameters and decouples updates between GPU and CPU. ZenFlow performs in-place updates of important gradients on GPU, while asynchronously offloading and accumulating less important ones on CPU, fully overlapping CPU work with GPU computation. To scale across GPUs, ZenFlow introduces a lightweight gradient selection method that exploits a novel spatial and temporal locality property of important gradients, avoiding costly global synchronization. ZenFlow achieves up to 5x end-to-end speedup, 2x lower PCIe traffic, and reduces GPU stalls by over 85 percent, all while preserving accuracy.
Authors: Yiwei Ou, Xiaobin Ren, Ronggui Sun, Guansong Gao, Ziyi Jiang, Kaiqi Zhao, Manfredo Manfredini
Abstract: Existing visual place recognition (VPR) datasets predominantly rely on vehicle-mounted imagery, lack multimodal diversity and underrepresent dense, mixed-use street-level spaces, especially in non-Western urban contexts. To address these gaps, we introduce MMS-VPR, a large-scale multimodal dataset for street-level place recognition in complex, pedestrian-only environments. The dataset comprises 78,575 annotated images and 2,512 video clips captured across 207 locations in a ~70,800 $\mathrm{m}^2$ open-air commercial district in Chengdu, China. Each image is labeled with precise GPS coordinates, timestamp, and textual metadata, and covers varied lighting conditions, viewpoints, and timeframes. MMS-VPR follows a systematic and replicable data collection protocol with minimal device requirements, lowering the barrier for scalable dataset creation. Importantly, the dataset forms an inherent spatial graph with 125 edges, 81 nodes, and 1 subgraph, enabling structure-aware place recognition. We further define two application-specific subsets -- Dataset_Edges and Dataset_Points -- to support fine-grained and graph-based evaluation tasks. Extensive benchmarks using conventional VPR models, graph neural networks, and multimodal baselines show substantial improvements when leveraging multimodal and structural cues. MMS-VPR facilitates future research at the intersection of computer vision, geospatial understanding, and multimodal reasoning. The dataset is publicly available at https://huggingface.co/datasets/Yiwei-Ou/MMS-VPR.
Authors: Pratim Chowdhary
Abstract: Understanding which neural components drive specific capabilities in mid-sized language models ($\leq$10B parameters) remains a key challenge. We introduce the $(\bm{K}, \epsilon)$-Minimum Sufficient Head Circuit ($K$-MSHC), a methodology to identify minimal sets of attention heads crucial for classification tasks as well as Search-K-MSHC, an efficient algorithm for discovering these circuits. Applying our Search-K-MSHC algorithm to Gemma-9B, we analyze three syntactic task families: grammar acceptability, arithmetic verification, and arithmetic word problems. Our findings reveal distinct task-specific head circuits, with grammar tasks predominantly utilizing early layers, word problems showing pronounced activity in both shallow and deep regions, and arithmetic verification demonstrating a more distributed pattern across the network. We discover non-linear circuit overlap patterns, where different task pairs share computational components at varying levels of importance. While grammar and arithmetic share many "weak" heads, arithmetic and word problems share more consistently critical "strong" heads. Importantly, we find that each task maintains dedicated "super-heads" with minimal cross-task overlap, suggesting that syntactic and numerical competencies emerge from specialized yet partially reusable head circuits.
Authors: Lingzhi Wang, Pengcheng Huang, Haotian Li, Yuliang Wei, Guodong Xin, Rui Zhang, Donglin Zhang, Zhenzhou Ji, Wei Wang
Abstract: Knowledge graphs (KGs) serve as fundamental structures for organizing interconnected data across diverse domains. However, most KGs remain incomplete, limiting their effectiveness in downstream applications. Knowledge graph completion (KGC) aims to address this issue by inferring missing links, but existing methods face critical challenges: deep graph neural networks (GNNs) suffer from over-smoothing, while embedding-based models fail to capture abstract relational features. This study aims to overcome these limitations by proposing a unified framework that integrates GNN distillation and abstract probabilistic interaction modeling (APIM). GNN distillation approach introduces an iterative message-feature filtering process to mitigate over-smoothing, preserving the discriminative power of node representations. APIM module complements this by learning structured, abstract interaction patterns through probabilistic signatures and transition matrices, allowing for a richer, more flexible representation of entity and relation interactions. We apply these methods to GNN-based models and the APIM to embedding-based KGC models, conducting extensive evaluations on the widely used WN18RR and FB15K-237 datasets. Our results demonstrate significant performance gains over baseline models, showcasing the effectiveness of the proposed techniques. The findings highlight the importance of both controlling information propagation and leveraging structured probabilistic modeling, offering new avenues for advancing knowledge graph completion. And our codes are available at https://anonymous.4open.science/r/APIM_and_GNN-Distillation-461C.
URLs: https://anonymous.4open.science/r/APIM_and_GNN-Distillation-461C.
Authors: Kingsley Yeon, Promit Ghosal, Mihai Anitescu
Abstract: Efficient matrix trace estimation is essential for scalable computation of log-determinants, matrix norms, and distributional divergences. In many large-scale applications, the matrices involved are too large to store or access in full, making even a single matrix-vector (mat-vec) product infeasible. Instead, one often has access only to small subblocks of the matrix or localized matrix-vector products on restricted index sets. Hutch++ achieves optimal convergence rate but relies on randomized SVD and assumes full mat-vec access, making it difficult to apply in these constrained settings. We propose the Block-Orthonormal Stochastic Lanczos Quadrature (BOLT), which matches Hutch++ accuracy with a simpler implementation based on orthonormal block probes and Lanczos iterations. BOLT builds on the Stochastic Lanczos Quadrature (SLQ) framework, which combines random probing with Krylov subspace methods to efficiently approximate traces of matrix functions, and performs better than Hutch++ in near flat-spectrum regimes. To address memory limitations and partial access constraints, we introduce Subblock SLQ, a variant of BOLT that operates only on small principal submatrices. As a result, this framework yields a proxy KL divergence estimator and an efficient method for computing the Wasserstein-2 distance between Gaussians - both compatible with low-memory and partial-access regimes. We provide theoretical guarantees and demonstrate strong empirical performance across a range of high-dimensional settings.
Authors: Haiyu Deng, Yanna Jiang, Guangsheng Yu, Qin Wang, Xu Wang, Baihe Ma, Wei Ni, Ren Ping Liu
Abstract: Machine learning models are increasingly shared and outsourced, raising requirements of verifying training effort (Proof-of-Learning, PoL) to ensure claimed performance and establishing ownership (Proof-of-Ownership, PoO) for transactions. When models are trained by untrusted parties, PoL and PoO must be enforced together to enable protection, attribution, and compensation. However, existing studies typically address them separately, which not only weakens protection against forgery and privacy breaches but also leads to high verification overhead. We propose PoLO, a unified framework that simultaneously achieves PoL and PoO using chained watermarks. PoLO splits the training process into fine-grained training shards and embeds a dedicated watermark in each shard. Each watermark is generated using the hash of the preceding shard, certifying the training process of the preceding shard. The chained structure makes it computationally difficult to forge any individual part of the whole training process. The complete set of watermarks serves as the PoL, while the final watermark provides the PoO. PoLO offers more efficient and privacy-preserving verification compared to the vanilla PoL solutions that rely on gradient-based trajectory tracing and inadvertently expose training data during verification, while maintaining the same level of ownership assurance of watermark-based PoO schemes. Our evaluation shows that PoLO achieves 99% watermark detection accuracy for ownership verification, while preserving data privacy and cutting verification costs to just 1.5-10% of traditional methods. Forging PoLO demands 1.1-4x more resources than honest proof generation, with the original proof retaining over 90% detection accuracy even after attacks.
Authors: Qi Feng (Kyoto University), Hidetoshi Shimodaira (Kyoto University, RIKEN)
Abstract: Video-based spatial cognition is vital for robotics and embodied AI but challenges current Vision-Language Models (VLMs). This paper makes two key contributions. First, we introduce ViCA (Visuospatial Cognitive Assistant)-322K, a diverse dataset of 322,003 QA pairs from real-world indoor videos (ARKitScenes, ScanNet, ScanNet++), offering supervision for 3D metadata-grounded queries and video-based complex reasoning. Second, we develop ViCA-7B, fine-tuned on ViCA-322K, which achieves new state-of-the-art on all eight VSI-Bench tasks, outperforming existing models, including larger ones (e.g., +26.1 on Absolute Distance). For interpretability, we present ViCA-Thinking-2.68K, a dataset with explicit reasoning chains, and fine-tune ViCA-7B to create ViCA-7B-Thinking, a model that articulates its spatial reasoning. Our work highlights the importance of targeted data and suggests paths for improved temporal-spatial modeling. We release all resources to foster research in robust visuospatial intelligence.
Authors: Ruoqi Wang, Haitao Wang, Shaojie Guo, Qiong Luo
Abstract: Out-of-domain (OOD) robustness under domain adaptation settings, where labeled source data and unlabeled target data come from different distributions, is a key challenge in real-world applications. A common approach to improving OOD robustness is through data augmentations. However, in real-world scenarios, models trained with generic augmentations can only improve marginally when generalized under distribution shifts toward unlabeled target domains. While dataset-specific targeted augmentations can address this issue, they typically require expert knowledge and extensive prior data analysis to identify the nature of the datasets and domain shift. To address these challenges, we propose Frequency-Pixel Connect, a domain-adaptation framework that enhances OOD robustness by introducing a targeted augmentation in both the frequency space and pixel space. Specifically, we mix the amplitude spectrum and pixel content of a source image and a target image to generate augmented samples that introduce domain diversity while preserving the semantic structure of the source image. Unlike previous targeted augmentation methods that are both dataset-specific and limited to the pixel space, Frequency-Pixel Connect is dataset-agnostic, enabling broader and more flexible applicability beyond natural image datasets. We further analyze the effectiveness of Frequency-Pixel Connect by evaluating the performance of our method connecting same-class cross-domain samples while separating different-class examples. We demonstrate that Frequency-Pixel Connect significantly improves cross-domain connectivity and outperforms previous generic methods on four diverse real-world benchmarks across vision, medical, audio, and astronomical domains, and it also outperforms other dataset-specific targeted augmentation methods.
Authors: Albert Zhao, Stefano Soatto
Abstract: We describe a robust planning method for autonomous driving that mixes normal and adversarial agent predictions output by a diffusion model trained for motion prediction. We first train a diffusion model to learn an unbiased distribution of normal agent behaviors. We then generate a distribution of adversarial predictions by biasing the diffusion model at test time to generate predictions that are likely to collide with a candidate plan. We score plans using expected cost with respect to a mixture distribution of normal and adversarial predictions, leading to a planner that is robust against adversarial behaviors but not overly conservative when agents behave normally. Unlike current approaches, we do not use risk measures that over-weight adversarial behaviors while placing little to no weight on low-cost normal behaviors or use hard safety constraints that may not be appropriate for all driving scenarios. We show the effectiveness of our method on single-agent and multi-agent jaywalking scenarios as well as a red light violation scenario.
Authors: Yuancheng Jiang, Roland Yap, Zhenkai Liang
Abstract: In light of the rapid adoption of AI coding assistants, LLM-assisted development has become increasingly prevalent, creating an urgent need for robust evaluation of generated code quality. Existing benchmarks often require extensive manual effort to create static datasets, rely on indirect or insufficiently challenging tasks, depend on non-scalable ground truth, or neglect critical low-level security evaluations, particularly memory-safety issues. In this work, we introduce OSS-Bench, a benchmark generator that automatically constructs large-scale, live evaluation tasks from real-world open-source software. OSS-Bench replaces functions with LLM-generated code and evaluates them using three natural metrics: compilability, functional correctness, and memory safety, leveraging robust signals like compilation failures, test-suite violations, and sanitizer alerts as ground truth. In our evaluation, the benchmark, instantiated as OSS-Bench(php) and OSS-Bench(sql), profiles 17 diverse LLMs, revealing insights such as intra-family behavioral patterns and inconsistencies between model size and performance. Our results demonstrate that OSS-Bench mitigates overfitting by leveraging the evolving complexity of OSS and highlights LLMs' limited understanding of low-level code security via extended fuzzing experiments. Overall, OSS-Bench offers a practical and scalable framework for benchmarking the real-world coding capabilities of LLMs.
Authors: Axel Abels, Tom Lenaerts
Abstract: Despite their performance, large language models (LLMs) can inadvertently perpetuate biases found in the data they are trained on. By analyzing LLM responses to bias-eliciting headlines, we find that these models often mirror human biases. To address this, we explore crowd-based strategies for mitigating bias through response aggregation. We first demonstrate that simply averaging responses from multiple LLMs, intended to leverage the "wisdom of the crowd", can exacerbate existing biases due to the limited diversity within LLM crowds. In contrast, we show that locally weighted aggregation methods more effectively leverage the wisdom of the LLM crowd, achieving both bias mitigation and improved accuracy. Finally, recognizing the complementary strengths of LLMs (accuracy) and humans (diversity), we demonstrate that hybrid crowds containing both significantly enhance performance and further reduce biases across ethnic and gender-related contexts.
Authors: Siwen Zhang, Xizeng Zhao, Zhengzhi Deng, Zhaoyuan Huang, Gang Tao, Nuo Xu, Zhouteng Ye
Abstract: Accelerating the solution of nonlinear partial differential equations (PDEs) while maintaining accuracy at coarse spatiotemporal resolution remains a key challenge in scientific computing. Physics-informed machine learning (ML) methods such as Physics-Informed Neural Networks (PINNs) introduce prior knowledge through loss functions to ensure physical consistency, but their "soft constraints" are usually not strictly satisfied. Here, we propose LaPON, an operator network inspired by the Lagrange's mean value theorem, which embeds prior knowledge directly into the neural network architecture instead of the loss function, making the neural network naturally satisfy the given constraints. This is a hybrid framework that combines neural operators with traditional numerical methods, where neural operators are used to compensate for the effect of discretization errors on the analytical scale in under-resolution simulations. As evaluated on turbulence problem modeled by the Navier-Stokes equations (NSE), the multiple time step extrapolation accuracy and stability of LaPON exceed the direct numerical simulation baseline at 8x coarser grids and 8x larger time steps, while achieving a vorticity correlation of more than 0.98 with the ground truth. It is worth noting that the model can be well generalized to unseen flow states, such as turbulence with different forcing, without retraining. In addition, with the same training data, LaPON's comprehensive metrics on the out-of-distribution test set are at least approximately twice as good as two popular ML baseline methods. By combining numerical computing with machine learning, LaPON provides a scalable and reliable solution for high-fidelity fluid dynamics simulation, showing the potential for wide application in fields such as weather forecasting and engineering design.
Authors: Qi Feng (Kyoto University), Hidetoshi Shimodaira (Kyoto University, RIKEN)
Abstract: While Multimodal Large Language Models (MLLMs) excel at general vision-language tasks, visuospatial cognition - reasoning about spatial layouts, relations, and dynamics - remains a significant challenge. Existing models often lack the necessary architectural components and specialized training data for fine-grained spatial understanding. We introduce ViCA2 (Visuospatial Cognitive Assistant 2), a novel MLLM designed to enhance spatial reasoning. ViCA2 features a dual vision encoder architecture integrating SigLIP for semantics and Hiera for spatial structure, coupled with a token ratio control mechanism for efficiency. We also developed ViCA-322K, a new large-scale dataset with over 322,000 spatially grounded question-answer pairs for targeted instruction tuning. On the challenging VSI-Bench benchmark, our ViCA2-7B model achieves a state-of-the-art average score of 56.8, significantly surpassing larger open-source models (e.g., LLaVA-NeXT-Video-72B, 40.9) and leading proprietary models (Gemini-1.5 Pro, 45.4). This demonstrates the effectiveness of our approach in achieving strong visuospatial intelligence with a compact model. We release ViCA2, its codebase, and the ViCA-322K dataset to facilitate further research.
Authors: Fernando Zhapa-Camacho, Robert Hoehndorf
Abstract: Geometric embedding methods have shown to be useful for multi-hop reasoning on knowledge graphs by mapping entities and logical operations to geometric regions and geometric transformations, respectively. Geometric embeddings provide direct interpretability framework for queries. However, current methods have only leveraged the geometric construction of entities, failing to map logical operations to geometric transformations and, instead, using neural components to learn these operations. We introduce GeometrE, a geometric embedding method for multi-hop reasoning, which does not require learning the logical operations and enables full geometric interpretability. Additionally, unlike previous methods, we introduce a transitive loss function and show that it can preserve the logical rule $\forall a,b,c: r(a,b) \land r(b,c) \to r(a,c)$. Our experiments show that GeometrE outperforms current state-of-the-art methods on standard benchmark datasets.
Authors: Yinghao Zhu, Ziyi He, Haoran Hu, Xiaochen Zheng, Xichen Zhang, Zixiang Wang, Junyi Gao, Liantao Ma, Lequan Yu
Abstract: The rapid advancement of Large Language Models (LLMs) has stimulated interest in multi-agent collaboration for addressing complex medical tasks. However, the practical advantages of multi-agent collaboration approaches remain insufficiently understood. Existing evaluations often lack generalizability, failing to cover diverse tasks reflective of real-world clinical practice, and frequently omit rigorous comparisons against both single-LLM-based and established conventional methods. To address this critical gap, we introduce MedAgentBoard, a comprehensive benchmark for the systematic evaluation of multi-agent collaboration, single-LLM, and conventional approaches. MedAgentBoard encompasses four diverse medical task categories: (1) medical (visual) question answering, (2) lay summary generation, (3) structured Electronic Health Record (EHR) predictive modeling, and (4) clinical workflow automation, across text, medical images, and structured EHR data. Our extensive experiments reveal a nuanced landscape: while multi-agent collaboration demonstrates benefits in specific scenarios, such as enhancing task completeness in clinical workflow automation, it does not consistently outperform advanced single LLMs (e.g., in textual medical QA) or, critically, specialized conventional methods that generally maintain better performance in tasks like medical VQA and EHR-based prediction. MedAgentBoard offers a vital resource and actionable insights, emphasizing the necessity of a task-specific, evidence-based approach to selecting and developing AI solutions in medicine. It underscores that the inherent complexity and overhead of multi-agent collaboration must be carefully weighed against tangible performance gains. All code, datasets, detailed prompts, and experimental results are open-sourced at https://medagentboard.netlify.app/.
Authors: Kapil Dev (RMIT University, Melbourne, Australia)
Abstract: Human aesthetic preferences for 3D shapes are central to industrial design, virtual reality, and consumer product development. However, most computational models of 3D aesthetics lack empirical grounding in large-scale human judgments, limiting their practical relevance. We present a large-scale study of human preferences. We collected 22,301 pairwise comparisons across five object categories (chairs, tables, mugs, lamps, and dining chairs) via Amazon Mechanical Turk. Building on a previously published dataset~\cite{dev2020learning}, we introduce new non-linear modeling and cross-category analysis to uncover the geometric drivers of aesthetic preference. We apply the Bradley-Terry model to infer latent aesthetic scores and use Random Forests with SHAP analysis to identify and interpret the most influential geometric features (e.g., symmetry, curvature, compactness). Our cross-category analysis reveals both universal principles and domain-specific trends in aesthetic preferences. We focus on human interpretable geometric features to ensure model transparency and actionable design insights, rather than relying on black-box deep learning approaches. Our findings bridge computational aesthetics and cognitive science, providing practical guidance for designers and a publicly available dataset to support reproducibility. This work advances the understanding of 3D shape aesthetics through a human-centric, data-driven framework.
Authors: Andreas Floros, Seyed-Mohsen Moosavi-Dezfooli, Pier Luigi Dragotti
Abstract: We consider the problem of trustworthy image restoration, taking the form of a constrained optimization over the prior density. To this end, we develop generative models for the task of image super-resolution that respect the degradation process and that can be made asymptotically consistent with the low-resolution measurements, outperforming existing methods by a large margin in that respect.
Authors: Andi Han, Pierre-Louis Poirion, Akiko Takeda
Abstract: Optimization with orthogonality constraints frequently arises in various fields such as machine learning. Riemannian optimization offers a powerful framework for solving these problems by equipping the constraint set with a Riemannian manifold structure and performing optimization intrinsically on the manifold. This approach typically involves computing a search direction in the tangent space and updating variables via a retraction operation. However, as the size of the variables increases, the computational cost of the retraction can become prohibitively high, limiting the applicability of Riemannian optimization to large-scale problems. To address this challenge and enhance scalability, we propose a novel approach that restricts each update on a random submanifold, thereby significantly reducing the per-iteration complexity. We introduce two sampling strategies for selecting the random submanifolds and theoretically analyze the convergence of the proposed methods. We provide convergence results for general nonconvex functions and functions that satisfy Riemannian Polyak-Lojasiewicz condition as well as for stochastic optimization settings. Additionally, we demonstrate how our approach can be generalized to quotient manifolds derived from the orthogonal manifold. Extensive experiments verify the benefits of the proposed method, across a wide variety of problems.
Authors: Yepeng Weng, Qiao Hu, Xujie Chen, Li Liu, Dianwen Mei, Huishi Qiu, Jiang Tian, Zhongchao Shi
Abstract: Speculative decoding is a promising approach for accelerating large language models. The primary idea is to use a lightweight draft model to speculate the output of the target model for multiple subsequent timesteps, and then verify them in parallel to determine whether the drafted tokens should be accepted or rejected. To enhance acceptance rates, existing frameworks typically construct token trees containing multiple candidates in each timestep. However, their reliance on token-level verification mechanisms introduces two critical limitations: First, the probability distribution of a sequence differs from that of individual tokens, leading to suboptimal acceptance length. Second, current verification schemes begin from the root node and proceed layer by layer in a top-down manner. Once a parent node is rejected, all its child nodes should be discarded, resulting in inefficient utilization of speculative candidates. This paper introduces Traversal Verification, a novel speculative decoding algorithm that fundamentally rethinks the verification paradigm through leaf-to-root traversal. Our approach considers the acceptance of the entire token sequence from the current node to the root, and preserves potentially valid subsequences that would be prematurely discarded by existing methods. We theoretically prove that the probability distribution obtained through Traversal Verification is identical to that of the target model, guaranteeing lossless inference while achieving substantial acceleration gains. Experimental results across different large language models and multiple tasks show that our method consistently improves acceptance length and throughput over existing methods
Authors: Tim Y. J. Wang, Juan Kuntz, O. Deniz Akyildiz
Abstract: We introduce a novel particle-based algorithm for end-to-end training of latent diffusion models. We reformulate the training task as minimizing a free energy functional and obtain a gradient flow that does so. By approximating the latter with a system of interacting particles, we obtain the algorithm, which we underpin it theoretically by providing error guarantees. The novel algorithm compares favorably in experiments with previous particle-based methods and variational inference analogues.
Authors: Haodong Yang, Lei Wang, Md Zakir Hossain
Abstract: Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient fine-tuning (PEFT) method that injects two trainable low-rank matrices (A and B) into frozen pretrained models. While efficient, LoRA constrains updates to a fixed low-rank subspace (Delta W = BA), which can limit representational capacity and hinder downstream performance. We introduce Subspace Recomposition in Low-Rank Adaptation (SRLoRA) via importance-based fusion and reinitialization, a novel approach that enhances LoRA's expressiveness without compromising its lightweight structure. SRLoRA assigns importance scores to each LoRA pair (a column of B and the corresponding row of A), and dynamically recomposes the subspace during training. Less important pairs are fused into the frozen backbone, freeing capacity to reinitialize new pairs along unused principal directions derived from the pretrained weight's singular value decomposition. This mechanism enables continual subspace refreshment and richer adaptation over time, without increasing the number of trainable parameters. We evaluate SRLoRA on both language and vision tasks, including the GLUE benchmark and various image classification datasets. SRLoRA consistently achieves faster convergence and improved accuracy over standard LoRA, demonstrating its generality, efficiency, and potential for broader PEFT applications.
Authors: Shuguang Yu, Fan Zhou, Yingjie Zhang, Ziqi Chen, Hongtu Zhu
Abstract: This paper introduces a novel nonparametric method for estimating high-dimensional dynamic covariance matrices with multiple conditioning covariates, leveraging random forests and supported by robust theoretical guarantees. Unlike traditional static methods, our dynamic nonparametric covariance models effectively capture distributional heterogeneity. Furthermore, unlike kernel-smoothing methods, which are restricted to a single conditioning covariate, our approach accommodates multiple covariates in a fully nonparametric framework. To the best of our knowledge, this is the first method to use random forests for estimating high-dimensional dynamic covariance matrices. In high-dimensional settings, we establish uniform consistency theory, providing nonasymptotic error rates and model selection properties, even when the response dimension grows sub-exponentially with the sample size. These results hold uniformly across a range of conditioning variables. The method's effectiveness is demonstrated through simulations and a stock dataset analysis, highlighting its ability to model complex dynamics in high-dimensional scenarios.
Authors: Yuyang Ding, Dan Qiao, Juntao Li, Jiajie Xu, Pingfu Chao, Xiaofang Zhou, Min Zhang
Abstract: Distantly supervised named entity recognition (DS-NER) has emerged as a cheap and convenient alternative to traditional human annotation methods, enabling the automatic generation of training data by aligning text with external resources. Despite the many efforts in noise measurement methods, few works focus on the latent noise distribution between different distant annotation methods. In this work, we explore the effectiveness and robustness of DS-NER by two aspects: (1) distant annotation techniques, which encompasses both traditional rule-based methods and the innovative large language model supervision approach, and (2) noise assessment, for which we introduce a novel framework. This framework addresses the challenges by distinctly categorizing them into the unlabeled-entity problem (UEP) and the noisy-entity problem (NEP), subsequently providing specialized solutions for each. Our proposed method achieves significant improvements on eight real-world distant supervision datasets originating from three different data sources and involving four distinct annotation techniques, confirming its superiority over current state-of-the-art methods.
Authors: Antonio Candelieri, Andrea Ponti, Francesco Archetti
Abstract: Gaussian Process based Bayesian Optimization is a widely applied algorithm to learn and optimize under uncertainty, well-known for its sample efficiency. However, recently -- and more frequently -- research studies have empirically demonstrated that the Gaussian Process fitting procedure at its core could be its most relevant weakness. Fitting a Gaussian Process means tuning its kernel's hyperparameters to a set of observations, but the common Maximum Likelihood Estimation technique, usually appropriate for learning tasks, has shown different criticalities in Bayesian Optimization, making theoretical analysis of this algorithm an open challenge. Exploiting the analogy between Gaussian Processes and Gaussian Distributions, we present a new approach which uses a prefixed set of hyperparameters values to fit as many Gaussian Processes and then combines them into a unique model as a Wasserstein Barycenter of Gaussian Processes. We considered both "easy" test problems and others known to undermine the \textit{vanilla} Bayesian Optimization algorithm. The new method, namely Wasserstein Barycenter Gausssian Process based Bayesian Optimization (WBGP-BO), resulted promising and able to converge to the optimum, contrary to vanilla Bayesian Optimization, also on the most "tricky" test problems.
Authors: Yu Gui, Cong Ma, Zongming Ma
Abstract: Multi-modal contrastive learning as a self-supervised representation learning technique has achieved great success in foundation model training, such as CLIP~\citep{radford2021learning}. In this paper, we study the theoretical properties of the learned representations from multi-modal contrastive learning beyond linear representations and specific data distributions. Our analysis reveals that, enabled by temperature optimization, multi-modal contrastive learning not only maximizes mutual information between modalities but also adapts to intrinsic dimensions of data, which can be much lower than user-specified dimensions for representation vectors. Experiments on both synthetic and real-world datasets demonstrate the ability of contrastive learning to learn low-dimensional and informative representations, bridging theoretical insights and practical performance.
Authors: Amit Cohen, Lev Gloukhenki, Ravid Hadar, Eden Itah, Yehuda Shvut, Michael Schapira
Abstract: Congestion control (CC) crucially impacts user experience across Internet services like streaming, gaming, AR/VR, and connected cars. Traditionally, CC algorithm design seeks universal control rules that yield high performance across diverse application domains and networks. However, varying service needs and network conditions challenge this approach. We share operational experience with a system that automatically customizes congestion control logic to service needs and network conditions. We discuss design, deployment challenges, and solutions, highlighting performance benefits through case studies in streaming, gaming, connected cars, and more. Our system leverages PCC Vivace, an online-learning based congestion control protocol developed by researchers. Hence, along with insights from customizing congestion control, we also discuss lessons learned and modifications made to adapt PCC Vivace for real-world deployment.
Authors: Rohit Goswami (Science Institute and Faculty of Physical Sciences, University of Iceland, Reykjav\'ik, Iceland), Maxim Masterov (SURF, Amsterdam, The Netherlands), Satish Kamath (SURF, Amsterdam, The Netherlands), Alejandro Pe\~na-Torres (Science Institute and Faculty of Physical Sciences, University of Iceland, Reykjav\'ik, Iceland), Hannes J\'onsson (Science Institute and Faculty of Physical Sciences, University of Iceland, Reykjav\'ik, Iceland)
Abstract: The task of locating first order saddle points on high-dimensional surfaces describing the variation of energy as a function of atomic coordinates is an essential step for identifying the mechanism and estimating the rate of thermally activated events within the harmonic approximation of transition state theory. When combined directly with electronic structure calculations, the number of energy and atomic force evaluations needed for convergence is a primary issue. Here, we describe an efficient implementation of Gaussian process regression (GPR) acceleration of the minimum mode following method where a dimer is used to estimate the lowest eigenmode of the Hessian. A surrogate energy surface is constructed and updated after each electronic structure calculation. The method is applied to a test set of 500 molecular reactions previously generated by Hermez and coworkers [J. Chem. Theory Comput. 18, 6974 (2022)]. An order of magnitude reduction in the number of electronic structure calculations needed to reach the saddle point configurations is obtained by using the GPR compared to the dimer method. Despite the wide range in stiffness of the molecular degrees of freedom, the calculations are carried out using Cartesian coordinates and are found to require similar number of electronic structure calculations as an elaborate internal coordinate method implemented in the Sella software package. The present implementation of the GPR surrogate model in C++ is efficient enough for the wall time of the saddle point searches to be reduced in 3 out of 4 cases even though the calculations are carried out at a low Hartree-Fock level.
Authors: Guoyu Hu, Shaofeng Cai, Tien Tuan Anh Dinh, Zhongle Xie, Cong Yue, Gang Chen, Beng Chin Ooi
Abstract: Modern deep learning models capture the semantics of complex data by transforming them into high-dimensional embedding vectors. Emerging applications, such as retrieval-augmented generation, use approximate nearest neighbor (ANN) search in the embedding vector space to find similar data. Existing vector databases provide indexes for efficient ANN searches, with graph-based indexes being the most popular due to their low latency and high recall in real-world high-dimensional datasets. However, these indexes are costly to build, suffer from significant contention under concurrent read-write workloads, and scale poorly to multiple servers. Our goal is to build a vector database that achieves high throughput and high recall under concurrent read-write workloads. To this end, we first propose an ANN index with an explicit two-stage design combining a fast filter stage with highly compressed vectors and a refine stage to ensure recall, and we devise a novel lightweight machine learning technique to fine-tune the index parameters. We introduce an early termination check to dynamically adapt the search process for each query. Next, we add support for writes while maintaining search performance by decoupling the management of the learned parameters. Finally, we design HAKES, a distributed vector database that serves the new index in a disaggregated architecture. We evaluate our index and system against 12 state-of-the-art indexes and three distributed vector databases, using high-dimensional embedding datasets generated by deep learning models. The experimental results show that our index outperforms index baselines in the high recall region and under concurrent read-write workloads. Furthermore, \namesys{} is scalable and achieves up to $16\times$ higher throughputs than the baselines. The HAKES project is open-sourced at https://www.comp.nus.edu.sg/~dbsystem/hakes/.
Authors: Yuxin Ma, Dmitriy Kunisky
Abstract: We introduce a new family of algorithms for detecting and estimating a rank-one signal from a noisy observation under prior information about that signal's direction, focusing on examples where the signal is known to have entries biased to be positive. Given a matrix observation $\mathbf{Y}$, our algorithms construct a nonlinear Laplacian, another matrix of the form $\mathbf{Y} + \mathrm{diag}(\sigma(\mathbf{Y}\mathbf{1}))$ for a nonlinear $\sigma: \mathbb{R} \to \mathbb{R}$, and examine the top eigenvalue and eigenvector of this matrix. When $\mathbf{Y}$ is the (suitably normalized) adjacency matrix of a graph, our approach gives a class of algorithms that search for unusually dense subgraphs by computing a spectrum of the graph "deformed" by the degree profile $\mathbf{Y}\mathbf{1}$. We study the performance of such algorithms compared to direct spectral algorithms (the case $\sigma = 0$) on models of sparse principal component analysis with biased signals, including the Gaussian planted submatrix problem. For such models, we rigorously characterize the critical threshold strength of rank-one signal, as a function of the nonlinearity $\sigma$, at which an outlier eigenvalue appears in the spectrum of a nonlinear Laplacian. While identifying the $\sigma$ that minimizes this critical signal strength in closed form seems intractable, we explore three approaches to design $\sigma$ numerically: exhaustively searching over simple classes of $\sigma$, learning $\sigma$ from datasets of problem instances, and tuning $\sigma$ using black-box optimization of the critical signal strength. We find both theoretically and empirically that, if $\sigma$ is chosen appropriately, then nonlinear Laplacian spectral algorithms substantially outperform direct spectral algorithms, while avoiding the complexity of broader classes of algorithms like approximate message passing or general first order methods.
Authors: Ahmet Bilican, M. Ak{\i}n Y{\i}lmaz, A. Murat Tekalp, R. G\"okberk Cinbi\c{s}
Abstract: Efficiently adapting large foundation models is critical, especially with tight compute and memory budgets. Parameter-Efficient Fine-Tuning (PEFT) methods such as LoRA offer limited granularity and effectiveness in few-parameter regimes. We propose Wavelet Fine-Tuning (WaveFT), a novel PEFT method that learns highly sparse updates in the wavelet domain of residual matrices. WaveFT allows precise control of trainable parameters, offering fine-grained capacity adjustment and excelling with remarkably low parameter count, potentially far fewer than LoRA's minimum -- ideal for extreme parameter-efficient scenarios. In order to demonstrate the effect of the wavelet transform, we compare WaveFT with a special case, called SHiRA, that entails applying sparse updates directly in the weight domain. Evaluated on personalized text-to-image generation using Stable Diffusion XL as baseline, WaveFT significantly outperforms LoRA and other PEFT methods, especially at low parameter counts; achieving superior subject fidelity, prompt alignment, and image diversity.
Authors: Zahi Mizrahi, Shai Berkovitz, Nimrod Talmon, Michael Fire
Abstract: Keeping track of how lawmakers vote is essential for government transparency. While many parliamentary voting records are available online, they are often difficult to interpret, making it challenging to understand legislative behavior across parliaments and predict voting outcomes. Accurate prediction of votes has several potential benefits, from simplifying parliamentary work by filtering out bills with a low chance of passing to refining proposed legislation to increase its likelihood of approval. In this study, we leverage advanced machine learning and data analysis techniques to develop a comprehensive framework for predicting parliamentary voting outcomes across multiple legislatures. We introduce the Voting Prediction Framework (VPF) - a data-driven framework designed to forecast parliamentary voting outcomes at the individual legislator level and for entire bills. VPF consists of three key components: (1) Data Collection - gathering parliamentary voting records from multiple countries using APIs, web crawlers, and structured databases; (2) Parsing and Feature Integration - processing and enriching the data with meaningful features, such as legislator seniority, and content-based characteristics of a given bill; and (3) Prediction Models - using machine learning to forecast how each parliament member will vote and whether a bill is likely to pass. The framework will be open source, enabling anyone to use or modify the framework. To evaluate VPF, we analyzed over 5 million voting records from five countries - Canada, Israel, Tunisia, the United Kingdom and the USA. Our results show that VPF achieves up to 85% precision in predicting individual votes and up to 84% accuracy in predicting overall bill outcomes. These findings highlight VPF's potential as a valuable tool for political analysis, policy research, and enhancing public access to legislative decision-making.
Authors: A. Feder Cooper, Aaron Gokaslan, Amy B. Cyphert, Christopher De Sa, Mark A. Lemley, Daniel E. Ho, Percy Liang
Abstract: Plaintiffs and defendants in copyright lawsuits over generative AI often make sweeping, opposing claims about the extent to which large language models (LLMs) have memorized plaintiffs' protected expression. Drawing on adversarial ML and copyright law, we show that these polarized positions dramatically oversimplify the relationship between memorization and copyright. To do so, we leverage a recent probabilistic extraction technique to extract pieces of the Books3 dataset from 13 open-weight LLMs. Through numerous experiments, we show that it's possible to extract substantial parts of at least some books from different LLMs. This is evidence that the LLMs have memorized the extracted text; this memorized content is copied inside the model parameters. But the results are complicated: the extent of memorization varies both by model and by book. With our specific experiments, we find that the largest LLMs don't memorize most books -- either in whole or in part. However, we also find that Llama 3.1 70B memorizes some books, like Harry Potter and 1984, almost entirely. We discuss why our results have significant implications for copyright cases, though not ones that unambiguously favor either side.
Authors: Florent Chiaroni, Ali Ayub, Ola Ahmad
Abstract: In robotics applications, few-shot segmentation is crucial because it allows robots to perform complex tasks with minimal training data, facilitating their adaptation to diverse, real-world environments. However, pixel-level annotations of even small amount of images is highly time-consuming and costly. In this paper, we present a novel few-shot binary segmentation method based on bounding-box annotations instead of pixel-level labels. We introduce, ProMi, an efficient prototype-mixture-based method that treats the background class as a mixture of distributions. Our approach is simple, training-free, and effective, accommodating coarse annotations with ease. Compared to existing baselines, ProMi achieves the best results across different datasets with significant gains, demonstrating its effectiveness. Furthermore, we present qualitative experiments tailored to real-world mobile robot tasks, demonstrating the applicability of our approach in such scenarios. Our code: https://github.com/ThalesGroup/promi.
Authors: Junliang Ye, Lei Wang, Md Zakir Hossain
Abstract: Reconstructing natural images from functional magnetic resonance imaging (fMRI) data remains a core challenge in natural decoding due to the mismatch between the richness of visual stimuli and the noisy, low resolution nature of fMRI signals. While recent two-stage models, combining deep variational autoencoders (VAEs) with diffusion models, have advanced this task, they treat all spatial-frequency components of the input equally. This uniform treatment forces the model to extract meaning features and suppress irrelevant noise simultaneously, limiting its effectiveness. We introduce FreqSelect, a lightweight, adaptive module that selectively filters spatial-frequency bands before encoding. By dynamically emphasizing frequencies that are most predictive of brain activity and suppressing those that are uninformative, FreqSelect acts as a content-aware gate between image features and natural data. It integrates seamlessly into standard very deep VAE-diffusion pipelines and requires no additional supervision. Evaluated on the Natural Scenes dataset, FreqSelect consistently improves reconstruction quality across both low- and high-level metrics. Beyond performance gains, the learned frequency-selection patterns offer interpretable insights into how different visual frequencies are represented in the brain. Our method generalizes across subjects and scenes, and holds promise for extension to other neuroimaging modalities, offering a principled approach to enhancing both decoding accuracy and neuroscientific interpretability.
Authors: Qiang Fu, Andre Wibisono
Abstract: We study the Hamiltonian flow for optimization (HF-opt), which simulates the Hamiltonian dynamics for some integration time and resets the velocity to $0$ to decrease the objective function; this is the optimization analogue of the Hamiltonian Monte Carlo algorithm for sampling. For short integration time, HF-opt has the same convergence rates as gradient descent for minimizing strongly and weakly convex functions. We show that by randomizing the integration time in HF-opt, the resulting randomized Hamiltonian flow (RHF) achieves accelerated convergence rates in continuous time, similar to the rates for the accelerated gradient flow. We study a discrete-time implementation of RHF as the randomized Hamiltonian gradient descent (RHGD) algorithm. We prove that RHGD achieves the same accelerated convergence rates as Nesterov's accelerated gradient descent (AGD) for minimizing smooth strongly and weakly convex functions. We provide numerical experiments to demonstrate that RHGD is competitive with classical accelerated methods such as AGD across all settings and outperforms them in certain regimes.
Authors: Carl Edwards, Chi Han, Gawon Lee, Thao Nguyen, Bowen Jin, Chetan Kumar Prasad, Sara Szymku\'c, Bartosz A. Grzybowski, Ying Diao, Jiawei Han, Ge Liu, Hao Peng, Martin D. Burke, Heng Ji
Abstract: Despite their ability to understand chemical knowledge and accurately generate sequential representations, large language models (LLMs) remain limited in their capacity to propose novel molecules with drug-like properties. In addition, the molecules that LLMs propose can often be challenging to make in the lab. To more effectively enable the discovery of functional small molecules, LLMs need to learn a molecular language. However, LLMs are currently limited by encoding molecules from atoms. In this paper, we argue that just like tokenizing texts into (sub-)word tokens instead of characters, molecules should be decomposed and reassembled at the level of functional building blocks, i.e., parts of molecules that bring unique functions and serve as effective building blocks for real-world automated laboratory synthesis. This motivates us to propose mCLM, a modular Chemical-Language Model tokenizing molecules into building blocks and learning a bilingual language model of both natural language descriptions of functions and molecule building blocks. By reasoning on such functional building blocks, mCLM guarantees to generate efficiently synthesizable molecules thanks to recent progress in block-based chemistry, while also improving the functions of molecules in a principled manner. In experiments on 430 FDA-approved drugs, we find mCLM capable of significantly improving 5 out of 6 chemical functions critical to determining drug potentials. More importantly, mCLM can reason on multiple functions and improve the FDA-rejected drugs (``fallen angels'') over multiple iterations to greatly improve their shortcomings.
Authors: Paulo C. Marques F
Abstract: We consider the conformalization of a stacked ensemble of predictive models, showing that the potentially simple form of the meta-learner at the top of the stack enables a procedure with manageable computational cost that achieves approximate marginal validity without requiring the use of a separate calibration sample. Empirical results indicate that the method compares favorably to a standard inductive alternative.
Authors: Takeshi Kojima, Yaonan Zhu, Yusuke Iwasawa, Toshinori Kitamura, Gang Yan, Shu Morikuni, Ryosuke Takanami, Alfredo Solano, Tatsuya Matsushima, Akiko Murakami, Yutaka Matsuo
Abstract: Recent Foundation Model-enabled robotics (FMRs) display greatly improved general-purpose skills, enabling more adaptable automation than conventional robotics. Their ability to handle diverse tasks thus creates new opportunities to replace human labor. However, unlike general foundation models, FMRs interact with the physical world, where their actions directly affect the safety of humans and surrounding objects, requiring careful deployment and control. Based on this proposition, our survey comprehensively summarizes robot control approaches to mitigate physical risks by covering all the lifespan of FMRs ranging from pre-deployment to post-accident stage. Specifically, we broadly divide the timeline into the following three phases: (1) pre-deployment phase, (2) pre-incident phase, and (3) post-incident phase. Throughout this survey, we find that there is much room to study (i) pre-incident risk mitigation strategies, (ii) research that assumes physical interaction with humans, and (iii) essential issues of foundation models themselves. We hope that this survey will be a milestone in providing a high-resolution analysis of the physical risks of FMRs and their control, contributing to the realization of a good human-robot relationship.
Authors: Aditeya Baral, Allen George Ajith, Roshan Nayak, Mrityunjay Abhijeet Bhanja
Abstract: Code-mixed languages, characterized by frequent within-sentence language transitions, present structural challenges that standard language models fail to address. In this work, we propose CMLFormer, an enhanced multi-layer dual-decoder Transformer with a shared encoder and synchronized decoder cross-attention, designed to model the linguistic and semantic dynamics of code-mixed text. CMLFormer is pre-trained on an augmented Hinglish corpus with switching point and translation annotations with multiple new objectives specifically aimed at capturing switching behavior, cross-lingual structure, and code-mixing complexity. Our experiments show that CMLFormer improves F1 score, precision, and accuracy over other approaches on the HASOC-2021 benchmark under select pre-training setups. Attention analyses further show that it can identify and attend to switching points, validating its sensitivity to code-mixed structure. These results demonstrate the effectiveness of CMLFormer's architecture and multi-task pre-training strategy for modeling code-mixed languages.
Authors: Bohan Zhou, Shu Liu, Xinzhe Zuo, Wuchen Li
Abstract: We propose a class of discrete state sampling algorithms based on Nesterov's accelerated gradient method, which extends the classical Metropolis-Hastings (MH) algorithm. The evolution of the discrete states probability distribution governed by MH can be interpreted as a gradient descent direction of the Kullback--Leibler (KL) divergence, via a mobility function and a score function. Specifically, this gradient is defined on a probability simplex equipped with a discrete Wasserstein-2 metric with a mobility function. This motivates us to study a momentum-based acceleration framework using damped Hamiltonian flows on the simplex set, whose stationary distribution matches the discrete target distribution. Furthermore, we design an interacting particle system to approximate the proposed accelerated sampling dynamics. The extension of the algorithm with a general choice of potentials and mobilities is also discussed. In particular, we choose the accelerated gradient flow of the relative Fisher information, demonstrating the advantages of the algorithm in estimating discrete score functions without requiring the normalizing constant and keeping positive probabilities. Numerical examples, including sampling on a Gaussian mixture supported on lattices or a distribution on a hypercube, demonstrate the effectiveness of the proposed discrete-state sampling algorithm.
Authors: Thai Bui, Hoa T. Vu
Abstract: We study the densest subgraph problem and its variants through the lens of learning-augmented algorithms. For this problem, the greedy algorithm by Charikar (APPROX 2000) provides a linear-time $ 1/2 $-approximation, while computing the exact solution typically requires solving a linear program or performing maximum flow computations.We show that given a partial solution, i.e., one produced by a machine learning classifier that captures at least a $ (1 - \epsilon) $-fraction of nodes in the optimal subgraph, it is possible to design an extremely simple linear-time algorithm that achieves a provable $ (1 - \epsilon) $-approximation. Our approach also naturally extends to the directed densest subgraph problem and several NP-hard variants.An experiment on the Twitch Ego Nets dataset shows that our learning-augmented algorithm outperforms Charikar's greedy algorithm and a baseline that directly returns the predicted densest subgraph without additional algorithmic processing.
Authors: Toshihiro Ota, Yuma Fujimoto
Abstract: Understanding a dynamical system fundamentally relies on establishing an appropriate Hamiltonian function and elucidating its symmetries. By formulating agents' strategies and cumulative payoffs as canonically conjugate variables, we identify the Hamiltonian function that generates the dynamics of poly-matrix zero-sum games. We reveal the symmetries of our Hamiltonian and derive the associated conserved quantities, showing how the conservation of probability and the invariance of the Fenchel coupling are intrinsically encoded within the system. Furthermore, we propose the dissipation FTRL (DFTRL) dynamics by introducing a perturbation that dissipates the Fenchel coupling, proving convergence to the Nash equilibrium and linking DFTRL to last-iterate convergent algorithms. Our results highlight the potential of Hamiltonian dynamics in uncovering the structural properties of learning dynamics in games, and pave the way for broader applications of Hamiltonian dynamics in game theory and machine learning.
Authors: Ali Naseh, Harsh Chaudhari, Jaechul Roh, Mingshi Wu, Alina Oprea, Amir Houmansadr
Abstract: DeepSeek recently released R1, a high-performing large language model (LLM) optimized for reasoning tasks. Despite its efficient training pipeline, R1 achieves competitive performance, even surpassing leading reasoning models like OpenAI's o1 on several benchmarks. However, emerging reports suggest that R1 refuses to answer certain prompts related to politically sensitive topics in China. While existing LLMs often implement safeguards to avoid generating harmful or offensive outputs, R1 represents a notable shift - exhibiting censorship-like behavior on politically charged queries. In this paper, we investigate this phenomenon by first introducing a large-scale set of heavily curated prompts that get censored by R1, covering a range of politically sensitive topics, but are not censored by other models. We then conduct a comprehensive analysis of R1's censorship patterns, examining their consistency, triggers, and variations across topics, prompt phrasing, and context. Beyond English-language queries, we explore censorship behavior in other languages. We also investigate the transferability of censorship to models distilled from the R1 language model. Finally, we propose techniques for bypassing or removing this censorship. Our findings reveal possible additional censorship integration likely shaped by design choices during training or alignment, raising concerns about transparency, bias, and governance in language model deployment.
Authors: Ping Xu, Zhiyuan Ning, Pengjiang Li, Wenhao Liu, Pengyang Wang, Jiaxu Cui, Yuanchun Zhou, Pengfei Wang
Abstract: Single-cell RNA sequencing (scRNA-seq) reveals cell heterogeneity, with cell clustering playing a key role in identifying cell types and marker genes. Recent advances, especially graph neural networks (GNNs)-based methods, have significantly improved clustering performance. However, the analysis of scRNA-seq data remains challenging due to noise, sparsity, and high dimensionality. Compounding these challenges, GNNs often suffer from over-smoothing, limiting their ability to capture complex biological information. In response, we propose scSiameseClu, a novel Siamese Clustering framework for interpreting single-cell RNA-seq data, comprising of 3 key steps: (1) Dual Augmentation Module, which applies biologically informed perturbations to the gene expression matrix and cell graph relationships to enhance representation robustness; (2) Siamese Fusion Module, which combines cross-correlation refinement and adaptive information fusion to capture complex cellular relationships while mitigating over-smoothing; and (3) Optimal Transport Clustering, which utilizes Sinkhorn distance to efficiently align cluster assignments with predefined proportions while maintaining balance. Comprehensive evaluations on seven real-world datasets demonstrate that~\methodname~outperforms state-of-the-art methods in single-cell clustering, cell type annotation, and cell type classification, providing a powerful tool for scRNA-seq data interpretation.
Authors: Yunseok Jang, Yeda Song, Sungryull Sohn, Lajanugen Logeswaran, Tiange Luo, Dong-Ki Kim, Kyunghoon Bae, Honglak Lee
Abstract: Recent advancements in Large Language Models (LLMs) and Vision-Language Models (VLMs) have sparked significant interest in developing GUI visual agents. We introduce MONDAY (Mobile OS Navigation Task Dataset for Agents from YouTube), a large-scale dataset of 313K annotated frames from 20K instructional videos capturing diverse real-world mobile OS navigation across multiple platforms. Models that include MONDAY in their pre-training phases demonstrate robust cross-platform generalization capabilities, consistently outperforming models trained on existing single OS datasets while achieving an average performance gain of 18.11%p on an unseen mobile OS platform. To enable continuous dataset expansion as mobile platforms evolve, we present an automated framework that leverages publicly available video content to create comprehensive task datasets without manual annotation. Our framework comprises robust OCR-based scene detection (95.04% F1score), near-perfect UI element detection (99.87% hit ratio), and novel multi-step action identification to extract reliable action sequences across diverse interface configurations. We contribute both the MONDAY dataset and our automated collection framework to facilitate future research in mobile OS navigation.
Authors: Yifeng Jiao, Yuchen Liu, Yu Zhang, Xin Guo, Yushuai Wu, Chen Jiang, Jiyang Li, Hongwei Zhang, Limei Han, Xin Gao, Yuan Qi, Yuan Cheng
Abstract: The advent of single-cell Assay for Transposase-Accessible Chromatin using sequencing (scATAC-seq) offers an innovative perspective for deciphering regulatory mechanisms by assembling a vast repository of single-cell chromatin accessibility data. While foundation models have achieved significant success in single-cell transcriptomics, there is currently no foundation model for scATAC-seq that supports zero-shot high-quality cell identification and comprehensive multi-omics analysis simultaneously. Key challenges lie in the high dimensionality and sparsity of scATAC-seq data, as well as the lack of a standardized schema for representing open chromatin regions (OCRs). Here, we present \textbf{ChromFound}, a foundation model tailored for scATAC-seq. ChromFound utilizes a hybrid architecture and genome-aware tokenization to effectively capture genome-wide long contexts and regulatory signals from dynamic chromatin landscapes. Pretrained on 1.97 million cells from 30 tissues and 6 disease conditions, ChromFound demonstrates broad applicability across 6 diverse tasks. Notably, it achieves robust zero-shot performance in generating universal cell representations and exhibits excellent transferability in cell type annotation and cross-omics prediction. By uncovering enhancer-gene links undetected by existing computational methods, ChromFound offers a promising framework for understanding disease risk variants in the noncoding genome.
Authors: Ziqing Xing, Zhaoyang Zhang, Zirui Chen, Hongning Ruan, Zhaohui Yang
Abstract: In this paper, we incorporate physical knowledge into learning-based high-precision target sensing using the multi-view channel state information (CSI) between multiple base stations (BSs) and user equipment (UEs). Such kind of multi-view sensing problem can be naturally cast into a conditional generation framework. To this end, we design a bipartite neural network architecture, the first part of which uses an elaborately designed encoder to fuse the latent target features embedded in the multi-view CSI, and then the second uses them as conditioning inputs of a powerful generative model to guide the target's reconstruction. Specifically, the encoder is designed to capture the physical correlation between the CSI and the target, and also be adaptive to the numbers and positions of BS-UE pairs. Therein the view-specific nature of CSI is assimilated by introducing a spatial positional embedding scheme, which exploits the structure of electromagnetic(EM)-wave propagation channels. Finally, a conditional diffusion model with a weighted loss is employed to generate the target's point cloud from the fused features. Extensive numerical results demonstrate that the proposed generative multi-view (Gen-MV) sensing framework exhibits excellent flexibility and significant performance improvement on the reconstruction quality of target's shape and EM properties.
Authors: Mingyuan Zhou, Yi Gu, Zhendong Wang
Abstract: Diffusion distillation has emerged as a promising strategy for accelerating text-to-image (T2I) diffusion models by distilling a pretrained score network into a one- or few-step generator. While existing methods have made notable progress, they often rely on real or teacher-synthesized images to perform well when distilling high-resolution T2I diffusion models such as Stable Diffusion XL (SDXL), and their use of classifier-free guidance (CFG) introduces a persistent trade-off between text-image alignment and generation diversity. We address these challenges by optimizing Score identity Distillation (SiD) -- a data-free, one-step distillation framework -- for few-step generation. Backed by theoretical analysis that justifies matching a uniform mixture of outputs from all generation steps to the data distribution, our few-step distillation algorithm avoids step-specific networks and integrates seamlessly into existing pipelines, achieving state-of-the-art performance on SDXL at 1024x1024 resolution. To mitigate the alignment-diversity trade-off when real text-image pairs are available, we introduce a Diffusion GAN-based adversarial loss applied to the uniform mixture and propose two new guidance strategies: Zero-CFG, which disables CFG in the teacher and removes text conditioning in the fake score network, and Anti-CFG, which applies negative CFG in the fake score network. This flexible setup improves diversity without sacrificing alignment. Comprehensive experiments on SD1.5 and SDXL demonstrate state-of-the-art performance in both one-step and few-step generation settings, along with robustness to the absence of real images. Our efficient PyTorch implementation, along with the resulting one- and few-step distilled generators, will be released publicly as a separate branch at https://github.com/mingyuanzhou/SiD-LSG.
Authors: Haoyu Zhao, Yihan Geng, Shange Tang, Yong Lin, Bohan Lyu, Hongzhou Lin, Chi Jin, Sanjeev Arora
Abstract: LLM-based formal proof assistants (e.g., in Lean) hold great promise for automating mathematical discovery. But beyond syntactic correctness, do these systems truly understand mathematical structure as humans do? We investigate this question through the lens of mathematical inequalities -- a fundamental tool across many domains. While modern provers can solve basic inequalities, we probe their ability to handle human-intuitive compositionality. We introduce Ineq-Comp, a benchmark built from elementary inequalities through systematic transformations, including variable duplication, algebraic rewriting, and multi-step composition. Although these problems remain easy for humans, we find that most provers -- including Goedel, STP, and Kimina-7B -- struggle significantly. DeepSeek-Prover-V2-7B shows relative robustness -- possibly because it is trained to decompose the problems into sub-problems -- but still suffers a 20\% performance drop (pass@32). Strikingly, performance remains poor for all models even when formal proofs of the constituent parts are provided in context, revealing that the source of weakness is indeed in compositional reasoning. Our results expose a persisting gap between the generalization behavior of current AI provers and human mathematical intuition.
Authors: Joel Jang, Seonghyeon Ye, Zongyu Lin, Jiannan Xiang, Johan Bjorck, Yu Fang, Fengyuan Hu, Spencer Huang, Kaushil Kundalia, Yen-Chen Lin, Loic Magne, Ajay Mandlekar, Avnish Narayan, You Liang Tan, Guanzhi Wang, Jing Wang, Qi Wang, Yinzhen Xu, Xiaohui Zeng, Kaiyuan Zheng, Ruijie Zheng, Ming-Yu Liu, Luke Zettlemoyer, Dieter Fox, Jan Kautz, Scott Reed, Yuke Zhu, Linxi Fan
Abstract: We introduce DreamGen, a simple yet highly effective 4-stage pipeline for training robot policies that generalize across behaviors and environments through neural trajectories - synthetic robot data generated from video world models. DreamGen leverages state-of-the-art image-to-video generative models, adapting them to the target robot embodiment to produce photorealistic synthetic videos of familiar or novel tasks in diverse environments. Since these models generate only videos, we recover pseudo-action sequences using either a latent action model or an inverse-dynamics model (IDM). Despite its simplicity, DreamGen unlocks strong behavior and environment generalization: a humanoid robot can perform 22 new behaviors in both seen and unseen environments, while requiring teleoperation data from only a single pick-and-place task in one environment. To evaluate the pipeline systematically, we introduce DreamGen Bench, a video generation benchmark that shows a strong correlation between benchmark performance and downstream policy success. Our work establishes a promising new axis for scaling robot learning well beyond manual data collection.
Authors: Subhayan Saha, Giovanni Barbarino, Nicolas Gillis
Abstract: Tensor decompositions have become a central tool in data science, with applications in areas such as data analysis, signal processing, and machine learning. A key property of many tensor decompositions, such as the canonical polyadic decomposition, is identifiability: the factors are unique, up to trivial scaling and permutation ambiguities. This allows one to recover the groundtruth sources that generated the data. The Tucker decomposition (TD) is a central and widely used tensor decomposition model. However, it is in general not identifiable. In this paper, we study the identifiability of the nonnegative TD (nTD). By adapting and extending identifiability results of nonnegative matrix factorization (NMF), we provide uniqueness results for nTD. Our results require the nonnegative matrix factors to have some degree of sparsity (namely, satisfy the separability condition, or the sufficiently scattered condition), while the core tensor only needs to have some slices (or linear combinations of them) or unfoldings with full column rank (but does not need to be nonnegative). Under such conditions, we derive several procedures, using either unfoldings or slices of the input tensor, to obtain identifiable nTDs by minimizing the volume of unfoldings or slices of the core tensor.
Authors: Filippo Leveni, Matteo Mistura, Francesco Iubatti, Carmine Giangregorio, Nicol\`o Pastore, Cesare Alippi, Giacomo Boracchi
Abstract: Malware are malicious programs that are grouped into families based on their penetration technique, source code, and other characteristics. Classifying malware programs into their respective families is essential for building effective defenses against cyber threats. Machine learning models have a huge potential in malware detection on mobile devices, as malware families can be recognized by classifying permission data extracted from Android manifest files. Still, the malware classification task is challenging due to the high-dimensional nature of permission data and the limited availability of training samples. In particular, the steady emergence of new malware families makes it impossible to acquire a comprehensive training set covering all the malware classes. In this work, we present a malware classification system that, on top of classifying known malware, detects new ones. In particular, we combine an open-set recognition technique developed within the computer vision community, namely MaxLogit, with a tree-based Gradient Boosting classifier, which is particularly effective in classifying high-dimensional data. Our solution turns out to be very practical, as it can be seamlessly employed in a standard classification workflow, and efficient, as it adds minimal computational overhead. Experiments on public and proprietary datasets demonstrate the potential of our solution, which has been deployed in a business environment.
Authors: Matias Quintana, Youlong Gu, Xiucheng Liang, Yujun Hou, Koichi Ito, Yihan Zhu, Mahmoud Abdelrahman, Filip Biljecki
Abstract: Understanding people's preferences and needs is crucial for urban planning decisions, yet current approaches often combine them from multi-cultural and multi-city populations, obscuring important demographic differences and risking amplifying biases. We conducted a large-scale urban visual perception survey of streetscapes worldwide using street view imagery, examining how demographics -- including gender, age, income, education, race and ethnicity, and, for the first time, personality traits -- shape perceptions among 1,000 participants, with balanced demographics, from five countries and 45 nationalities. This dataset, introduced as Street Perception Evaluation Considering Socioeconomics (SPECS), exhibits statistically significant differences in perception scores in six traditionally used indicators (safe, lively, wealthy, beautiful, boring, and depressing) and four new ones we propose (live nearby, walk, cycle, green) among demographics and personalities. We revealed that location-based sentiments are carried over in people's preferences when comparing urban streetscapes with other cities. Further, we compared the perception scores based on where participants and streetscapes are from. We found that an off-the-shelf machine learning model trained on an existing global perception dataset tends to overestimate positive indicators and underestimate negative ones compared to human responses, suggesting that targeted intervention should consider locals' perception. Our study aspires to rectify the myopic treatment of street perception, which rarely considers demographics or personality traits.
Authors: Yiling Tao, Shuyi Wang, Jiaxi Yang, Guido Zuccon
Abstract: This paper reports on findings from a comparative study on the effectiveness and efficiency of federated unlearning strategies within Federated Online Learning to Rank (FOLTR), with specific attention to systematically analysing the unlearning capabilities of methods in a verifiable manner. Federated approaches to ranking of search results have recently garnered attention to address users privacy concerns. In FOLTR, privacy is safeguarded by collaboratively training ranking models across decentralized data sources, preserving individual user data while optimizing search results based on implicit feedback, such as clicks. Recent legislation introduced across numerous countries is establishing the so called "the right to be forgotten", according to which services based on machine learning models like those in FOLTR should provide capabilities that allow users to remove their own data from those used to train models. This has sparked the development of unlearning methods, along with evaluation practices to measure whether unlearning of a user data successfully occurred. Current evaluation practices are however often controversial, necessitating the use of multiple metrics for a more comprehensive assessment -- but previous proposals of unlearning methods only used single evaluation metrics. This paper addresses this limitation: our study rigorously assesses the effectiveness of unlearning strategies in managing both under-unlearning and over-unlearning scenarios using adapted, and newly proposed evaluation metrics. Thanks to our detailed analysis, we uncover the strengths and limitations of five unlearning strategies, offering valuable insights into optimizing federated unlearning to balance data privacy and system performance within FOLTR. We publicly release our code and complete results at https://github.com/Iris1026/Unlearning-for-FOLTR.git.
Authors: Shibo Hong, Jiahao Ying, Haiyuan Liang, Mengdi Zhang, Jun Kuang, Jiazheng Zhang, Yixin Cao
Abstract: Evaluating the open-ended outputs of large language models (LLMs) has become a bottleneck as model capabilities, task diversity, and modality coverage rapidly expand. Existing "LLM-as-a-Judge" evaluators are typically narrow in a few tasks, aspects, or modalities, and easily suffer from low consistency. In this paper, we argue that explicit, fine-grained aspect specification is the key to both generalizability and objectivity in automated evaluation. To do so, we introduce a hierarchical aspect taxonomy spanning 112 aspects that unifies evaluation across four representative settings - Natural Language Generation, Image Understanding, Image Generation, and Interleaved Text-and-Image Generation. Building on this taxonomy, we create FRAbench, a benchmark comprising 60.4k pairwise samples with 325k aspect-level labels obtained from a combination of human and LLM annotations. FRAbench provides the first large-scale, multi-modal resource for training and meta-evaluating fine-grained LMM judges. Leveraging FRAbench, we develop GenEval, a fine-grained evaluator generalizable across tasks and modalities. Experiments show that GenEval (i) attains high agreement with GPT-4o and expert annotators, (ii) transfers robustly to unseen tasks and modalities, and (iii) reveals systematic weaknesses of current LMMs on evaluation.
Authors: Konstantina Lelova, Gregory F. Cooper, Sofia Triantafillou
Abstract: Transporting causal information learned from experiments in one population to another is a critical challenge in clinical research and decision-making. Causal transportability uses causal graphs to model differences between the source and target populations and identifies conditions under which causal effects learned from experiments can be reused in a different population. Similarly, causal identifiability identifies conditions under which causal effects can be estimated from observational data. However, these approaches rely on knowing the causal graph, which is often unavailable in real-world settings. In this work, we propose a Bayesian method for assessing whether Z-specific (conditional) causal effects are both identifiable and transportable, without knowing the causal graph. Our method combines experimental data from the source population with observational data from the target population to compute the probability that a causal effect is both identifiable from observational data and transportable. When this holds, we leverage both observational data from the target domain and experimental data from the source domain to obtain an unbiased, efficient estimator of the causal effect in the target population. Using simulations, we demonstrate that our method correctly identifies transportable causal effects and improves causal effect estimation.
Authors: Jiawen Xu, Odej Kao, Margret Keuper
Abstract: Open set recognition (OSR) is devised to address the problem of detecting novel classes during model inference. Even in recent vision models, this remains an open issue which is receiving increasing attention. Thereby, a crucial challenge is to learn features that are relevant for unseen categories from given data, for which these features might not be discriminative. To facilitate this process and "optimize to learn" more diverse features, we propose GradMix, a data augmentation method that dynamically leverages gradient-based attribution maps of the model during training to mask out already learned concepts. Thus GradMix encourages the model to learn a more complete set of representative features from the same data source. Extensive experiments on open set recognition, close set classification, and out-of-distribution detection reveal that our method can often outperform the state-of-the-art. GradMix can further increase model robustness to corruptions as well as downstream classification performance for self-supervised learning, indicating its benefit for model generalization.
Authors: Yanbin Yin, Kun Zhou, Zhen Wang, Xiangdong Zhang, Yifei Shao, Shibo Hao, Yi Gu, Jieyuan Liu, Somanshu Singla, Tianyang Liu, Eric P. Xing, Zhengzhong Liu, Haojian Jin, Zhiting Hu
Abstract: The recent explosion of large language models (LLMs), each with its own general or specialized strengths, makes scalable, reliable benchmarking more urgent than ever. Standard practices nowadays face fundamental trade-offs: closed-ended question-based benchmarks (eg MMLU) struggle with saturation as newer models emerge, while crowd-sourced leaderboards (eg Chatbot Arena) rely on costly and slow human judges. Recently, automated methods (eg LLM-as-a-judge) shed light on the scalability, but risk bias by relying on one or a few "authority" models. To tackle these issues, we propose Decentralized Arena (dearena), a fully automated framework leveraging collective intelligence from all LLMs to evaluate each other. It mitigates single-model judge bias by democratic, pairwise evaluation, and remains efficient at scale through two key components: (1) a coarse-to-fine ranking algorithm for fast incremental insertion of new models with sub-quadratic complexity, and (2) an automatic question selection strategy for the construction of new evaluation dimensions. Across extensive experiments across 66 LLMs, dearena attains up to 97% correlation with human judgements, while significantly reducing the cost. Our code and data will be publicly released on https://github.com/maitrix-org/de-arena.
Authors: Wei-Chen Liao, Ti-Rong Wu, I-Chen Wu
Abstract: Multi-agent reinforcement Learning (MARL) is often challenged by the sight range dilemma, where agents either receive insufficient or excessive information from their environment. In this paper, we propose a novel method, called Dynamic Sight Range Selection (DSR), to address this issue. DSR utilizes an Upper Confidence Bound (UCB) algorithm and dynamically adjusts the sight range during training. Experiment results show several advantages of using DSR. First, we demonstrate using DSR achieves better performance in three common MARL environments, including Level-Based Foraging (LBF), Multi-Robot Warehouse (RWARE), and StarCraft Multi-Agent Challenge (SMAC). Second, our results show that DSR consistently improves performance across multiple MARL algorithms, including QMIX and MAPPO. Third, DSR offers suitable sight ranges for different training steps, thereby accelerating the training process. Finally, DSR provides additional interpretability by indicating the optimal sight range used during training. Unlike existing methods that rely on global information or communication mechanisms, our approach operates solely based on the individual sight ranges of agents. This approach offers a practical and efficient solution to the sight range dilemma, making it broadly applicable to real-world complex environments.
Authors: Muhamed Kuric, Martin Zach, Andreas Habring, Michael Unser, Thomas Pock
Abstract: We consider the problem of sampling from a product-of-experts-type model that encompasses many standard prior and posterior distributions commonly found in Bayesian imaging. We show that this model can be easily lifted into a novel latent variable model, which we refer to as a Gaussian latent machine. This leads to a general sampling approach that unifies and generalizes many existing sampling algorithms in the literature. Most notably, it yields a highly efficient and effective two-block Gibbs sampling approach in the general case, while also specializing to direct sampling algorithms in particular cases. Finally, we present detailed numerical experiments that demonstrate the efficiency and effectiveness of our proposed sampling approach across a wide range of prior and posterior sampling problems from Bayesian imaging.
Authors: Adarsh Singh
Abstract: The development of novel pharmaceuticals represents a significant challenge in modern science, with substantial costs and time investments. Deep generative models have emerged as promising tools for accelerating drug discovery by efficiently exploring the vast chemical space. However, this rapidly evolving field lacks standardized evaluation protocols, impeding fair comparison between approaches. This research presents an extensive analysis of the Molecular Sets (MOSES) platform, a comprehensive benchmarking framework designed to standardize evaluation of deep generative models in molecular design. Through rigorous assessment of multiple generative architectures, including recurrent neural networks, variational autoencoders, and generative adversarial networks, we examine their capabilities in generating valid, unique, and novel molecular structures while maintaining specific chemical properties. Our findings reveal that different architectures exhibit complementary strengths across various metrics, highlighting the complex trade-offs between exploration and exploitation in chemical space. This study provides detailed insights into the current state of the art in molecular generation and establishes a foundation for future advancements in AI-driven drug discovery.
Authors: Yu Fan, Jingwei Ni, Jakob Merane, Etienne Salimbeni, Yang Tian, Yoan Hermstr\"uwer, Yinya Huang, Mubashara Akhtar, Florian Geering, Oliver Dreyer, Daniel Brunner, Markus Leippold, Mrinmaya Sachan, Alexander Stremitzer, Christoph Engel, Elliott Ash, Joel Niklaus
Abstract: Long-form legal reasoning remains a key challenge for large language models (LLMs) in spite of recent advances in test-time scaling. We introduce LEXam, a novel benchmark derived from 340 law exams spanning 116 law school courses across a range of subjects and degree levels. The dataset comprises 4,886 law exam questions in English and German, including 2,841 long-form, open-ended questions and 2,045 multiple-choice questions. Besides reference answers, the open questions are also accompanied by explicit guidance outlining the expected legal reasoning approach such as issue spotting, rule recall, or rule application. Our evaluation on both open-ended and multiple-choice questions present significant challenges for current LLMs; in particular, they notably struggle with open questions that require structured, multi-step legal reasoning. Moreover, our results underscore the effectiveness of the dataset in differentiating between models with varying capabilities. Adopting an LLM-as-a-Judge paradigm with rigorous human expert validation, we demonstrate how model-generated reasoning steps can be evaluated consistently and accurately. Our evaluation setup provides a scalable method to assess legal reasoning quality beyond simple accuracy metrics. Project page: https://lexam-benchmark.github.io/
Authors: Marin \v{S}ola, Peter B\"uhlmann, Xinwei Shen
Abstract: Distributional robustness is a central goal of prediction algorithms due to the prevalent distribution shifts in real-world data. The prediction model aims to minimize the worst-case risk among a class of distributions, a.k.a., an uncertainty set. Causality provides a modeling framework with a rigorous robustness guarantee in the above sense, where the uncertainty set is data-driven rather than pre-specified as in traditional distributional robustness optimization. However, current causality-inspired robustness methods possess finite-radius robustness guarantees only in the linear settings, where the causal relationships among the covariates and the response are linear. In this work, we propose a nonlinear method under a causal framework by incorporating recent developments in identifiable representation learning and establish a distributional robustness guarantee. To our best knowledge, this is the first causality-inspired robustness method with such a finite-radius robustness guarantee in nonlinear settings. Empirical validation of the theoretical findings is conducted on both synthetic data and real-world single-cell data, also illustrating that finite-radius robustness is crucial.
Authors: Maytus Piriyajitakonkij, Rujikorn Charakorn, Weicheng Tao, Wei Pan, Mingfei Sun, Cheston Tan, Mengmi Zhang
Abstract: Early cavemen relied on gestures, vocalizations, and simple signals to coordinate, plan, avoid predators, and share resources. Today, humans collaborate using complex languages to achieve remarkable results. What drives this evolution in communication? How does language emerge, adapt, and become vital for teamwork? Understanding the origins of language remains a challenge. A leading hypothesis in linguistics and anthropology posits that language evolved to meet the ecological and social demands of early human cooperation. Language did not arise in isolation, but through shared survival goals. Inspired by this view, we investigate the emergence of language in multi-agent Foraging Games. These environments are designed to reflect the cognitive and ecological constraints believed to have influenced the evolution of communication. Agents operate in a shared grid world with only partial knowledge about other agents and the environment, and must coordinate to complete games like picking up high-value targets or executing temporally ordered actions. Using end-to-end deep reinforcement learning, agents learn both actions and communication strategies from scratch. We find that agents develop communication protocols with hallmark features of natural language: arbitrariness, interchangeability, displacement, cultural transmission, and compositionality. We quantify each property and analyze how different factors, such as population size and temporal dependencies, shape specific aspects of the emergent language. Our framework serves as a platform for studying how language can evolve from partial observability, temporal reasoning, and cooperative goals in embodied multi-agent settings. We will release all data, code, and models publicly.
Authors: Yeonsu Kim, Junhan Lee, John T. Hwang, Bingran Wang, Dongjin Lee
Abstract: Forward uncertainty quantification in dynamic systems is challenging due to non-smooth or locally oscillating nonlinear behaviors. Spline dimensional decomposition (SDD) effectively addresses such nonlinearity by partitioning input coordinates via knot placement, yet its accuracy is highly sensitive to the location of internal knots. Optimizing knots through sequential quadratic programming can be effective, yet the optimization process becomes computationally intense. We propose a computationally efficient, interpolation-based method for optimal knot selection in SDD. The method involves three steps: (1) interpolating input-output profiles, (2) defining subinterval-based reference regions, and (3) selecting optimal knot locations at maximum gradient points within each region. The resulting knot vector is then applied to SDD for accurate approximation of non-smooth and locally oscillating responses. A modal analysis of a lower control arm demonstrates that SDD with the proposed knot selection achieves higher accuracy than SDD with uniformly or randomly spaced knots, and also a Gaussian process surrogate model. The proposed SDD exhibits the lowest relative variance error (2.89%), compared to SDD with uniformly spaced knots (12.310%), randomly spaced knots (15.274%), and Gaussian process (5.319%) in the first natural frequency distribution. All surrogate models are constructed using the same 401 simulation datasets, and the relative errors are evaluated against a 2000-sample Monte Carlo simulation. The scalability and applicability of proposed method are demonstrated through stochastic and reliability analyses of mathematical functions (N=1, 3) and a lower control arm system (N=10). The results confirm that both second-moment statistics and reliability estimates can be accurately achieved with only a few hundred function evaluations or finite element simulations.
Authors: Chenxi Liu, Yongqiang Chen, Tongliang Liu, James Cheng, Bo Han, Kun Zhang
Abstract: System 2 reasoning is one of the defining characteristics of intelligence, which requires slow and logical thinking. Human conducts System 2 reasoning via the language of thoughts that organizes the reasoning process as a causal sequence of mental language, or thoughts. Recently, it has been observed that System 2 reasoning can be elicited from Large Language Models (LLMs) pre-trained on large-scale natural languages. However, in this work, we show that there is a significant gap between the modeling of languages and thoughts. As language is primarily a tool for humans to share knowledge and thinking, modeling human language can easily absorb language biases into LLMs deviated from the chain of thoughts in minds. Furthermore, we show that the biases will mislead the eliciting of "thoughts" in LLMs to focus only on a biased part of the premise. To this end, we propose a new prompt technique termed Language-of-Thoughts (LoT) to demonstrate and alleviate this gap. Instead of directly eliciting the chain of thoughts from partial information, LoT instructs LLMs to adjust the order and token used for the expressions of all the relevant information. We show that the simple strategy significantly reduces the language modeling biases in LLMs and improves the performance of LLMs across a variety of reasoning tasks.
Authors: Hao Fang, Kai Huang, Hao Ye, Chongtao Guo, Le Liang, Xiao Li, Shi Jin
Abstract: The pursuit of rate maximization in wireless communication frequently encounters substantial challenges associated with user fairness. This paper addresses these challenges by exploring a novel power allocation approach for delay optimization, utilizing graph neural networks (GNNs)-based reinforcement learning (RL) in device-to-device (D2D) communication. The proposed approach incorporates not only channel state information but also factors such as packet delay, the number of backlogged packets, and the number of transmitted packets into the components of the state information. We adopt a centralized RL method, where a central controller collects and processes the state information. The central controller functions as an agent trained using the proximal policy optimization (PPO) algorithm. To better utilize topology information in the communication network and enhance the generalization of the proposed method, we embed GNN layers into both the actor and critic networks of the PPO algorithm. This integration allows for efficient parameter updates of GNNs and enables the state information to be parameterized as a low-dimensional embedding, which is leveraged by the agent to optimize power allocation strategies. Simulation results demonstrate that the proposed method effectively reduces average delay while ensuring user fairness, outperforms baseline methods, and exhibits scalability and generalization capability.
Authors: Xiao Wang, Yu Jin, Lan Chen, Bo Jiang, Lin Zhu, Yonghong Tian, Jin Tang, Bin Luo
Abstract: Event-based Vision Sensors (EVS) have demonstrated significant advantages over traditional RGB frame-based cameras in low-light conditions, high-speed motion capture, and low latency. Consequently, object detection based on EVS has attracted increasing attention from researchers. Current event stream object detection algorithms are typically built upon Convolutional Neural Networks (CNNs) or Transformers, which either capture limited local features using convolutional filters or incur high computational costs due to the utilization of self-attention. Recently proposed vision heat conduction backbone networks have shown a good balance between efficiency and accuracy; however, these models are not specifically designed for event stream data. They exhibit weak capability in modeling object contour information and fail to exploit the benefits of multi-scale features. To address these issues, this paper proposes a novel dynamic graph induced contour-aware heat conduction network for event stream based object detection, termed CvHeat-DET. The proposed model effectively leverages the clear contour information inherent in event streams to predict the thermal diffusivity coefficients within the heat conduction model, and integrates hierarchical structural graph features to enhance feature learning across multiple scales. Extensive experiments on three benchmark datasets for event stream-based object detection fully validated the effectiveness of the proposed model. The source code of this paper will be released on https://github.com/Event-AHU/OpenEvDET.
Authors: Zhihe Yang, Xufang Luo, Zilong Wang, Dongqi Han, Zhiyuan He, Dongsheng Li, Yunjian Xu
Abstract: Reinforcement learning (RL) has become a cornerstone for enhancing the reasoning capabilities of large language models (LLMs), with recent innovations such as Group Relative Policy Optimization (GRPO) demonstrating exceptional effectiveness. In this study, we identify a critical yet underexplored issue in RL training: low-probability tokens disproportionately influence model updates due to their large gradient magnitudes. This dominance hinders the effective learning of high-probability tokens, whose gradients are essential for LLMs' performance but are substantially suppressed. To mitigate this interference, we propose two novel methods: Advantage Reweighting and Low-Probability Token Isolation (Lopti), both of which effectively attenuate gradients from low-probability tokens while emphasizing parameter updates driven by high-probability tokens. Our approaches promote balanced updates across tokens with varying probabilities, thereby enhancing the efficiency of RL training. Experimental results demonstrate that they substantially improve the performance of GRPO-trained LLMs, achieving up to a 46.2% improvement in K&K Logic Puzzle reasoning tasks. Our implementation is available at https://github.com/zhyang2226/AR-Lopti.
Authors: Jeffrey T. H. Wong, Cheng Zhang, Xinye Cao, Pedro Gimenes, George A. Constantinides, Wayne Luk, Yiren Zhao
Abstract: Large language models have demonstrated remarkable performance; however, their massive parameter counts make deployment highly expensive. Low-rank approximation offers a promising compression solution, yet existing approaches have two main limitations: (1) They focus on minimizing the output error of individual linear layers, without considering the architectural characteristics of Transformers, and (2) they decompose a large weight matrix into two small low-rank matrices. Consequently, these methods often fall short compared to other compression techniques like pruning and quantization, and introduce runtime overhead such as the extra GEMM kernel launches for decomposed small matrices. To address these limitations, we propose $\tt A^\tt 3$, a post-training low-rank approximation framework. $\tt A^\tt 3$ splits a Transformer layer into three functional components, namely $\tt QK$, $\tt OV$, and $\tt MLP$. For each component, $\tt A^\tt 3$ provides an analytical solution that reduces the hidden dimension size inside each component while minimizing the component's functional loss ($\it i.e.$, error in attention scores, attention outputs, and MLP outputs). This approach directly reduces model sizes, KV cache sizes, and FLOPs without introducing any runtime overheads. In addition, it provides a new narrative in advancing the optimization problem from singular linear layer loss optimization toward improved end-to-end performance. Through extensive experiments, we show that $\tt A^\tt 3$ maintains superior performance compared to SoTAs. For example, under the same reduction budget in computation and memory, our low-rank approximated LLaMA 3.1-70B achieves a perplexity of 4.69 on WikiText-2, outperforming the previous SoTA's 7.87 by 3.18. We also demonstrate the versatility of $\tt A^\tt 3$, including KV cache compression, quantization, and mixed-rank assignments for enhanced performance.
Authors: Anthony Bardou, Patrick Thiran
Abstract: Time-Varying Bayesian Optimization (TVBO) is the go-to framework for optimizing a time-varying black-box objective function that may be noisy and expensive to evaluate. Is it possible for the instantaneous regret of a TVBO algorithm to vanish asymptotically, and if so, when? We answer this question of great theoretical importance by providing algorithm-independent lower regret bounds and upper regret bounds for TVBO algorithms, from which we derive sufficient conditions for a TVBO algorithm to have the no-regret property. Our analysis covers all major classes of stationary kernel functions.
Authors: Federico Del Pup, Andrea Zanola, Louis Fabrice Tshimanga, Alessandra Bertoldo, Livio Finos, Manfredo Atzori
Abstract: Deep learning is significantly advancing the analysis of electroencephalography (EEG) data by effectively discovering highly nonlinear patterns within the signals. Data partitioning and cross-validation are crucial for assessing model performance and ensuring study comparability, as they can produce varied results and data leakage due to specific signal properties (e.g., biometric). Such variability leads to incomparable studies and, increasingly, overestimated performance claims, which are detrimental to the field. Nevertheless, no comprehensive guidelines for proper data partitioning and cross-validation exist in the domain, nor is there a quantitative evaluation of their impact on model accuracy, reliability, and generalizability. To assist researchers in identifying optimal experimental strategies, this paper thoroughly investigates the role of data partitioning and cross-validation in evaluating EEG deep learning models. Five cross-validation settings are compared across three supervised cross-subject classification tasks (BCI, Parkinson's, and Alzheimer's disease detection) and four established architectures of increasing complexity (ShallowConvNet, EEGNet, DeepConvNet, and Temporal-based ResNet). The comparison of over 100,000 trained models underscores, first, the importance of using subject-based cross-validation strategies for evaluating EEG deep learning models, except when within-subject analyses are acceptable (e.g., BCI). Second, it highlights the greater reliability of nested approaches (N-LNSO) compared to non-nested counterparts, which are prone to data leakage and favor larger models overfitting to validation data. In conclusion, this work provides EEG deep learning researchers with an analysis of data partitioning and cross-validation and offers guidelines to avoid data leakage, currently undermining the domain with potentially overestimated performance claims.
Authors: Tuan Thai, TrungTin Nguyen, Dat Do, Nhat Ho, Christopher Drovandi
Abstract: Mixture of Experts (MoE) models constitute a widely utilized class of ensemble learning approaches in statistics and machine learning, known for their flexibility and computational efficiency. They have become integral components in numerous state-of-the-art deep neural network architectures, particularly for analyzing heterogeneous data across diverse domains. Despite their practical success, the theoretical understanding of model selection, especially concerning the optimal number of mixture components or experts, remains limited and poses significant challenges. These challenges primarily stem from the inclusion of covariates in both the Gaussian gating functions and expert networks, which introduces intrinsic interactions governed by partial differential equations with respect to their parameters. In this paper, we revisit the concept of dendrograms of mixing measures and introduce a novel extension to Gaussian-gated Gaussian MoE models that enables consistent estimation of the true number of mixture components and achieves the pointwise optimal convergence rate for parameter estimation in overfitted scenarios. Notably, this approach circumvents the need to train and compare a range of models with varying numbers of components, thereby alleviating the computational burden, particularly in high-dimensional or deep neural network settings. Experimental results on synthetic data demonstrate the effectiveness of the proposed method in accurately recovering the number of experts. It outperforms common criteria such as the Akaike information criterion, the Bayesian information criterion, and the integrated completed likelihood, while achieving optimal convergence rates for parameter estimation and accurately approximating the regression function.
Authors: Jonathan Ott, Maximilian Stahlke, Tobias Feigl, Bjoern M. Eskofier, Christopher Mutschler
Abstract: We introduce the Sparse pretrained Radio Transformer (SpaRTran), an unsupervised representation learning approach based on the concept of compressed sensing for radio channels. Our approach learns embeddings that focus on the physical properties of radio propagation, to create the optimal basis for fine-tuning on radio-based downstream tasks. SpaRTran uses a sparse gated autoencoder that induces a simplicity bias to the learned representations, resembling the sparse nature of radio propagation. For signal reconstruction, it learns a dictionary that holds atomic features, which increases flexibility across signal waveforms and spatiotemporal signal patterns. Our experiments show that SpaRTran reduces errors by up to 85 % compared to state-of-the-art methods when fine-tuned on radio fingerprinting, a challenging downstream task. In addition, our method requires less pretraining effort and offers greater flexibility, as we train it solely on individual radio signals. SpaRTran serves as an excellent base model that can be fine-tuned for various radio-based downstream tasks, effectively reducing the cost for labeling. In addition, it is significantly more versatile than existing methods and demonstrates superior generalization.
Authors: Xiang Fei, Jinghui Lu, Qi Sun, Hao Feng, Yanjie Wang, Wei Shi, An-Lan Wang, Jingqun Tang, Can Huang
Abstract: Autoregressive models have become the de facto choice for sequence generation tasks, but standard approaches treat digits as independent tokens and apply cross-entropy loss, overlooking the coherent structure of numerical sequences. This paper introduces Numerical Token Integrity Loss (NTIL) to address this gap. NTIL operates at two levels: (1) token-level, where it extends the Earth Mover's Distance (EMD) to preserve ordinal relationships between numerical values, and (2) sequence-level, where it penalizes the overall discrepancy between the predicted and actual sequences. This dual approach improves numerical prediction and integrates effectively with LLMs/MLLMs. Extensive experiments show significant performance improvements with NTIL.
Authors: Biel Tura Vecino, Subhadeep Maji, Aravind Varier, Antonio Bonafonte, Ivan Valles, Michael Owen, Leif Radel, Grant Strimmel, Seyi Feyisetan, Roberto Barra Chicote, Ariya Rastrow, Constantinos Papayiannis, Volker Leutnant, Trevor Wood
Abstract: The use of audio recordings of human speech to train LLMs poses privacy concerns due to these models' potential to generate outputs that closely resemble artifacts in the training data. In this study, we propose a speaker privacy-preserving representation learning method through the Universal Speech Codec (USC), a computationally efficient encoder-decoder model that disentangles speech into: $\textit{(i)}$ privacy-preserving semantically rich representations, capturing content and speech paralinguistics, and $\textit{(ii)}$ residual acoustic and speaker representations that enables high-fidelity reconstruction. Extensive evaluations presented show that USC's semantic representation preserves content, prosody, and sentiment, while removing potentially identifiable speaker attributes. Combining both representations, USC achieves state-of-the-art speech reconstruction. Additionally, we introduce an evaluation methodology for measuring privacy-preserving properties, aligning with perceptual tests. We compare USC against other codecs in the literature and demonstrate its effectiveness on privacy-preserving representation learning, illustrating the trade-offs of speaker anonymization, paralinguistics retention and content preservation in the learned semantic representations. Audio samples are shared in $\href{https://www.amazon.science/usc-samples}{https://www.amazon.science/usc-samples}$.
URLs: https://www.amazon.science/usc-samples, https://www.amazon.science/usc-samples
Authors: Zhaoyi Wang, Shengyu Huang, Jemil Avers Butt, Yuanzhou Cai, Matej Varga, Andreas Wieser
Abstract: Point cloud registration has seen significant advancements with the application of deep learning techniques. However, existing approaches often overlook the potential of integrating radiometric information from RGB images. This limitation reduces their effectiveness in aligning point clouds pairs, especially in regions where geometric data alone is insufficient. When used effectively, radiometric information can enhance the registration process by providing context that is missing from purely geometric data. In this paper, we propose CoFF, a novel Cross-modal Feature Fusion method that utilizes both point cloud geometry and RGB images for pairwise point cloud registration. Assuming that the co-registration between point clouds and RGB images is available, CoFF explicitly addresses the challenges where geometric information alone is unclear, such as in regions with symmetric similarity or planar structures, through a two-stage fusion of 3D point cloud features and 2D image features. It incorporates a cross-modal feature fusion module that assigns pixel-wise image features to 3D input point clouds to enhance learned 3D point features, and integrates patch-wise image features with superpoint features to improve the quality of coarse matching. This is followed by a coarse-to-fine matching module that accurately establishes correspondences using the fused features. We extensively evaluate CoFF on four common datasets: 3DMatch, 3DLoMatch, IndoorLRS, and the recently released ScanNet++ datasets. In addition, we assess CoFF on specific subset datasets containing geometrically ambiguous cases. Our experimental results demonstrate that CoFF achieves state-of-the-art registration performance across all benchmarks, including remarkable registration recalls of 95.9% and 81.6% on the widely-used 3DMatch and 3DLoMatch datasets, respectively...(Truncated to fit arXiv abstract length)
Authors: Rodrigo Maulen-Soto (SU), Claire Boyer (EPFL), Pierre Marion (EPFL)
Abstract: Transformers have emerged as a powerful neural network architecture capable of tackling a wide range of learning tasks. In this work, we provide a theoretical analysis of their ability to automatically extract structure from data in an unsupervised setting. In particular, we demonstrate their suitability for clustering when the input data is generated from a Gaussian mixture model. To this end, we study a simplified two-head attention layer and define a population risk whose minimization with unlabeled data drives the head parameters to align with the true mixture centroids.
Authors: Debarpan Bhattacharya, Apoorva Kulkarni, Sriram Ganapathy
Abstract: The popular success of text-based large language models (LLM) has streamlined the attention of the multimodal community to combine other modalities like vision and audio along with text to achieve similar multimodal capabilities. In this quest, large audio language models (LALMs) have to be evaluated on reasoning related tasks which are different from traditional classification or generation tasks. Towards this goal, we propose a novel dataset called temporal reasoning evaluation of audio (TREA). We benchmark open-source LALMs and observe that they are consistently behind human capabilities on the tasks in the TREA dataset. While evaluating LALMs, we also propose an uncertainty metric, which computes the invariance of the model to semantically identical perturbations of the input. Our analysis shows that the accuracy and uncertainty metrics are not necessarily correlated and thus, points to a need for wholesome evaluation of LALMs for high-stakes applications.
Authors: Marouane Il Idrissi (UQAM, IID), Agathe Fernandes Machado (UQAM), Ewen Gallic (AMSE), Arthur Charpentier (UQAM)
Abstract: Cooperative game theory methods, notably Shapley values, have significantly enhanced machine learning (ML) interpretability. However, existing explainable AI (XAI) frameworks mainly attribute average model predictions, overlooking predictive uncertainty. This work addresses that gap by proposing a novel, model-agnostic uncertainty attribution (UA) method grounded in conformal prediction (CP). By defining cooperative games where CP interval properties-such as width and bounds-serve as value functions, we systematically attribute predictive uncertainty to input features. Extending beyond the traditional Shapley values, we use the richer class of Harsanyi allocations, and in particular the proportional Shapley values, which distribute attribution proportionally to feature importance. We propose a Monte Carlo approximation method with robust statistical guarantees to address computational feasibility, significantly improving runtime efficiency. Our comprehensive experiments on synthetic benchmarks and real-world datasets demonstrate the practical utility and interpretative depth of our approach. By combining cooperative game theory and conformal prediction, we offer a rigorous, flexible toolkit for understanding and communicating predictive uncertainty in high-stakes ML applications.
Authors: Snehashis Majhi, Giacomo D'Amicantonio, Antitza Dantcheva, Quan Kong, Lorenzo Garattoni, Gianpiero Francesca, Egor Bondarev, Francois Bremond
Abstract: Weakly-supervised methods for video anomaly detection (VAD) are conventionally based merely on RGB spatio-temporal features, which continues to limit their reliability in real-world scenarios. This is due to the fact that RGB-features are not sufficiently distinctive in setting apart categories such as shoplifting from visually similar events. Therefore, towards robust complex real-world VAD, it is essential to augment RGB spatio-temporal features by additional modalities. Motivated by this, we introduce the Poly-modal Induced framework for VAD: "PI-VAD", a novel approach that augments RGB representations by five additional modalities. Specifically, the modalities include sensitivity to fine-grained motion (Pose), three dimensional scene and entity representation (Depth), surrounding objects (Panoptic masks), global motion (optical flow), as well as language cues (VLM). Each modality represents an axis of a polygon, streamlined to add salient cues to RGB. PI-VAD includes two plug-in modules, namely Pseudo-modality Generation module and Cross Modal Induction module, which generate modality-specific prototypical representation and, thereby, induce multi-modal information into RGB cues. These modules operate by performing anomaly-aware auxiliary tasks and necessitate five modality backbones -- only during training. Notably, PI-VAD achieves state-of-the-art accuracy on three prominent VAD datasets encompassing real-world scenarios, without requiring the computational overhead of five modality backbones at inference.
Authors: Anton Ehrmanntraut, Julia Wunderle, Jan Pfister, Fotis Jannidis, Andreas Hotho
Abstract: Despite the prominence of decoder-only language models, encoders remain crucial for resource-constrained applications. We introduce ModernGBERT (134M, 1B), a fully transparent family of German encoder models trained from scratch, incorporating architectural innovations from ModernBERT. To evaluate the practical trade-offs of training encoders from scratch, we also present LL\"aMmlein2Vec (120M, 1B, 7B), a family of encoders derived from German decoder-only models via LLM2Vec. We benchmark all models on natural language understanding, text embedding, and long-context reasoning tasks, enabling a controlled comparison between dedicated encoders and converted decoders. Our results show that ModernGBERT 1B outperforms prior state-of-the-art German encoders as well as encoders adapted via LLM2Vec, with regard to performance and parameter-efficiency. All models, training data, checkpoints and code are publicly available, advancing the German NLP ecosystem with transparent, high-performance encoder models.
Authors: Philipp Scholl, Alexander Dietrich, Sebastian Wolf, Jinoh Lee, Alin-Albu Sch\"affer, Gitta Kutyniok, Maged Iskandar
Abstract: Accurately modeling the friction torque in robotic joints has long been challenging due to the request for a robust mathematical description. Traditional model-based approaches are often labor-intensive, requiring extensive experiments and expert knowledge, and they are difficult to adapt to new scenarios and dependencies. On the other hand, data-driven methods based on neural networks are easier to implement but often lack robustness, interpretability, and trustworthiness--key considerations for robotic hardware and safety-critical applications such as human-robot interaction. To address the limitations of both approaches, we propose the use of symbolic regression (SR) to estimate the friction torque. SR generates interpretable symbolic formulas similar to those produced by model-based methods while being flexible to accommodate various dynamic effects and dependencies. In this work, we apply SR algorithms to approximate the friction torque using collected data from a KUKA LWR-IV+ robot. Our results show that SR not only yields formulas with comparable complexity to model-based approaches but also achieves higher accuracy. Moreover, SR-derived formulas can be seamlessly extended to include load dependencies and other dynamic factors.
Authors: Giacomo Greco
Abstract: We adopt a Gamma and Malliavin Calculi point of view in order to generalize Score-based diffusion Generative Models (SGMs) to an infinite-dimensional abstract Hilbertian setting. Particularly, we define the forward noising process using Dirichlet forms associated to the Cameron-Martin space of Gaussian measures and Wiener chaoses; whereas by relying on an abstract time-reversal formula, we show that the score function is a Malliavin derivative and it corresponds to a conditional expectation. This allows us to generalize SGMs to the infinite-dimensional setting. Moreover, we extend existing finite-dimensional entropic convergence bounds to this Hilbertian setting by highlighting the role played by the Cameron-Martin norm in the Fisher information of the data distribution. Lastly, we specify our discussion for spherical random fields, considering as source of noise a Whittle-Mat\'ern random spherical field.
Authors: Yanfeng Yang, Kenji Fukumizu
Abstract: Creating large-scale datasets for training high-performance generative models is often prohibitively expensive, especially when associated attributes or annotations must be provided. As a result, merging existing datasets has become a common strategy. However, the sets of attributes across datasets are often inconsistent, and their naive concatenation typically leads to block-wise missing conditions. This presents a significant challenge for conditional generative modeling when the multiple attributes are used jointly as conditions, thereby limiting the model's controllability and applicability. To address this issue, we propose a novel generative approach, Diffusion Model with Double Guidance, which enables precise conditional generation even when no training samples contain all conditions simultaneously. Our method maintains rigorous control over multiple conditions without requiring joint annotations. We demonstrate its effectiveness in molecular and image generation tasks, where it outperforms existing baselines both in alignment with target conditional distributions and in controllability under missing condition settings.
Authors: Junyi Chen, Alap Kshirsagar, Frederik Heller, Mario G\'omez Andreu, Boris Belousov, Tim Schneider, Lisa P. Y. Lin, Katja Doerschner, Knut Drewing, Jan Peters
Abstract: One of the most important object properties that humans and robots perceive through touch is hardness. This paper investigates information-theoretic active sampling strategies for sample-efficient hardness classification with vision-based tactile sensors. We evaluate three probabilistic classifier models and two model-uncertainty-based sampling strategies on a robotic setup as well as on a previously published dataset of samples collected by human testers. Our findings indicate that the active sampling approaches, driven by uncertainty metrics, surpass a random sampling baseline in terms of accuracy and stability. Additionally, while in our human study, the participants achieve an average accuracy of 48.00%, our best approach achieves an average accuracy of 88.78% on the same set of objects, demonstrating the effectiveness of vision-based tactile sensors for object hardness classification.
Authors: Dang Hoai Nam, Huynh Tong Dang Khoa, Vo Nguyen Le Duy
Abstract: Humans can quickly generalize handwriting styles from a single example by intuitively separating content from style. Machines, however, struggle with this task, especially in low-data settings, often missing subtle spatial and stylistic cues. Motivated by this gap, we introduce WriteViT, a one-shot handwritten text synthesis framework that incorporates Vision Transformers (ViT), a family of models that have shown strong performance across various computer vision tasks. WriteViT integrates a ViT-based Writer Identifier for extracting style embeddings, a multi-scale generator built with Transformer encoder-decoder blocks enhanced by conditional positional encoding (CPE), and a lightweight ViT-based recognizer. While previous methods typically rely on CNNs or CRNNs, our design leverages transformers in key components to better capture both fine-grained stroke details and higher-level style information. Although handwritten text synthesis has been widely explored, its application to Vietnamese -- a language rich in diacritics and complex typography -- remains limited. Experiments on Vietnamese and English datasets demonstrate that WriteViT produces high-quality, style-consistent handwriting while maintaining strong recognition performance in low-resource scenarios. These results highlight the promise of transformer-based designs for multilingual handwriting generation and efficient style adaptation.
Authors: Wenbin Zhou, Agni Orfanoudaki, Shixiang Zhu
Abstract: High-stakes decisions in domains such as healthcare, energy, and public policy are often made by human experts using domain knowledge and heuristics, yet are increasingly supported by predictive and optimization-based tools. A dominant approach in operations research is the predict-then-optimize paradigm, where a predictive model estimates uncertain inputs, and an optimization model recommends a decision. However, this approach often lacks interpretability and can fail under distributional uncertainty -- particularly when the outcome distribution is multi-modal or complex -- leading to brittle or misleading decisions. In this paper, we introduce CREDO, a novel framework that quantifies, for any candidate decision, a distribution-free upper bound on the probability that the decision is suboptimal. By combining inverse optimization geometry with conformal prediction and generative modeling, CREDO produces risk certificates that are both statistically rigorous and practically interpretable. This framework enables human decision-makers to audit and validate their own decisions under uncertainty, bridging the gap between algorithmic tools and real-world judgment.
Authors: Jieying Xue, Phuong Minh Nguyen, Minh Le Nguyen, Xin Liu
Abstract: With the rapid advancement of global digitalization, users from different countries increasingly rely on social media for information exchange. In this context, multilingual multi-label emotion detection has emerged as a critical research area. This study addresses SemEval-2025 Task 11: Bridging the Gap in Text-Based Emotion Detection. Our paper focuses on two sub-tracks of this task: (1) Track A: Multi-label emotion detection, and (2) Track B: Emotion intensity. To tackle multilingual challenges, we leverage pre-trained multilingual models and focus on two architectures: (1) a fine-tuned BERT-based classification model and (2) an instruction-tuned generative LLM. Additionally, we propose two methods for handling multi-label classification: the base method, which maps an input directly to all its corresponding emotion labels, and the pairwise method, which models the relationship between the input text and each emotion category individually. Experimental results demonstrate the strong generalization ability of our approach in multilingual emotion recognition. In Track A, our method achieved Top 4 performance across 10 languages, ranking 1st in Hindi. In Track B, our approach also secured Top 5 performance in 7 languages, highlighting its simplicity and effectiveness\footnote{Our code is available at https://github.com/yingjie7/mlingual_multilabel_emo_detection.
URLs: https://github.com/yingjie7/mlingual_multilabel_emo_detection.
Authors: Zilu Tang, Afra Feyza Aky\"urek, Ekin Aky\"urek, Derry Wijaya
Abstract: Preference alignment has become a standard pipeline in finetuning models to follow \emph{generic} human preferences. Majority of work seeks to optimize model to produce responses that would be preferable \emph{on average}, simplifying the diverse and often \emph{contradicting} space of human preferences. While research has increasingly focused on personalized alignment: adapting models to individual user preferences, there is a lack of personalized preference dataset which focus on nuanced individual-level preferences. To address this, we introduce WikiPersona: the first fine-grained personalization using well-documented, famous individuals. Our dataset challenges models to align with these personas through an interpretable process: generating verifiable textual descriptions of a persona's background and preferences in addition to alignment. We systematically evaluate different personalization approaches and find that as few-shot prompting with preferences and fine-tuning fail to simultaneously ensure effectiveness and efficiency, using \textit{inferred personal preferences} as prefixes enables effective personalization, especially in topics where preferences clash while leading to more equitable generalization across unseen personas.
Authors: Livia Qian, Carol Figueroa, Gabriel Skantze
Abstract: Vocal feedback (e.g., `mhm', `yeah', `okay') is an important component of spoken dialogue and is crucial to ensuring common ground in conversational systems. The exact meaning of such feedback is conveyed through both lexical and prosodic form. In this work, we investigate the perceived prosodic similarity of vocal feedback with the same lexical form, and to what extent existing speech representations reflect such similarities. A triadic comparison task with recruited participants is used to measure perceived similarity of feedback responses taken from two different datasets. We find that spectral and self-supervised speech representations encode prosody better than extracted pitch features, especially in the case of feedback from the same speaker. We also find that it is possible to further condense and align the representations to human perception through contrastive learning.
Authors: Lucas Berry, Axel Brando, Wei-Di Chang, Juan Camilo Gamboa Higuera, David Meger
Abstract: Estimating uncertainty in text-to-image diffusion models is challenging because of their large parameter counts (often exceeding 100 million) and operation in complex, high-dimensional spaces with virtually infinite input possibilities. In this paper, we propose Epistemic Mixture of Experts (EMoE), a novel framework for efficiently estimating epistemic uncertainty in diffusion models. EMoE leverages pre-trained networks without requiring additional training, enabling direct uncertainty estimation from a prompt. We leverage a latent space within the diffusion process that captures epistemic uncertainty better than existing methods. Experimental results on the COCO dataset demonstrate EMoE's effectiveness, showing a strong correlation between uncertainty and image quality. Additionally, EMoE identifies under-sampled languages and regions with higher uncertainty, revealing hidden biases in the training set. This capability demonstrates the relevance of EMoE as a tool for addressing fairness and accountability in AI-generated content.
Authors: Likai Chen, Georg Keilbar, Wei Biao Wu
Abstract: This paper considers the estimation of quantiles via a smoothed version of the stochastic gradient descent (SGD) algorithm. By smoothing the score function in the conventional SGD quantile algorithm, we achieve monotonicity in the quantile level in that the estimated quantile curves do not cross. We derive non-asymptotic tail probability bounds for the smoothed SGD quantile estimate both for the case with and without Polyak-Ruppert averaging. For the latter, we also provide a uniform Bahadur representation and a resulting Gaussian approximation result. Numerical studies show good finite sample behavior for our theoretical results.
Authors: Gabriele Spadaro, Alberto Presta, Jhony H. Giraldo, Marco Grangetto, Wei Hu, Giuseppe Valenzise, Attilio Fiandrotti, Enzo Tartaglione
Abstract: Efficient compression of low-bit-rate point clouds is critical for bandwidth-constrained applications. However, existing techniques mainly focus on high-fidelity reconstruction, requiring many bits for compression. This paper proposes a "Denoising Diffusion Probabilistic Model" (DDPM) architecture for point cloud compression (DDPM-PCC) at low bit-rates. A PointNet encoder produces the condition vector for the generation, which is then quantized via a learnable vector quantizer. This configuration allows to achieve a low bitrates while preserving quality. Experiments on ShapeNet and ModelNet40 show improved rate-distortion at low rates compared to standardized and state-of-the-art approaches. We publicly released the code at https://github.com/EIDOSLAB/DDPM-PCC.
Authors: Paula Feldman, Martin Sinnona, Viviana Siless, Claudio Delrieux, Emmanuel Iarussi
Abstract: Anatomical trees are critical for clinical diagnosis and treatment planning, yet their complex and diverse geometry make accurate representation a significant challenge. Motivated by the latest advances in large language models, we introduce an autoregressive method for synthesizing anatomical trees. Our approach first embeds vessel structures into a learned discrete vocabulary using a VQ-VAE architecture, then models their generation autoregressively with a GPT-2 model. This method effectively captures intricate geometries and branching patterns, enabling realistic vascular tree synthesis. Comprehensive qualitative and quantitative evaluations reveal that our technique achieves high-fidelity tree reconstruction with compact discrete representations. Moreover, our B-spline representation of vessel cross-sections preserves critical morphological details that are often overlooked in previous' methods parameterizations. To the best of our knowledge, this work is the first to generate blood vessels in an autoregressive manner. Code, data, and trained models will be made available.
Authors: Galit Shmueli, David Martens, Jaewon Yoo, Travis Greene
Abstract: Counterfactuals play a pivotal role in the two distinct data science fields of causal inference (CI) and explainable artificial intelligence (XAI). While the core idea behind counterfactuals remains the same in both fields--the examination of what would have happened under different circumstances--there are key differences in how they are used and interpreted. We introduce a formal definition that encompasses the multi-faceted concept of the counterfactual in CI and XAI. We then discuss how counterfactuals are used, evaluated, generated, and operationalized in CI vs. XAI, highlighting conceptual and practical differences. By comparing and contrasting the two, we hope to identify opportunities for cross-fertilization across CI and XAI.
Authors: Ho-Chun Herbert Chang
Abstract: Measuring social influence is difficult due to the lack of counter-factuals and comparisons. By combining machine learning-based modeling and network science, we present general properties of social value, a recent measure for social influence using synthetic control applicable to political behavior. Social value diverges from centrality measures on in that it relies on an external regressor to predict an output variable of interest, generates a synthetic measure of influence, then distributes individual contribution based on a social network. Through theoretical derivations, we show the properties of SV under linear regression with and without interaction, across lattice networks, power-law networks, and random graphs. A reduction in computation can be achieved for any ensemble model. Through simulation, we find that the generalized friendship paradox holds -- that in certain situations, your friends have on average more influence than you do.
Authors: Ahmet Berke Gokmen, Yigit Ekin, Bahri Batuhan Bilecen, Aysegul Dundar
Abstract: We propose RoPECraft, a training-free video motion transfer method for diffusion transformers that operates solely by modifying their rotary positional embeddings (RoPE). We first extract dense optical flow from a reference video, and utilize the resulting motion offsets to warp the complex-exponential tensors of RoPE, effectively encoding motion into the generation process. These embeddings are then further optimized during denoising time steps via trajectory alignment between the predicted and target velocities using a flow-matching objective. To keep the output faithful to the text prompt and prevent duplicate generations, we incorporate a regularization term based on the phase components of the reference video's Fourier transform, projecting the phase angles onto a smooth manifold to suppress high-frequency artifacts. Experiments on benchmarks reveal that RoPECraft outperforms all recently published methods, both qualitatively and quantitatively.
Authors: Adam \v{S}torek, Mukur Gupta, Samira Hajizadeh, Prashast Srivastava, Suman Jana
Abstract: Although modern Large Language Models (LLMs) support extremely large contexts, their effectiveness in utilizing long context for code reasoning remains unclear. This paper investigates LLM reasoning ability over code snippets within large repositories and how it relates to their recall ability. Specifically, we differentiate between lexical code recall (verbatim retrieval) and semantic code recall (remembering what the code does). To measure semantic recall, we propose SemTrace, a code reasoning technique where the impact of specific statements on output is attributable and unpredictable. We also present a method to quantify semantic recall sensitivity in existing benchmarks. Our evaluation of state-of-the-art LLMs reveals a significant drop in code reasoning accuracy as a code snippet approaches the middle of the input context, particularly with techniques requiring high semantic recall like SemTrace. Moreover, we find that lexical recall varies by granularity, with models excelling at function retrieval but struggling with line-by-line recall. Notably, a disconnect exists between lexical and semantic recall, suggesting different underlying mechanisms. Finally, our findings indicate that current code reasoning benchmarks may exhibit low semantic recall sensitivity, potentially underestimating LLM challenges in leveraging in-context information.
Authors: Rebecca Pelke, Nils Bosbach, Lennart M. Reimann, Rainer Leupers
Abstract: Accelerating Machine Learning (ML) workloads requires efficient methods due to their large optimization space. Autotuning has emerged as an effective approach for systematically evaluating variations of implementations. Traditionally, autotuning requires the workloads to be executed on the target hardware (HW). We present an interface that allows executing autotuning workloads on simulators. This approach offers high scalability when the availability of the target HW is limited, as many simulations can be run in parallel on any accessible HW. Additionally, we evaluate the feasibility of using fast instruction-accurate simulators for autotuning. We train various predictors to forecast the performance of ML workload implementations on the target HW based on simulation statistics. Our results demonstrate that the tuned predictors are highly effective. The best workload implementation in terms of actual run time on the target HW is always within the top 3 % of predictions for the tested x86, ARM, and RISC-V-based architectures. In the best case, this approach outperforms native execution on the target HW for embedded architectures when running as few as three samples on three simulators in parallel.
Authors: Christopher Kolloff, Tobias H\"oppe, Emmanouil Angelis, Mathias Jacob Schreiner, Stefan Bauer, Andrea Dittadi, Simon Olsson
Abstract: We propose a regularization framework inspired by thermodynamic work for guiding pre-trained probability flow generative models (e.g., continuous normalizing flows or diffusion models) by minimizing excess work, a concept rooted in statistical mechanics and with strong conceptual connections to optimal transport. Our approach enables efficient guidance in sparse-data regimes common to scientific applications, where only limited target samples or partial density constraints are available. We introduce two strategies: Path Guidance for sampling rare transition states by concentrating probability mass on user-defined subsets, and Observable Guidance for aligning generated distributions with experimental observables while preserving entropy. We demonstrate the framework's versatility on a coarse-grained protein model, guiding it to sample transition configurations between folded/unfolded states and correct systematic biases using experimental data. The method bridges thermodynamic principles with modern generative architectures, offering a principled, efficient, and physics-inspired alternative to standard fine-tuning in data-scarce domains. Empirical results highlight improved sample efficiency and bias reduction, underscoring its applicability to molecular simulations and beyond.
Authors: David Anugraha, Zilu Tang, Lester James V. Miranda, Hanyang Zhao, Mohammad Rifqi Farhansyah, Garry Kuwanto, Derry Wijaya, Genta Indra Winata
Abstract: Reward models are essential for aligning language model outputs with human preferences, yet existing approaches often lack both controllability and interpretability. These models are typically optimized for narrow objectives, limiting their generalizability to broader downstream tasks. Moreover, their scalar outputs are difficult to interpret without contextual reasoning. To address these limitations, we introduce R3, a novel reward modeling framework that is rubric-agnostic, generalizable across evaluation dimensions, and provides interpretable, reasoned score assignments. R3 enables more transparent and flexible evaluation of language models, supporting robust alignment with diverse human values and use cases. Our models, data, and code are available as open source at https://github.com/rubricreward/r3
Authors: Miko{\l}aj Ma{\l}ki\'nski, Jacek Ma\'ndziuk
Abstract: The abstract visual reasoning (AVR) domain presents a diverse suite of analogy-based tasks devoted to studying model generalization. Recent years have brought dynamic progress in the field, particularly in i.i.d. scenarios, in which models are trained and evaluated on the same data distributions. Nevertheless, o.o.d. setups that assess model generalization to new test distributions remain challenging even for the most recent models. To advance generalization in AVR tasks, we present the Pathways of Normalized Group Convolution model (PoNG), a novel neural architecture that features group convolution, normalization, and a parallel design. We consider a wide set of AVR benchmarks, including Raven's Progressive Matrices and visual analogy problems with both synthetic and real-world images. The experiments demonstrate strong generalization capabilities of the proposed model, which in several settings outperforms the existing literature methods.
Authors: Jiajie Zhang, Nianyi Lin, Lei Hou, Ling Feng, Juanzi Li
Abstract: Recently, large reasoning models have achieved impressive performance on various tasks by employing human-like deep thinking. However, the lengthy thinking process substantially increases inference overhead, making efficiency a critical bottleneck. In this work, we first demonstrate that NoThinking, which prompts the reasoning model to skip thinking and directly generate the final solution, is a better choice for relatively simple tasks in terms of both performance and efficiency. Motivated by this, we propose AdaptThink, a novel RL algorithm to teach reasoning models to choose the optimal thinking mode adaptively based on problem difficulty. Specifically, AdaptThink features two core components: (1) a constrained optimization objective that encourages the model to choose NoThinking while maintaining the overall performance; (2) an importance sampling strategy that balances Thinking and NoThinking samples during on-policy training, thereby enabling cold start and allowing the model to explore and exploit both thinking modes throughout the training process. Our experiments indicate that AdaptThink significantly reduces the inference costs while further enhancing performance. Notably, on three math datasets, AdaptThink reduces the average response length of DeepSeek-R1-Distill-Qwen-1.5B by 53% and improves its accuracy by 2.4%, highlighting the promise of adaptive thinking-mode selection for optimizing the balance between reasoning quality and efficiency. Our codes and models are available at https://github.com/THU-KEG/AdaptThink.
Authors: Lotem Peled-Cohen, Maya Zadok, Nitay Calderon, Hila Gonen, Roi Reichart
Abstract: Cognitive decline often surfaces in language years before diagnosis. It is frequently non-experts, such as those closest to the patient, who first sense a change and raise concern. As LLMs become integrated into daily communication and used over prolonged periods, it may even be an LLM that notices something is off. But what exactly do they notice--and should be noticing--when making that judgment? This paper investigates how dementia is perceived through language by non-experts. We presented transcribed picture descriptions to non-expert humans and LLMs, asking them to intuitively judge whether each text was produced by someone healthy or with dementia. We introduce an explainable method that uses LLMs to extract high-level, expert-guided features representing these picture descriptions, and use logistic regression to model human and LLM perceptions and compare with clinical diagnoses. Our analysis reveals that human perception of dementia is inconsistent and relies on a narrow, and sometimes misleading, set of cues. LLMs, by contrast, draw on a richer, more nuanced feature set that aligns more closely with clinical patterns. Still, both groups show a tendency toward false negatives, frequently overlooking dementia cases. Through our interpretable framework and the insights it provides, we hope to help non-experts better recognize the linguistic signs that matter.
Authors: Connor Lennon, Edward Rubin, Glen Waddell
Abstract: Machine learning (ML) primarily evolved to solve "prediction problems." The first stage of two-stage least squares (2SLS) is a prediction problem, suggesting potential gains from ML first-stage assistance. However, little guidance exists on when ML helps 2SLS$\unicode{x2014}$or when it hurts. We investigate the implications of inserting ML into 2SLS, decomposing the bias into three informative components. Mechanically, ML-in-2SLS procedures face issues common to prediction and causal-inference settings$\unicode{x2014}$and their interaction. Through simulation, we show linear ML methods (e.g., post-Lasso) work well, while nonlinear methods (e.g., random forests, neural nets) generate substantial bias in second-stage estimates$\unicode{x2014}$potentially exceeding the bias of endogenous OLS.
Authors: Huawei Lin, Tong Geng, Zhaozhuo Xu, Weijie Zhao
Abstract: Autoregressive (AR) models have recently shown strong performance in image generation, where a critical component is the visual tokenizer (VT) that maps continuous pixel inputs to discrete token sequences. The quality of the VT largely defines the upper bound of AR model performance. However, current discrete VTs fall significantly behind continuous variational autoencoders (VAEs), leading to degraded image reconstructions and poor preservation of details and text. Existing benchmarks focus on end-to-end generation quality, without isolating VT performance. To address this gap, we introduce VTBench, a comprehensive benchmark that systematically evaluates VTs across three core tasks: Image Reconstruction, Detail Preservation, and Text Preservation, and covers a diverse range of evaluation scenarios. We systematically assess state-of-the-art VTs using a set of metrics to evaluate the quality of reconstructed images. Our findings reveal that continuous VAEs produce superior visual representations compared to discrete VTs, particularly in retaining spatial structure and semantic detail. In contrast, the degraded representations produced by discrete VTs often lead to distorted reconstructions, loss of fine-grained textures, and failures in preserving text and object integrity. Furthermore, we conduct experiments on GPT-4o image generation and discuss its potential AR nature, offering new insights into the role of visual tokenization. We release our benchmark and codebase publicly to support further research and call on the community to develop strong, general-purpose open-source VTs.
Authors: Vladimir Vovk
Abstract: This paper discusses a counterpart of conformal prediction for e-values, conformal e-prediction. Conformal e-prediction is conceptually simpler and had been developed in the 1990s as a precursor of conformal prediction. When conformal prediction emerged as result of replacing e-values by p-values, it seemed to have important advantages over conformal e-prediction without obvious disadvantages. This paper re-examines relations between conformal prediction and conformal e-prediction systematically from a modern perspective. Conformal e-prediction has advantages of its own, such as the ease of designing conditional conformal e-predictors and the guaranteed validity of cross-conformal e-predictors (whereas for cross-conformal predictors validity is only an empirical fact and can be broken with excessive randomization). Even where conformal prediction has clear advantages, conformal e-prediction can often emulate those advantages, more or less successfully.
Authors: Krikamol Muandet
Abstract: Modern applications of AI involve training and deploying machine learning models across heterogeneous and potentially massive environments. Emerging diversity of data not only brings about new possibilities to advance AI systems, but also restricts the extent to which information can be shared across environments due to pressing concerns such as privacy, security, and equity. Based on a novel characterization of learning algorithms as choice correspondences on a hypothesis space, this work provides a minimum requirement in terms of intuitive and reasonable axioms under which the only rational learning algorithm in heterogeneous environments is an empirical risk minimization (ERM) that unilaterally learns from a single environment without information sharing across environments. Our (im)possibility result underscores the fundamental trade-off that any algorithms will face in order to achieve Collective Intelligence (CI), i.e., the ability to learn across heterogeneous environments. Ultimately, collective learning in heterogeneous environments are inherently hard because, in critical areas of machine learning such as out-of-distribution generalization, federated/collaborative learning, algorithmic fairness, and multi-modal learning, it can be infeasible to make meaningful comparisons of model predictive performance across environments.
Authors: Francesco Ardizzon, Stefano Tomasin
Abstract: In physical layer authentication (PLA) mechanisms, a verifier decides whether a received message has been transmitted by a legitimate user or an intruder, according to some features of the physical channel over which the message traveled. To design the authentication check implemented at the verifier, typically either the statistics or a dataset of features are available for the channel from the legitimate user, while no information is available when under attack. When the statistics are known, a well-known good solution is the likelihood test (LT). When a dataset is available, the decision problem is one-class classification (OCC) and a good understanding of the machine learning (ML) techniques used for its solution is important to ensure security. Thus, in this paper, we aim at obtaining ML PLA verifiers that operate as the LT. We show how to do it with the neural network (NN) and the one-class least-squares support vector machine (OCLSSVM) models, trained as two-class classifiers on the single-class dataset and an artificial dataset. The artificial dataset for the negative class is obtained by generating channel feature (CF) vectors uniformly distributed over the domain of the legitimate class dataset. We also derive a modified stochastic gradient descent (SGD) algorithm that trains a PLA verifier operating as LT without the need for the artificial dataset. Furthermore, we show that the one-class least-squares support vector machine with suitable kernels operates as the LT at convergence. Lastly, we show that the widely used autoencoder classifier generally does not provide the LT. Numerical results are provided considering PLA on both wireless and underwater acoustic channels.
Authors: Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongxuan Li, Jun Zhu
Abstract: Diffusion probabilistic models (DPMs) have achieved impressive success in high-resolution image synthesis, especially in recent large-scale text-to-image generation applications. An essential technique for improving the sample quality of DPMs is guided sampling, which usually needs a large guidance scale to obtain the best sample quality. The commonly-used fast sampler for guided sampling is DDIM, a first-order diffusion ODE solver that generally needs 100 to 250 steps for high-quality samples. Although recent works propose dedicated high-order solvers and achieve a further speedup for sampling without guidance, their effectiveness for guided sampling has not been well-tested before. In this work, we demonstrate that previous high-order fast samplers suffer from instability issues, and they even become slower than DDIM when the guidance scale grows large. To further speed up guided sampling, we propose DPM-Solver++, a high-order solver for the guided sampling of DPMs. DPM-Solver++ solves the diffusion ODE with the data prediction model and adopts thresholding methods to keep the solution matches training data distribution. We further propose a multistep variant of DPM-Solver++ to address the instability issue by reducing the effective step size. Experiments show that DPM-Solver++ can generate high-quality samples within only 15 to 20 steps for guided sampling by pixel-space and latent-space DPMs.
Authors: Yijun Bian, Kun Zhang
Abstract: The concern about hidden discrimination in ML models is growing, as their widespread real-world application increasingly impacts human lives. Various techniques, including commonly used group fairness measures and several fairness-aware ensemble-based methods, have been developed to enhance fairness. However, existing fairness measures typically focus on only one aspect -- either group or individual fairness, and the hard compatibility among them indicates a possibility of remaining biases even when one of them is satisfied. Moreover, existing mechanisms to boost fairness usually present empirical results to show validity, yet few of them discuss whether fairness can be boosted with certain theoretical guarantees. To address these issues, we propose a fairness quality measure named 'discriminative risk (DR)' to reflect both individual and group fairness aspects. Furthermore, we investigate its properties and establish the first- and second-order oracle bounds to show that fairness can be boosted via ensemble combination with theoretical learning guarantees. The analysis is suitable for both binary and multi-class classification. A pruning method is also proposed to utilise our proposed measure and comprehensive experiments are conducted to evaluate the effectiveness of the proposed methods.
Authors: Puning Yang, Jian Liang, Jie Cao, Ran He
Abstract: Out-of-distribution (OOD) detection aims to detect test samples that do not fall into any training in-distribution (ID) classes. Prior efforts focus on regularizing models with ID data only, largely underperforming counterparts that utilize auxiliary outliers. However, data safety and privacy make it infeasible to collect task-specific outliers in advance for different scenarios. Besides, using task-irrelevant outliers leads to inferior OOD detection performance. To address the above issue, we present a new setup called test-time OOD detection, which allows the deployed model to utilize real OOD data from the unlabeled data stream during testing. We propose Adaptive Outlier Optimization (AUTO) which allows for continuous adaptation of the OOD detector. Specifically, AUTO consists of three key components: 1) an in-out-aware filter to selectively annotate test samples with pseudo-ID and pseudo-OOD and ingeniously trigger the updating process while encountering each pseudo-OOD sample; 2) a dynamic-updated memory to overcome the catastrophic forgetting led by frequent parameter updates; 3) a prediction-aligning objective to calibrate the rough OOD objective during testing. Extensive experiments show that AUTO significantly improves OOD detection performance over state-of-the-art methods. Besides, evaluations on complicated scenarios (e.g. multi-OOD, time-series OOD) also conduct the superiority of AUTO.
Authors: Yunfei Luo, Zhangqi Duan
Abstract: Stock trading is one of the popular ways for financial management. However, the market and the environment of economy is unstable and usually not predictable. Furthermore, engaging in stock trading requires time and effort to analyze, create strategies, and make decisions. It would be convenient and effective if an agent could assist or even do the task of analyzing and modeling the past data and then generate a strategy for autonomous trading. Recently, reinforcement learning has been shown to be robust in various tasks that involve achieving a goal with a decision making strategy based on time-series data. In this project, we have developed a pipeline that simulates the stock trading environment and have trained an agent to automate the stock trading process with deep reinforcement learning methods, including deep Q-learning, deep SARSA, and the policy gradient method. We evaluate our platform during relatively good (before 2021) and bad (2021 - 2022) situations. The stocks we've evaluated on including Google, Apple, Tesla, Meta, Microsoft, and IBM. These stocks are among the popular ones, and the changes in trends are representative in terms of having good and bad situations. We showed that before 2021, the three reinforcement methods we have tried always provide promising profit returns with total annual rates around $70\%$ to $90\%$, while maintain a positive profit return after 2021 with total annual rates around 2% to 7%.
Authors: Yuwei Luo, Mohsen Bayati
Abstract: This paper is motivated by recent research in the $d$-dimensional stochastic linear bandit literature, which has revealed an unsettling discrepancy: algorithms like Thompson sampling and Greedy demonstrate promising empirical performance, yet this contrasts with their pessimistic theoretical regret bounds. The challenge arises from the fact that while these algorithms may perform poorly in certain problem instances, they generally excel in typical instances. To address this, we propose a new data-driven technique that tracks the geometric properties of the uncertainty ellipsoid around the main problem parameter. This methodology enables us to formulate a data-driven frequentist regret bound, which incorporates the geometric information, for a broad class of base algorithms, including Greedy, OFUL, and Thompson sampling. This result allows us to identify and ``course-correct" problem instances in which the base algorithms perform poorly. The course-corrected algorithms achieve the minimax optimal regret of order $\tilde{\mathcal{O}}(d\sqrt{T})$ for a $T$-period decision-making scenario, effectively maintaining the desirable attributes of the base algorithms, including their empirical efficacy. We present simulation results to validate our findings using synthetic and real data.
Authors: Eric Hsiung, Joydeep Biswas, Swarat Chaudhuri
Abstract: Active automata learning from membership and equivalence queries is a foundational problem with numerous applications. We propose a novel variant of the active automata learning problem: actively learn finite automata using preference queries -- i.e., queries about the relative position of two sequences in a total order -- instead of membership queries. Our solution is REMAP, a novel algorithm which leverages a symbolic observation table along with unification and constraint solving to navigate a space of symbolic hypotheses (each representing a set of automata), and uses satisfiability-solving to construct a concrete automaton from a symbolic hypothesis. REMAP is guaranteed to correctly infer the minimal automaton with polynomial query complexity under exact equivalence queries, and achieves PAC-identification ($\varepsilon$-approximate, with high probability) of the minimal automaton using sampling-based equivalence queries. Our empirical evaluations of REMAP on the task of learning reward machines for two reinforcement learning domains indicate REMAP scales to large automata and is effective at learning correct automata from consistent teachers, under both exact and sampling-based equivalence queries.
Authors: Mathieu Chevalley, Jacob Sackett-Sanders, Yusuf Roohani, Pascal Notin, Artemy Bakulin, Dariusz Brzezinski, Kaiwen Deng, Yuanfang Guan, Justin Hong, Michael Ibrahim, Wojciech Kotlowski, Marcin Kowiel, Panagiotis Misiakos, Achille Nazaret, Markus P\"uschel, Chris Wendler, Arash Mehrjou, Patrick Schwab
Abstract: In drug discovery, mapping interactions between genes within cellular systems is a crucial early step. Such maps are not only foundational for understanding the molecular mechanisms underlying disease biology but also pivotal for formulating hypotheses about potential targets for new medicines. Recognizing the need to elevate the construction of these gene-gene interaction networks, especially from large-scale, real-world datasets of perturbed single cells, the CausalBench Challenge was initiated. This challenge aimed to inspire the machine learning community to enhance state-of-the-art methods, emphasizing better utilization of expansive genetic perturbation data. Using the framework provided by the CausalBench benchmark, participants were tasked with refining the current methodologies or proposing new ones. This report provides an analysis and summary of the methods submitted during the challenge to give a partial image of the state of the art at the time of the challenge. Notably, the winning solutions significantly improved performance compared to previous baselines, establishing a new state of the art for this critical task in biology and medicine.
Authors: Xuanming Zhang
Abstract: This study proposes a novel analytical framework that integrates DBSCAN clustering with the Elastic Net regression model to address multifactorial problems characterized by structural complexity and multicollinearity, exemplified by carbon emissions analysis. DBSCAN is employed for unsupervised learning to objectively cluster features, while the Elastic Net is utilized for high-dimensional feature selection and complexity control. The Elastic Net is specifically chosen for its ability to balance feature selection and regularization by combining L1 (lasso) and L2 (ridge) penalties, making it particularly suited for datasets with correlated predictors. Applying this framework to energy consumption data from 46 industries in China (2000-2019) resulted in the identification of 16 categories. Emission characteristics and drivers were quantitatively assessed for each category, demonstrating the framework's capacity to identify primary emission sources and provide actionable insights. This research underscores the global applicability of the framework for analyzing complex regional challenges, such as carbon emissions, and highlights its potential to identify opportunities for emission reduction.
Authors: Matthieu Jonckheere (LAAS-SARA), Chiara Mignacco (LMO, CELESTE), Gilles Stoltz (LMO, CELESTE)
Abstract: We revisit the reduction of learning in adversarial Markov decision processes [MDPs] to adversarial learning based on $Q$--values; this reduction has been considered in a number of recent articles as one building block to perform policy optimization. Namely, we first consider and extend this reduction in an ideal setting where an oracle provides value functions: it may involve any adversarial learning strategy (not just exponential weights) and it may be based indifferently on $Q$--values or on advantage functions. We then present two extensions: on the one hand, convergence of the last iterate for a vast class of adversarial learning strategies (again, not just exponential weights), satisfying a property called monotonicity of weights; on the other hand, stronger regret criteria for learning in MDPs, inherited from the stronger regret criteria of adversarial learning called strongly adaptive regret and tracking regret. Third, we demonstrate how adversarial learning, also referred to as aggregation of experts, relates to aggregation (orchestration) of expert policies: we obtain stronger forms of performance guarantees in this setting than existing ones, via yet another, simple reduction. Finally, we discuss the impact of the reduction of learning in adversarial MDPs to adversarial learning in the practical scenarios where transition kernels are unknown and value functions must be learned. In particular, we review the literature and note that many strategies for policy optimization feature a policy-improvement step based on exponential weights with estimated $Q$--values. Our main message is that this step may be replaced by the application of any adversarial learning strategy on estimated $Q$--values or on estimated advantage functions. We leave the empirical evaluation of these twists for future research.
Authors: Talha Mehboob, Noman Bashir, Jesus Omana Iglesias, Michael Zink, David Irwin
Abstract: Federated Learning (FL) distributes machine learning (ML) training across edge devices to reduce data transfer overhead and protect data privacy. Since FL model training may span hundreds of devices and is thus resource- and energy-intensive, it has a significant carbon footprint. Importantly, since energy's carbon-intensity differs substantially (by up to 60$\times$) across locations, training on the same device using the same amount of energy, but at different locations, can incur widely different carbon emissions. While prior work has focused on improving FL's resource- and energy-efficiency by optimizing time-to-accuracy, it implicitly assumes all energy has the same carbon intensity and thus does not optimize carbon efficiency, i.e., work done per unit of carbon emitted. To address the problem, we design EcoLearn, which minimizes FL's carbon footprint without significantly affecting model accuracy or training time. EcoLearn achieves a favorable tradeoff by integrating carbon awareness into multiple aspects of FL training, including i) selecting clients with high data utility and low carbon, ii) provisioning more clients during the initial training rounds, and iii) mitigating stragglers by dynamically adjusting client over-provisioning based on carbon. We implement EcoLearn and its carbon-aware FL training policies in the Flower framework and show that it reduces the carbon footprint of training (by up to $10.8$$\times$) while maintaining model accuracy and training time (within $\sim$$1$\%) compared to state-of-the-art approaches.
Authors: Edith Cohen, Benjamin Cohen-Wang, Xin Lyu, Jelani Nelson, Tamas Sarlos, Uri Stemmer
Abstract: The Private Aggregation of Teacher Ensembles (PATE) framework enables privacy-preserving machine learning by aggregating responses from disjoint subsets of sensitive data. Adaptations of PATE to tasks with inherent output diversity such as text generation face a core tension: preserving output diversity reduces teacher agreement, which in turn increases the noise required for differential privacy, degrading utility. Yet suppressing diversity is counterproductive, as modern large language models encapsulate knowledge in their output distributions. We propose Hot PATE, a variant tailored to settings where outputs are distributions. We formally define what it means to preserve diversity and introduce an efficient aggregation mechanism that transfers diversity to the randomized output without incurring additional privacy cost. Our method can be implemented with only API access to proprietary models and serves as a drop-in replacement for existing "cold" PATE aggregators. Empirically, Hot PATE achieves orders-of-magnitude improvement on in-context learning tasks.
Authors: Stefania Bellavia, Francesco Della Santa, Alessandra Papini
Abstract: This paper introduces novel alternate training procedures for hard-parameter sharing Multi-Task Neural Networks (MTNNs). Traditional MTNN training faces challenges in managing conflicting loss gradients, often yielding sub-optimal performance. The proposed alternate training method updates shared and task-specific weights alternately through the epochs, exploiting the multi-head architecture of the model. This approach reduces computational costs per epoch and memory requirements. Convergence properties similar to those of the classical stochastic gradient method are established. Empirical experiments demonstrate enhanced training regularization and reduced computational demands. In summary, our alternate training procedures offer a promising advancement for the training of hard-parameter sharing MTNNs.
Authors: Shanay Mehta, Shlok Mehendale, Nicole Fernandes, Jyotirmoy Sarkar, Santonu Sarkar, Snehanshu Saha
Abstract: Detection of anomalous situations for complex mission-critical systems hold paramount importance when their service continuity needs to be ensured. A major challenge in detecting anomalies from the operational data arises due to the imbalanced class distribution problem since the anomalies are supposed to be rare events. This paper evaluates a diverse array of Machine Learning (ML)-based anomaly detection algorithms through a comprehensive benchmark study. The paper contributes significantly by conducting an unbiased comparison of various anomaly detection algorithms, spanning classical ML, including various tree-based approaches to Deep Learning (DL) and outlier detection methods. The inclusion of 104 publicly available enhances the diversity of the study, allowing a more realistic evaluation of algorithm performance and emphasizing the importance of adaptability to real-world scenarios. The paper evaluates the general notion of DL as a universal solution, showing that, while powerful, it is not always the best fit for every scenario. The findings reveal that recently proposed tree-based evolutionary algorithms match DL methods and sometimes outperform them in many instances of univariate data where the size of the data is small and number of anomalies are less than 10%. Specifically, tree-based approaches successfully detect singleton anomalies in datasets where DL falls short. To the best of the authors' knowledge, such a study on a large number of state-of-the-art algorithms using diverse data sets, with the objective of guiding researchers and practitioners in making informed algorithmic choices, has not been attempted earlier.
Authors: Haochuan Xu, Ninh Pham
Abstract: We present sDBSCAN, a scalable density-based clustering algorithm in high dimensions with cosine distance. Utilizing the neighborhood-preserving property of random projections, sDBSCAN can quickly identify core points and their neighborhoods, the primary hurdle of density-based clustering. Theoretically, sDBSCAN outputs a clustering structure similar to DBSCAN under mild conditions with high probability. To further facilitate sDBSCAN, we present sOPTICS, a scalable OPTICS for interactive exploration of the intrinsic clustering structure. We also extend sDBSCAN and sOPTICS to L2, L1, $\chi^2$, and Jensen-Shannon distances via random kernel features. Empirically, sDBSCAN is significantly faster and provides higher accuracy than many other clustering algorithms on real-world million-point data sets. On these data sets, sDBSCAN and sOPTICS run in a few minutes, while the scikit-learn's counterparts demand several hours or cannot run due to memory constraints.
Authors: Xuansheng Wu, Haiyan Zhao, Yaochen Zhu, Yucheng Shi, Fan Yang, Lijie Hu, Tianming Liu, Xiaoming Zhai, Wenlin Yao, Jundong Li, Mengnan Du, Ninghao Liu
Abstract: Explainable AI (XAI) refers to techniques that provide human-understandable insights into the workings of AI models. Recently, the focus of XAI is being extended toward explaining Large Language Models (LLMs). This extension calls for a significant transformation in the XAI methodologies for two reasons. First, many existing XAI methods cannot be directly applied to LLMs due to their complexity and advanced capabilities. Second, as LLMs are increasingly deployed in diverse applications, the role of XAI shifts from merely opening the ``black box'' to actively enhancing the productivity and applicability of LLMs in real-world settings. Meanwhile, the conversation and generation abilities of LLMs can reciprocally enhance XAI. Therefore, in this paper, we introduce Usable XAI in the context of LLMs by analyzing (1) how XAI can explain and improve LLM-based AI systems and (2) how XAI techniques can be improved by using LLMs. We introduce 10 strategies, introducing the key techniques for each and discussing their associated challenges. We also provide case studies to demonstrate how to obtain and leverage explanations. The code used in this paper can be found at: https://github.com/JacksonWuxs/UsableXAI_LLM.
Authors: Chao Yi, Yu-Hang He, De-Chuan Zhan, Han-Jia Ye
Abstract: Vision Language Models (VLMs) excel in zero-shot image classification by pairing images with textual category names. The expanding variety of Pre-Trained VLMs enhances the likelihood of identifying a suitable VLM for specific tasks. To better reuse the VLM resource and fully leverage its potential on different zero-shot image classification tasks, a promising strategy is selecting appropriate Pre-Trained VLMs from the VLM Zoo, relying solely on the text data of the target dataset without access to the dataset's images. In this paper, we analyze two inherent challenges in assessing the ability of a VLM in this Language-Only VLM selection: the "Modality Gap" - the disparity in VLM's embeddings across two different modalities, making text a less reliable substitute for images; and the "Capability Gap" - the discrepancy between the VLM's overall ranking and its ranking for target dataset, hindering direct prediction of a model's dataset-specific performance from its general performance. We propose VLM Selection With gAp Bridging (SWAB) to mitigate the negative impact of two gaps. SWAB first adopts optimal transport to capture the relevance between open-source and target datasets with a transportation matrix. It then uses this matrix to transfer useful statistics of VLMs from open-source datasets to the target dataset for bridging two gaps. By bridging two gaps to obtain better substitutes for test images, SWAB can accurately predict the performance ranking of different VLMs on the target task without the need for the dataset's images. Experiments across various VLMs and image classification datasets validate SWAB's effectiveness.
Authors: Soumyendu Sarkar, Avisek Naug, Ricardo Luna, Antonio Guillen, Vineet Gundecha, Sahand Ghorbanpour, Sajad Mousavi, Dejan Markovikj, Ashwin Ramesh Babu
Abstract: As machine learning workloads significantly increase energy consumption, sustainable data centers with low carbon emissions are becoming a top priority for governments and corporations worldwide. This requires a paradigm shift in optimizing power consumption in cooling and IT loads, shifting flexible loads based on the availability of renewable energy in the power grid, and leveraging battery storage from the uninterrupted power supply in data centers, using collaborative agents. The complex association between these optimization strategies and their dependencies on variable external factors like weather and the power grid carbon intensity makes this a hard problem. Currently, a real-time controller to optimize all these goals simultaneously in a dynamic real-world setting is lacking. We propose a Data Center Carbon Footprint Reduction (DC-CFR) multi-agent Reinforcement Learning (MARL) framework that optimizes data centers for the multiple objectives of carbon footprint reduction, energy consumption, and energy cost. The results show that the DC-CFR MARL agents effectively resolved the complex interdependencies in optimizing cooling, load shifting, and energy storage in real-time for various locations under real-world dynamic weather and grid carbon intensity conditions. DC-CFR significantly outperformed the industry standard ASHRAE controller with a considerable reduction in carbon emissions (14.5%), energy usage (14.4%), and energy cost (13.7%) when evaluated over one year across multiple geographical regions.
Authors: Mahtab Sarvmaili, Hassan Sajjad, Ga Wu
Abstract: Existing example-based prediction explanation methods often bridge test and training data points through the model's parameters or latent representations. While these methods offer clues to the causes of model predictions, they often exhibit innate shortcomings, such as incurring significant computational overhead or producing coarse-grained explanations. This paper presents a Highly-precise and Data-centric Explan}ation (HD-Explain) prediction explanation method that exploits properties of Kernelized Stein Discrepancy (KSD). Specifically, the KSD uniquely defines a parameterized kernel function for a trained model that encodes model-dependent data correlation. By leveraging the kernel function, one can identify training samples that provide the best predictive support to a test point efficiently. We conducted thorough analyses and experiments across multiple classification domains, where we show that HD-Explain outperforms existing methods from various aspects, including 1) preciseness (fine-grained explanation), 2) consistency, and 3) computation efficiency, leading to a surprisingly simple, effective, and robust prediction explanation solution.
Authors: Marie Siew, Haoran Zhang, Jong-Ik Park, Yuezhou Liu, Yichen Ruan, Lili Su, Stratis Ioannidis, Edmund Yeh, Carlee Joe-Wong
Abstract: Federated learning (FL) enables collaborative learning across multiple clients. In most FL work, all clients train a single learning task. However, the recent proliferation of FL applications may increasingly require multiple FL tasks to be trained simultaneously, sharing clients' computing and communication resources, which we call Multiple-Model Federated Learning (MMFL). Current MMFL algorithms use naive average-based client-task allocation schemes that can lead to unfair performance when FL tasks have heterogeneous difficulty levels, e.g., tasks with larger models may need more rounds and data to train. Just as naively allocating resources to generic computing jobs with heterogeneous resource needs can lead to unfair outcomes, naive allocation of clients to FL tasks can lead to unfairness, with some tasks having excessively long training times, or lower converged accuracies. Furthermore, in the FL setting, since clients are typically not paid for their training effort, we face a further challenge that some clients may not even be willing to train some tasks, e.g., due to high computational costs, which may exacerbate unfairness in training outcomes across tasks. We address both challenges by firstly designing FedFairMMFL, a difficulty-aware algorithm that dynamically allocates clients to tasks in each training round. We provide guarantees on airness and FedFairMMFL's convergence rate. We then propose a novel auction design that incentivizes clients to train multiple tasks, so as to fairly distribute clients' training efforts across the tasks. We show how our fairness-based learning and incentive mechanisms impact training convergence and finally evaluate our algorithm with multiple sets of learning tasks on real world datasets.
Authors: Zhixiong Jin, Dimitrios Tsitsokas, Nikolas Geroliminis, Ludovic Leclercq
Abstract: In large-scale traffic optimization, models based on Macroscopic Fundamental Diagram (MFD) are recognized for their efficiency in broad network analyses. However, they fail to reflect variations in the individual traffic status of each road link, leading to a gap in detailed traffic optimization and analysis. To address the limitation, this study introduces a Local Correction Factor (LCF) that represents local speed deviations between the actual link speed and the MFD average speed based on the link configuration. The LCF is calculated using a deep learning function that takes as inputs the average speed from the MFD and the road network configuration. Our framework integrates Graph Attention Networks (GATs) with Gated Recurrent Units (GRUs) to capture both the spatial configurations and temporal correlations within the network. Coupled with a strategic network partitioning method, our model enhances the precision of link-level traffic speed estimations while preserving the computational advantages of aggregate models. In our experiments, we evaluate the proposed LCF across various urban traffic scenarios, including different levels of origin-destination trip demand and distribution, as well as diverse road configurations. The results demonstrate the robust adaptability and effectiveness of the proposed model. Furthermore, we validate the practicality of our model by calculating the travel time of each randomly generated path, achieving an average error reduction of approximately 84% relative to MFD-based results.
Authors: Maria-Florina Balcan, Dravyansh Sharma
Abstract: Decision trees and their ensembles are popular in machine learning as easy-to-understand models. Several techniques have been proposed in the literature for learning tree-based classifiers, with different techniques working well for data from different domains. In this work, we develop approaches to design tree-based learning algorithms given repeated access to data from the same domain. We study multiple formulations covering different aspects and popular techniques for learning decision tree based approaches. We propose novel parameterized classes of node splitting criteria in top-down algorithms, which interpolate between popularly used entropy and Gini impurity based criteria, and provide theoretical bounds on the number of samples needed to learn the splitting function appropriate for the data at hand. We also study the sample complexity of tuning prior parameters in Bayesian decision tree learning, and extend our results to decision tree regression. We further consider the problem of tuning hyperparameters in pruning the decision tree for classical pruning algorithms including min-cost complexity pruning. In addition, our techniques can be used to optimize the explainability versus accuracy trade-off when using decision trees. We extend our results to tuning popular tree-based ensembles, including random forests and gradient-boosted trees. We demonstrate the significance of our approach on real world datasets by learning data-specific decision trees which are simultaneously more accurate and interpretable.
Authors: Mathieu Chevalley, Patrick Schwab, Arash Mehrjou
Abstract: Targeted and uniform interventions to a system are crucial for unveiling causal relationships. While several methods have been developed to leverage interventional data for causal structure learning, their practical application in real-world scenarios often remains challenging. Recent benchmark studies have highlighted these difficulties, even when large numbers of single-variable intervention samples are available. In this work, we demonstrate, both theoretically and empirically, that such datasets contain a wealth of causal information that can be effectively extracted under realistic assumptions about the data distribution. More specifically, we introduce a novel variant of interventional faithfulness, which relies on comparisons between the marginal distributions of each variable across observational and interventional settings, and we introduce a score on causal orders. Under this assumption, we are able to prove strong theoretical guarantees on the optimum of our score that also hold for large-scale settings. To empirically verify our theory, we introduce Intersort, an algorithm designed to infer the causal order from datasets containing large numbers of single-variable interventions by approximately optimizing our score. Intersort outperforms baselines (GIES, DCDI, PC and EASE) on almost all simulated data settings replicating common benchmarks in the field. Our proposed novel approach to modeling interventional datasets thus offers a promising avenue for advancing causal inference, highlighting significant potential for further enhancements under realistic assumptions.
Authors: St\'ephane Rivaud (MLIA, TAU), Louis Fournier (MLIA), Thomas Pumir (MILA), Eugene Belilovsky (MILA), Michael Eickenberg, Edouard Oyallon
Abstract: Reversible architectures have been shown to be capable of performing on par with their non-reversible architectures, being applied in deep learning for memory savings and generative modeling. In this work, we show how reversible architectures can solve challenges in parallelizing deep model training. We introduce PETRA, a novel alternative to backpropagation for parallelizing gradient computations. PETRA facilitates effective model parallelism by enabling stages (i.e., a set of layers) to compute independently on different devices, while only needing to communicate activations and gradients between each other. By decoupling the forward and backward passes and keeping a single updated version of the parameters, the need for weight stashing is also removed. We develop a custom autograd-like training framework for PETRA, and we demonstrate its effectiveness on CIFAR-10, ImageNet32, and ImageNet, achieving competitive accuracies comparable to backpropagation using ResNet-18, ResNet-34, and ResNet-50 models.
Authors: Adel Nabli (MLIA, Mila), Louis Fournier (MLIA), Pierre Erbacher (MLIA), Louis Serrano (MLIA), Eugene Belilovsky (Mila), Edouard Oyallon (MLIA)
Abstract: Training LLMs relies on distributed implementations using multiple GPUs to compute gradients in parallel with sharded optimizers. However, synchronizing gradients in data parallel setups introduces communication overhead that grows with the number of workers, limiting parallelization efficiency. Local optimization algorithms reduce communications but incur high memory costs as they prevent optimizer state sharding, hindering scalability. To address this, we propose \textbf{AC}cumulate while \textbf{CO}mmunicate (\acco), a memory-efficient optimization algorithm for distributed LLM training. By synchronizing delayed gradients while computing new ones, \acco~reduces GPU idle time and supports heterogeneous hardware. To mitigate the convergence issues caused by delayed updates, we introduce a novel technique ensuring training dynamics align with standard distributed optimization. Compared to ZeRO-1, our approach is significantly faster and scales effectively across heterogeneous hardware.
Authors: Shojiro Yamabe, Kazuto Fukuchi, Jun Sakuma
Abstract: This study investigates behavior-targeted attacks on reinforcement learning and their countermeasures. Behavior-targeted attacks aim to manipulate the victim's behavior as desired by the adversary through adversarial interventions in state observations. Existing behavior-targeted attacks have some limitations, such as requiring white-box access to the victim's policy. To address this, we propose a novel attack method using imitation learning from adversarial demonstrations, which works under limited access to the victim's policy and is environment-agnostic. In addition, our theoretical analysis proves that the policy's sensitivity to state changes impacts defense performance, particularly in the early stages of the trajectory. Based on this insight, we propose time-discounted regularization, which enhances robustness against attacks while maintaining task performance. To the best of our knowledge, this is the first defense strategy specifically designed for behavior-targeted attacks.
Authors: Calarina Muslimani, Bram Grooten, Deepak Ranganatha Sastry Mamillapalli, Mykola Pechenizkiy, Decebal Constantin Mocanu, Matthew E. Taylor
Abstract: To integrate into human-centered environments, autonomous agents must learn from and adapt to humans in their native settings. Preference-based reinforcement learning (PbRL) can enable this by learning reward functions from human preferences. However, humans live in a world full of diverse information, most of which is irrelevant to completing any particular task. It then becomes essential that agents learn to focus on the subset of task-relevant state features. To that end, this work proposes R2N (Robust-to-Noise), the first PbRL algorithm that leverages principles of dynamic sparse training to learn robust reward models that can focus on task-relevant features. In experiments with a simulated teacher, we demonstrate that R2N can adapt the sparse connectivity of its neural networks to focus on task-relevant features, enabling R2N to significantly outperform several sparse training and PbRL algorithms across simulated robotic environments.
Authors: Jesse van Oostrum, Carlotta Langer, Nihat Ay
Abstract: In this paper we present a concise mathematical description of active inference in discrete time. The main part of the paper serves as a basic introduction to the topic, including a detailed example of the action selection mechanism. The appendix discusses the more subtle mathematical details, targeting readers who have already studied the active inference literature but struggle to make sense of the mathematical details and derivations. Throughout, we emphasize precise and standard mathematical notation, ensuring consistency with existing texts and linking all equations to widely used references on active inference. Additionally, we provide Python code that implements the action selection and learning mechanisms described in this paper and is compatible with pymdp environments.
Authors: Jacob E. Kooi, Mark Hoogendoorn, Vincent Fran\c{c}ois-Lavet
Abstract: Activation functions are one of the key components of a deep neural network. The most commonly used activation functions can be classed into the category of continuously differentiable (e.g. tanh) and piece-wise linear functions (e.g. ReLU), both having their own strengths and drawbacks with respect to downstream performance and representation capacity through learning. In reinforcement learning, the performance of continuously differentiable activations often falls short as compared to piece-wise linear functions. We show that the dying neuron problem in RL is not exclusive to ReLUs and actually leads to additional problems in the case of continuously differentiable activations such as tanh. To alleviate the dying neuron problem with these activations, we propose a Hadamard representation that unlocks the advantages of continuously differentiable activations. Using DQN, PPO and PQN in the Atari domain, we show faster learning, a reduction in dead neurons and increased effective rank.
Authors: M\'arton Kardos, Jan Kostkan, Arnault-Quentin Vermillet, Kristoffer Nielbo, Kenneth Enevoldsen, Roberta Rocca
Abstract: Topic models are useful tools for discovering latent semantic structures in large textual corpora. Recent efforts have been oriented at incorporating contextual representations in topic modeling and have been shown to outperform classical topic models. These approaches are typically slow, volatile, and require heavy preprocessing for optimal results. We present Semantic Signal Separation ($S^3$), a theory-driven topic modeling approach in neural embedding spaces. $S^3$ conceptualizes topics as independent axes of semantic space and uncovers these by decomposing contextualized document embeddings using Independent Component Analysis. Our approach provides diverse and highly coherent topics, requires no preprocessing, and is demonstrated to be the fastest contextual topic model, being, on average, 4.5x faster than the runner-up BERTopic. We offer an implementation of $S^3$, and all contextual baselines, in the Turftopic Python package.
Authors: Alexandre Jacquillat, Michael Lingzhi Li
Abstract: We define an online learning and optimization problem with discrete and irreversible decisions contributing toward a coverage target. In each period, a decision-maker selects facilities to open, receives information on the success of each one, and updates a classification model to guide future decisions. The goal is to minimize facility openings under a chance constraint reflecting the coverage target, in an asymptotic regime characterized by a large target number of facilities $m\to\infty$ but a finite horizon $T \in \mathcal{Z}_+$. We prove that, under statistical conditions, the online classifier converges to the Bayes-optimal classifier at a rate of at best $\mathcal{O}(1/\sqrt n)$. Thus, we formulate our online learning and optimization problem, with a generalized learning rate $r>0$ and a residual error $1-p$. We derive an asymptotically optimal algorithm and an asymptotically tight lower bound. The regret grows in $\Theta\left(m^{\frac{1-r}{1-r^T}}\right)$ if $p=1$ (perfect learning) or in $\Theta\left(\max\left\{m^{\frac{1-r}{1-r^T}},\sqrt{m}\right\}\right)$ otherwise; in particular, the regret rate is sub-linear and converges exponentially fast to its infinite-horizon limit. We extend this result to a more complicated facility location setting in a bipartite facility-customer graph with a target on customer coverage. Throughout, constructive proofs identify a policy featuring limited exploration initially and fast exploitation later on once uncertainty gets mitigated. These results uncover the benefits of limited online learning and optimization through pilot programs prior to full-fledged expansion.
Authors: Zhiqi Bu, Shiyun Xu
Abstract: We propose the generalized Newton's method (GeN) -- a Hessian-informed approach that applies to any optimizer such as SGD and Adam, and covers the Newton-Raphson method as a sub-case. Our method automatically and dynamically selects the learning rate that accelerates the convergence, without the intensive tuning of the learning rate scheduler. In practice, our method is easily implementable, since it only requires additional forward passes with almost zero computational overhead (in terms of training time and memory cost), if the overhead is amortized over many iterations. We present extensive experiments on language and vision tasks (e.g. GPT and ResNet) to showcase that GeN optimizers match the state-of-the-art performance, which was achieved with carefully tuned learning rate schedulers.
Authors: Hrishikesh Viswanath, Yue Chang, Aleksey Panas, Julius Berner, Peter Yichen Chen, Aniket Bera
Abstract: Simulating complex physical systems governed by Lagrangian dynamics often requires solving partial differential equations (PDEs) over high-resolution spatial domains, resulting in substantial computational costs. We present GIOROM (\textit{G}raph \textit{I}nf\textit{O}rmed \textit{R}educed \textit{O}rder \textit{M}odeling), a data-driven discretization invariant framework for accelerating Lagrangian simulations through reduced-order modeling (ROM). Previous discretization invariant ROM approaches rely on PDE time-steppers for spatiotemporally evolving low-dimensional reduced-order latent states. Instead, we leverage a data-driven graph-based neural approximation of the PDE solution operator. This operator estimates point-wise function values from a sparse set of input observations, reducing reliance on known governing equations of numerical solvers. Order reduction is achieved by embedding these point-wise estimates within the reduced-order latent space using a learned kernel parameterization. This latent representation enables the reconstruction of the solution at arbitrary spatial query points by evolving latent variables over local neighborhoods on the solution manifold, using the kernel. Empirically, GIOROM achieves a 6.6$\times$-32$\times$ reduction in input dimensionality while maintaining high-fidelity reconstructions across diverse Lagrangian regimes including fluid flows, granular media, and elastoplastic dynamics. The resulting framework enables learnable, data-driven and discretization-invariant order-reduction with reduced reliance on analytical PDE formulations. Our code is at \href{https://github.com/HrishikeshVish/GIOROM}{https://github.com/HrishikeshVish/GIOROM}
URLs: https://github.com/HrishikeshVish/GIOROM, https://github.com/HrishikeshVish/GIOROM
Authors: Mengzhao Chen, Wenqi Shao, Peng Xu, Jiahao Wang, Peng Gao, Kaipeng Zhang, Ping Luo
Abstract: Large language models (LLMs) are crucial in modern natural language processing and artificial intelligence. However, they face challenges in managing their significant memory requirements. Although quantization-aware training (QAT) offers a solution by reducing memory consumption through low-bit representations with minimal accuracy loss, it is impractical due to substantial training resources. To address this, we propose Efficient Quantization-Aware Training (EfficientQAT), a more feasible QAT algorithm. EfficientQAT involves two consecutive phases: Block-wise training of all parameters (Block-AP) and end-to-end training of quantization parameters (E2E-QP). To the best of our knowledge, Block-AP is the first method to enable direct training of all parameters in a block-wise manner, reducing accuracy loss in low-bit scenarios by enhancing the solution space during optimization. E2E-QP then trains only the quantization parameters (step sizes) end-to-end, further improving the performance of quantized models by considering interactions among all sub-modules. Extensive experiments demonstrate that EfficientQAT outperforms previous quantization methods across a range of models, including base LLMs, instruction-tuned LLMs, and multimodal LLMs, with scales from 7B to 70B parameters at various quantization bits. For instance, EfficientQAT obtains a 2-bit Llama-2-70B model on a single A100-80GB GPU in 41 hours, with less than 3 points accuracy degradation compared to the full precision (69.48 vs. 72.41). Code is available at https://github.com/OpenGVLab/EfficientQAT.
Authors: Chengzhi Lin, Shuchang Liu, Chuyuan Wang, Yongqi Liu
Abstract: Accurately predicting watch time is crucial for optimizing recommendations and user experience in short video platforms. However, existing methods that estimate a single average watch time often fail to capture the inherent uncertainty in user engagement patterns. In this paper, we propose Conditional Quantile Estimation (CQE) to model the entire conditional distribution of watch time. Using quantile regression, CQE characterizes the complex watch-time distribution for each user-video pair, providing a flexible and comprehensive approach to understanding user behavior. We further design multiple strategies to combine the quantile estimates, adapting to different recommendation scenarios and user preferences. Extensive offline experiments and online A/B tests demonstrate the superiority of CQE in watch-time prediction and user engagement modeling. Specifically, deploying CQE online on a large-scale platform with hundreds of millions of daily active users has led to substantial gains in key evaluation metrics, including active days, engagement time, and video views. These results highlight the practical impact of our proposed approach in enhancing the user experience and overall performance of the short video recommendation system. The code will be released https://github.com/justopit/CQE.
Authors: Yingru Li, Jiawei Xu, Baoxiang Wang, Zhi-Quan Luo
Abstract: Thompson Sampling is a principled method for balancing exploration and exploitation, but its real-world adoption faces computational challenges in large-scale or non-conjugate settings. While ensemble-based approaches offer partial remedies, they typically require prohibitively large ensemble sizes. We propose Ensemble++, a scalable exploration framework using a novel shared-factor ensemble architecture with random linear combinations. For linear bandits, we provide theoretical guarantees showing that Ensemble++ achieves regret comparable to exact Thompson Sampling with only $\Theta(d \log T)$ ensemble sizes--significantly outperforming prior methods. Crucially, this efficiency holds across both compact and finite action sets with either time-invariant or time-varying contexts without configuration changes. We extend this theoretical foundation to nonlinear rewards by replacing fixed features with learnable neural representations while preserving the same incremental update principle, effectively bridging theory and practice for real-world tasks. Comprehensive experiments across linear, quadratic, neural, and GPT-based contextual bandits validate our theoretical findings and demonstrate Ensemble++'s superior regret-computation tradeoff versus state-of-the-art methods.
Authors: Jingren Liu, Zhong Ji, YunLong Yu, Jiale Cao, Yanwei Pang, Jungong Han, Xuelong Li
Abstract: Parameter-efficient fine-tuning for continual learning (PEFT-CL) has shown promise in adapting pre-trained models to sequential tasks while mitigating catastrophic forgetting problem. However, understanding the mechanisms that dictate continual performance in this paradigm remains elusive. To unravel this mystery, we undertake a rigorous analysis of PEFT-CL dynamics to derive relevant metrics for continual scenarios using Neural Tangent Kernel (NTK) theory. With the aid of NTK as a mathematical analysis tool, we recast the challenge of test-time forgetting into the quantifiable generalization gaps during training, identifying three key factors that influence these gaps and the performance of PEFT-CL: training sample size, task-level feature orthogonality, and regularization. To address these challenges, we introduce NTK-CL, a novel framework that eliminates task-specific parameter storage while adaptively generating task-relevant features. Aligning with theoretical guidance, NTK-CL triples the feature representation of each sample, theoretically and empirically reducing the magnitude of both task-interplay and task-specific generalization gaps. Grounded in NTK analysis, our framework imposes an adaptive exponential moving average mechanism and constraints on task-level feature orthogonality, maintaining intra-task NTK forms while attenuating inter-task NTK forms. Ultimately, by fine-tuning optimizable parameters with appropriate regularization, NTK-CL achieves state-of-the-art performance on established PEFT-CL benchmarks. This work provides a theoretical foundation for understanding and improving PEFT-CL models, offering insights into the interplay between feature representation, task orthogonality, and generalization, contributing to the development of more efficient continual learning systems.
Authors: Borong Zhang, Mart\'in Guerra, Qin Li, Leonardo Zepeda-N\'u\~nez
Abstract: We present Wideband Back-Projection Diffusion, an end-to-end probabilistic framework for approximating the posterior distribution induced by the inverse scattering map from wideband scattering data. This framework produces highly accurate reconstructions, leveraging conditional diffusion models to draw samples, and also honors the symmetries of the underlying physics of wave-propagation. The procedure is factored into two steps: the first step, inspired by the filtered back-propagation formula, transforms data into a physics-based latent representation, while the second step learns a conditional score function conditioned on this latent representation. These two steps individually obey their associated symmetries and are amenable to compression by imposing the rank structure found in the filtered back-projection formula. Empirically, our framework has both low sample and computational complexity, with its number of parameters scaling only sub-linearly with the target resolution, and has stable training dynamics. It provides sharp reconstructions effortlessly and is capable of recovering even sub-Nyquist features in the multiple-scattering regime.
Authors: Yijun Bian, Yujie Luo, Ping Xu
Abstract: Discrimination mitigation within machine learning (ML) models could be complicated because multiple factors may be interwoven hierarchically and historically. Yet few existing fairness measures can capture the discrimination level within ML models in the face of multiple sensitive attributes (SAs). To bridge this gap, we propose a fairness measure based on distances between sets from a manifold perspective, named as 'Harmonic Fairness measure via Manifolds (HFM)' with two optional versions, which can deal with a fine-grained discrimination evaluation for several SAs of multiple values. Because directly computing HFM may be costly, to accelerate its subprocedure -- the computation of distances of sets, we further propose two approximation algorithms named 'Approximation of distance between sets for one sensitive attribute with multiple values (ApproxDist)' and 'Approximation of extended distance between sets for several sensitive attributes with multiple values (ExtendDist)' to respectively resolve bias evaluation of one single SA with multiple values and that of several SAs with multiple values. Moreover, we provide an algorithmic effectiveness analysis for ApproxDist under certain assumptions to explain how well it could work. The empirical results demonstrate that our proposed fairness measure HFM is valid and approximation algorithms (i.e. ApproxDist and ExtendDist) are effective and efficient.
Authors: Ilya Kuleshov, Evgenia Romanenkova, Vladislav Zhuzhel, Galina Boeva, Evgeni Vorsin, Alexey Zaytsev
Abstract: Neural CDEs provide a natural way to process the temporal evolution of irregular time series. The number of function evaluations (NFE) is these systems' natural analog of depth (the number of layers in traditional neural networks). It is usually regulated via solver error tolerance: lower tolerance means higher numerical precision, requiring more integration steps. However, lowering tolerances does not adequately increase the models' expressiveness. We propose a simple yet effective alternative: scaling the integration time horizon to increase NFEs and "deepen`` the model. Increasing the integration interval causes uncontrollable growth in conventional vector fields, so we also propose a way to stabilize the dynamics via Negative Feedback (NF). It ensures provable stability without constraining flexibility. It also implies robustness: we provide theoretical bounds for Neural ODE risk using Gaussian process theory. Experiments on four open datasets demonstrate that our method, DeNOTS, outperforms existing approaches~ -- ~including recent Neural RDEs and state space models,~ -- ~achieving up to $20\%$ improvement in metrics. DeNOTS combines expressiveness, stability, and robustness, enabling reliable modelling in continuous-time domains.
Authors: Haozhe Ma, Zhengding Luo, Thanh Vinh Vo, Kuankuan Sima, Tze-Yun Leong
Abstract: Reward shaping is effective in addressing the sparse-reward challenge in reinforcement learning by providing immediate feedback through auxiliary informative rewards. Based on the reward shaping strategy, we propose a novel multi-task reinforcement learning framework that integrates a centralized reward agent (CRA) and multiple distributed policy agents. The CRA functions as a knowledge pool, which aims to distill knowledge from various tasks and distribute it to individual policy agents to improve learning efficiency. Specifically, the shaped rewards serve as a straightforward metric to encode knowledge. This framework not only enhances knowledge sharing across established tasks but also adapts to new tasks by transferring meaningful reward signals. We validate the proposed method on both discrete and continuous domains, including the representative meta world benchmark, demonstrating its robustness in multi-task sparse-reward settings and its effective transferability to unseen tasks.
Authors: Qi Dong, Rubing Huang, Chenhui Cui, Dave Towey, Ling Zhou, Jinyu Tian, Jianzhou Wang
Abstract: Short-Term Electricity-Load Forecasting (STELF) refers to the prediction of the immediate demand (in the next few hours to several days) for the power system. Various external factors, such as weather changes and the emergence of new electricity consumption scenarios, can impact electricity demand, causing load data to fluctuate and become non-linear, which increases the complexity and difficulty of STELF. In the past decade, deep learning has been applied to STELF, modeling and predicting electricity demand with high accuracy, and contributing significantly to the development of STELF. This paper provides a comprehensive survey on deep-learning-based STELF over the past ten years. It examines the entire forecasting process, including data pre-processing, feature extraction, deep-learning modeling and optimization, and results evaluation. This paper also identifies some research challenges and potential research directions to be further investigated in future work.
Authors: Xinhe Wang, Pingbang Hu, Junwei Deng, Jiaqi W. Ma
Abstract: Data attribution aims to quantify the contribution of individual training data points to the outputs of an AI model, which has been used to measure the value of training data and compensate data providers. Given the impact on financial decisions and compensation mechanisms, a critical question arises concerning the adversarial robustness of data attribution methods. However, there has been little to no systematic research addressing this issue. In this work, we aim to bridge this gap by detailing a threat model with clear assumptions about the adversary's goal and capabilities and proposing principled adversarial attack methods on data attribution. We present two methods, Shadow Attack and Outlier Attack, which generate manipulated datasets to inflate the compensation adversarially. The Shadow Attack leverages knowledge about the data distribution in the AI applications, and derives adversarial perturbations through "shadow training", a technique commonly used in membership inference attacks. In contrast, the Outlier Attack does not assume any knowledge about the data distribution and relies solely on black-box queries to the target model's predictions. It exploits an inductive bias present in many data attribution methods - outlier data points are more likely to be influential - and employs adversarial examples to generate manipulated datasets. Empirically, in image classification and text generation tasks, the Shadow Attack can inflate the data-attribution-based compensation by at least 200%, while the Outlier Attack achieves compensation inflation ranging from 185% to as much as 643%. Our implementation is ready at https://github.com/TRAIS-Lab/adversarial-attack-data-attribution.
URLs: https://github.com/TRAIS-Lab/adversarial-attack-data-attribution.
Authors: Yuchen Shen, Haomin Wen, Leman Akoglu
Abstract: Outlier detection (OD) has a vast literature as it finds numerous real-world applications. Being an unsupervised task, model selection is a key bottleneck for OD without label supervision. Despite a long list of available OD algorithms with tunable hyperparameters, the lack of systematic approaches for unsupervised algorithm and hyperparameter selection limits their effective use in practice. In this paper, we present FoMo-0D, a pre-trained Foundation Model for zero/0-shot OD on tabular data, which bypasses the hurdle of model selection altogether. Having been pre-trained on synthetic data, FoMo-0D can directly predict the (outlier/inlier) label of test samples without parameter fine-tuning -- requiring no labeled data, and no additional training or hyperparameter tuning when given a new task. Extensive experiments on 57 real-world datasets against 26 baselines show that FoMo-0D is highly competitive; outperforming the majority of the baselines with no statistically significant difference from the 2nd best method. Further, FoMo-0D is efficient in inference time requiring only 7.7 ms per sample on average, with at least 7x speed-up compared to previous methods. To facilitate future research, our implementations for data synthesis and pre-training as well as model checkpoints are openly available at https://anonymous.4open.science/r/PFN40D.
Authors: Sitao Luan, Qincheng Lu, Chenqing Hua, Xinyu Wang, Jiaqi Zhu, Xiao-Wen Chang
Abstract: Over the past decade, Graph Neural Networks (GNNs) have achieved great success on machine learning tasks with relational data. However, recent studies have found that heterophily can cause significant performance degradation of GNNs, especially on node-level tasks. Numerous heterophilic benchmark datasets have been put forward to validate the efficacy of heterophily-specific GNNs, and various homophily metrics have been designed to help recognize these challenging datasets. Nevertheless, there still exist multiple pitfalls that severely hinder the proper evaluation of new models and metrics: 1) lack of hyperparameter tuning; 2) insufficient evaluation on the truly challenging heterophilic datasets; 3) missing quantitative evaluation for homophily metrics on synthetic graphs. To overcome these challenges, we first train and fine-tune baseline models on $27$ most widely used benchmark datasets, and categorize them into three distinct groups: malignant, benign and ambiguous heterophilic datasets. We identify malignant and ambiguous heterophily as the truly challenging subsets of tasks, and to our best knowledge, we are the first to propose such taxonomy. Then, we re-evaluate $11$ state-of-the-arts (SOTA) GNNs, covering six popular methods, with fine-tuned hyperparameters on different groups of heterophilic datasets. Based on the model performance, we comprehensively reassess the effectiveness of different methods on heterophily. At last, we evaluate $11$ popular homophily metrics on synthetic graphs with three different graph generation approaches. To overcome the unreliability of observation-based comparison and evaluation, we conduct the first quantitative evaluation and provide detailed analysis.
Authors: Bangchao Deng, Xin Jing, Tianyue Yang, Bingqing Qu, Dingqi Yang, Philippe Cudre-Mauroux
Abstract: Human trajectory data, which plays a crucial role in various applications such as crowd management and epidemic prevention, is challenging to obtain due to practical constraints and privacy concerns. In this context, synthetic human trajectory data is generated to simulate as close as possible to real-world human trajectories, often under summary statistics and distributional similarities. However, these similarities oversimplify complex human mobility patterns (a.k.a. ``Datasaurus''), resulting in intrinsic biases in both generative model design and benchmarks of the generated trajectories. Against this background, we propose MIRAGE, a huMan-Imitative tRAjectory GenErative model designed as a neural Temporal Point Process integrating an Exploration and Preferential Return model. It imitates the human decision-making process in trajectory generation, rather than fitting any specific statistical distributions as traditional methods do, thus avoiding the Datasaurus issue. We also propose a comprehensive task-based evaluation protocol beyond Datasaurus to systematically benchmark trajectory generative models on four typical downstream tasks, integrating multiple techniques and evaluation metrics for each task, to assess the ultimate utility of the generated trajectories. We conduct a thorough evaluation of MIRAGE on three real-world user trajectory datasets against a sizeable collection of baselines. Results show that compared to the best baselines, MIRAGE-generated trajectory data not only achieves the best statistical and distributional similarities with 59.0-67.7% improvement, but also yields the best performance in the task-based evaluation with 10.9-33.4% improvement. A series of ablation studies also validate the key design choices of MIRAGE.
Authors: Keitaro Sakamoto, Issei Sato
Abstract: Attention mechanism is a fundamental component of the transformer model and plays a significant role in its success. However, the theoretical understanding of how attention learns to select tokens is still an emerging area of research. In this work, we study the training dynamics and generalization ability of the attention mechanism under classification problems with label noise. We show that, with the characterization of signal-to-noise ratio (SNR), the token selection of attention mechanism achieves benign overfitting, i.e., maintaining high generalization performance despite fitting label noise. Our work also demonstrates an interesting delayed acquisition of generalization after an initial phase of overfitting. Finally, we provide experiments to support our theoretical analysis using both synthetic and real-world datasets.
Authors: Mayank Nagda, Phil Ostheimer, Thomas Specht, Frank Rhein, Fabian Jirasek, Stephan Mandt, Marius Kloft, Sophie Fellenz
Abstract: Physics-Informed Neural Networks (PINNs) solve partial differential equations using deep learning. However, conventional PINNs perform pointwise predictions that neglect dependencies within a domain, which may result in suboptimal solutions. We introduce SetPINNs, a framework that effectively captures local dependencies. With a finite element-inspired sampling scheme, we partition the domain into sets to model local dependencies while simultaneously enforcing physical laws. We provide a rigorous theoretical analysis showing that SetPINNs yield unbiased, lower-variance estimates of residual energy and its gradients, ensuring improved domain coverage and reduced residual error. Extensive experiments on synthetic and real-world tasks show improved accuracy, efficiency, and robustness.
Authors: Liangzu Peng, Juan Elenter, Joshua Agterberg, Alejandro Ribeiro, Ren\'e Vidal
Abstract: The goal of continual learning (CL) is to train a model that can solve multiple tasks presented sequentially. Recent CL approaches have achieved strong performance by leveraging large pre-trained models that generalize well to downstream tasks. However, such methods lack theoretical guarantees, making them prone to unexpected failures. Conversely, principled CL approaches often fail to achieve competitive performance. In this work, we aim to bridge this gap between theory and practice by designing a simple CL method that is theoretically sound and highly performant. Specifically, we lift pre-trained features into a higher dimensional space and formulate an over-parametrized minimum-norm least-squares problem. We find that the lifted features are highly ill-conditioned, potentially leading to large training errors (numerical instability) and increased generalization errors. We address these challenges by continually truncating the singular value decomposition of the lifted features. Our approach, termed LoRanPAC, is stable with respect to the choice of hyperparameters, can handle hundreds of tasks, and outperforms state-of-the-art CL methods on multiple datasets. Importantly, our method satisfies a recurrence relation throughout its continual learning process, which allows us to prove it maintains small training and test errors by appropriately truncating a fraction of SVD factors. This results in a stable continual learning method with strong empirical performance and theoretical guarantees. Code available: https://github.com/liangzu/loranpac.
Authors: Brahma S. Pavse, Yudong Chen, Qiaomin Xie, Josiah P. Hanna
Abstract: In reinforcement learning, offline value function learning is the procedure of using an offline dataset to estimate the expected discounted return from each state when taking actions according to a fixed target policy. The stability of this procedure, i.e., whether it converges to its fixed-point, critically depends on the representations of the state-action pairs. Poorly learned representations can make value function learning unstable, or even divergent. Therefore, it is critical to stabilize value function learning by explicitly shaping the state-action representations. Recently, the class of bisimulation-based algorithms have shown promise in shaping representations for control. However, it is still unclear if this class of methods can \emph{stabilize} value function learning. In this work, we investigate this question and answer it affirmatively. We introduce a bisimulation-based algorithm called kernel representations for offline policy evaluation (\textsc{krope}). \textsc{krope} uses a kernel to shape state-action representations such that state-action pairs that have similar immediate rewards and lead to similar next state-action pairs under the target policy also have similar representations. We show that \textsc{krope}: 1) learns stable representations and 2) leads to lower value error than baselines. Our analysis provides new theoretical insight into the stability properties of bisimulation-based methods and suggests that practitioners can use these methods to improve the stability and accuracy of offline evaluation of reinforcement learning agents.
Authors: Yuka Hashimoto, Tomoharu Iwata
Abstract: We propose deep Koopman-layered models with learnable parameters in the form of Toeplitz matrices for analyzing the transition of the dynamics of time-series data. The proposed model has both theoretical solidness and flexibility. By virtue of the universal property of Toeplitz matrices and the reproducing property underlying the model, we show its universality and generalization property. In addition, the flexibility of the proposed model enables the model to fit time-series data coming from nonautonomous dynamical systems. When training the model, we apply Krylov subspace methods for efficient computations, which establish a new connection between Koopman operators and numerical linear algebra. We also empirically demonstrate that the proposed model outperforms existing methods on eigenvalue estimation of multiple Koopman operators for nonautonomous systems.
Authors: Song Li, Yang Tan, Song Ke, Liang Hong, Bingxin Zhou
Abstract: Immunogenicity prediction is a central topic in reverse vaccinology for finding candidate vaccines that can trigger protective immune responses. Existing approaches typically rely on highly compressed features and simple model architectures, leading to limited prediction accuracy and poor generalizability. To address these challenges, we introduce VenusVaccine, a novel deep learning solution with a dual attention mechanism that integrates pre-trained latent vector representations of protein sequences and structures. We also compile the most comprehensive immunogenicity dataset to date, encompassing over 7000 antigen sequences, structures, and immunogenicity labels from bacteria, virus, and tumor. Extensive experiments demonstrate that VenusVaccine outperforms existing methods across a wide range of evaluation metrics. Furthermore, we establish a post-hoc validation protocol to assess the practical significance of deep learning models in tackling vaccine design challenges. Our work provides an effective tool for vaccine design and sets valuable benchmarks for future research. The implementation is at https://github.com/songleee/VenusVaccine.
Authors: JJ Wilson, Maya Bechler-Speicher, Petar Veli\v{c}kovi\'c
Abstract: In spite of the plethora of success stories with graph neural networks (GNNs) on modelling graph-structured data, they are notoriously vulnerable to over-squashing, whereby tasks necessitate the mixing of information between distance pairs of nodes. To address this problem, prior work suggests rewiring the graph structure to improve information flow. Alternatively, a significant body of research has dedicated itself to discovering and precomputing bottleneck-free graph structures to ameliorate over-squashing. One well regarded family of bottleneck-free graphs within the mathematical community are expander graphs, with prior work -- Expander Graph Propagation (EGP) -- proposing the use of a well-known expander graph family -- the Cayley graphs of the $\mathrm{SL}(2,\mathbb{Z}_n)$ special linear group -- as a computational template for GNNs. However, in EGP the computational graphs used are truncated to align with a given input graph. In this work, we show that truncation is detrimental to the coveted expansion properties. Instead, we propose CGP, a method to propagate information over a complete Cayley graph structure, thereby ensuring it is bottleneck-free to better alleviate over-squashing. Our empirical evidence across several real-world datasets not only shows that CGP recovers significant improvements as compared to EGP, but it is also akin to or outperforms computationally complex graph rewiring techniques.
Authors: Nan Huang, Haishuai Wang, Zihuai He, Marinka Zitnik, Xiang Zhang
Abstract: Medical time series (MedTS) classification suffers from poor generalizability in real-world deployment due to inter- and intra-dataset heterogeneity, such as varying numbers of channels, signal lengths, task definitions, and patient characteristics. To address this, we propose FORMED, a novel framework for repurposing a backbone foundation model, pre-trained on generic time series, to enable highly generalizable MedTS classification on unseen datasets. FORMED combines the backbone with a novel classifier comprising two components: (1) task-specific channel embeddings and label queries, dynamically sized to match any number of channels and target classes, and (2) a shared decoding attention layer, jointly trained across datasets to capture medical domain knowledge through task-agnostic feature-query interactions. After repurposing, FORMED achieves seamless adaptation to unseen MedTS datasets through lightweight label query training (0.1% of parameters), eliminating the need for full fine-tuning or architectural redesign. We evaluate FORMED on 5 diverse MedTS datasets, benchmarking against 11 Task-Specific Models (TSM) and 4 Task-Specific Adaptation (TSA) methods. Our results demonstrate FORMED's dominant performance, achieving up to 35% absolute improvement in F1-score (on ADFTD dataset) over specialized baselines. Further analysis reveals consistent generalization across varying channel configurations, time series lengths, and clinical tasks, which are key challenges in real-world deployment. By decoupling domain-invariant representation learning from task-specific adaptation, FORMED establishes a scalable and resource-efficient paradigm for foundation model repurposing in healthcare. This approach prioritizes clinical adaptability over rigid task-centric design, offering a practical pathway for real-world implementation.
Authors: Sijin Chen, Omar Hagrass, Jason M. Klusowski
Abstract: Decoding strategies play a pivotal role in text generation for modern language models, yet a puzzling gap divides theory and practice. Surprisingly, strategies that should intuitively be optimal, such as Maximum a Posteriori (MAP), often perform poorly in practice. Meanwhile, popular heuristic approaches like Top-$k$ and Nucleus sampling, which employ truncation and normalization of the conditional next-token probabilities, have achieved great empirical success but lack theoretical justifications. In this paper, we propose Decoding Game, a comprehensive theoretical framework which reimagines text generation as a two-player zero-sum game between Strategist, who seeks to produce text credible in the true distribution, and Nature, who distorts the true distribution adversarially. After discussing the decomposibility of multi-step generation, we derive the optimal strategy in closed form for one-step Decoding Game. It is shown that the adversarial Nature imposes an implicit regularization on likelihood maximization, and truncation-normalization methods are first-order approximations to the optimal strategy under this regularization. Additionally, by generalizing the objective and parameters of Decoding Game, near-optimal strategies encompass diverse methods such as greedy search, temperature scaling, and hybrids thereof. Numerical experiments are conducted to complement our theoretical analysis.
Authors: Junyi Tao, Xiaoyin Chen, Nelson F. Liu
Abstract: Large language models (LMs) are capable of in-context learning from a few demonstrations (example-label pairs) to solve new tasks during inference. Despite the intuitive importance of high-quality demonstrations, previous work has observed that, in some settings, ICL performance is minimally affected by irrelevant labels (Min et al., 2022). We hypothesize that LMs perform ICL with irrelevant labels via two sequential processes: an inference function that solves the task, followed by a verbalization function that maps the inferred answer to the label space. Importantly, we hypothesize that the inference function is invariant to remappings of the label space (e.g., "true"/"false" to "cat"/"dog"), enabling LMs to share the same inference function across settings with different label words. We empirically validate this hypothesis with controlled layer-wise interchange intervention experiments. Our findings confirm the hypotheses on multiple datasets and tasks (natural language inference, sentiment analysis, and topic classification) and further suggest that the two functions can be localized in specific layers across various open-sourced models, including GEMMA-7B, MISTRAL-7B-V0.3, GEMMA-2-27B, and LLAMA-3.1-70B.
Authors: Hyejun Jeong, Shiqing Ma, Amir Houmansadr
Abstract: Bias in Large Language Models remains a critical concern as these systems are increasingly deployed in high-stakes applications. Yet most fairness evaluations rely on scalar metrics or single-model analysis, overlooking how biases align -- or diverge -- across model families, scales, and tuning strategies. In this work, we reframe bias similarity as a form of functional similarity and evaluate 24 LLMs from four major families on over one million structured prompts spanning four bias dimensions. Our findings uncover that fairness is not strongly determined by model size, architecture, instruction tuning, or openness. Instead, bias behaviors are highly context-dependent and structurally persistent, often resistant to current alignment techniques. Contrary to common assumptions, we find that open-source models frequently match or outperform proprietary models in both fairness and utility. These results call into question the default reliance on proprietary systems and highlight the need for behaviorally grounded, model-specific audits to better understand how bias manifests and endures across the LLM landscape.
Authors: Sajad Movahedi, Antonio Orvieto, Seyed-Mohsen Moosavi-Dezfooli
Abstract: In this paper, we propose the $\textit{geometric invariance hypothesis (GIH)}$, which argues that the input space curvature of a neural network remains invariant under transformation in certain architecture-dependent directions during training. We investigate a simple, non-linear binary classification problem residing on a plane in a high dimensional space and observe that$\unicode{x2014}$unlike MLPs$\unicode{x2014}$ResNets fail to generalize depending on the orientation of the plane. Motivated by this example, we define a neural network's $\textbf{average geometry}$ and $\textbf{average geometry evolution}$ as compact $\textit{architecture-dependent}$ summaries of the model's input-output geometry and its evolution during training. By investigating the average geometry evolution at initialization, we discover that the geometry of a neural network evolves according to the data covariance projected onto its average geometry. This means that the geometry only changes in a subset of the input space when the average geometry is low-rank, such as in ResNets. This causes an architecture-dependent invariance property in the input space curvature, which we dub GIH. Finally, we present extensive experimental results to observe the consequences of GIH and how it relates to generalization in neural networks.
Authors: Cheng Huang, Pan Mu, Cong Bai, Peter AG Watson
Abstract: Precipitation from tropical cyclones (TCs) can cause disasters such as flooding, mudslides, and landslides. Predicting such precipitation in advance is crucial, giving people time to prepare and defend against these precipitation-induced disasters. Developing deep learning (DL) rainfall prediction methods offers a new way to predict potential disasters. However, one problem is that most existing methods suffer from cumulative errors and lack physical consistency. Second, these methods overlook the importance of meteorological factors in TC rainfall and their integration with the numerical weather prediction (NWP) model. Therefore, we propose Tropical Cyclone Precipitation Diffusion (TCP-Diffusion), a multi-modal model for global tropical cyclone precipitation forecasting. It forecasts TC rainfall around the TC center for the next 12 hours at 3 hourly resolution based on past rainfall observations and multi-modal environmental variables. Adjacent residual prediction (ARP) changes the training target from the absolute rainfall value to the rainfall trend and gives our model the ability of rainfall change awareness, reducing cumulative errors and ensuring physical consistency. Considering the influence of TC-related meteorological factors and the useful information from NWP model forecasts, we propose a multi-model framework with specialized encoders to extract richer information from environmental variables and results provided by NWP models. The results of extensive experiments show that our method outperforms other DL methods and the NWP method from the European Centre for Medium-Range Weather Forecasts (ECMWF).
Authors: Takeru Miyato, Sindy L\"owe, Andreas Geiger, Max Welling
Abstract: It has long been known in both neuroscience and AI that ``binding'' between neurons leads to a form of competitive learning where representations are compressed in order to represent more abstract concepts in deeper layers of the network. More recently, it was also hypothesized that dynamic (spatiotemporal) representations play an important role in both neuroscience and AI. Building on these ideas, we introduce Artificial Kuramoto Oscillatory Neurons (AKOrN) as a dynamical alternative to threshold units, which can be combined with arbitrary connectivity designs such as fully connected, convolutional, or attentive mechanisms. Our generalized Kuramoto updates bind neurons together through their synchronization dynamics. We show that this idea provides performance improvements across a wide spectrum of tasks such as unsupervised object discovery, adversarial robustness, calibrated uncertainty quantification, and reasoning. We believe that these empirical results show the importance of rethinking our assumptions at the most basic neuronal level of neural representation, and in particular show the importance of dynamical representations. Code:https://github.com/autonomousvision/akorn Project page:https://takerum.github.io/akorn_project_page/
URLs: https://github.com/autonomousvision/akorn, https://takerum.github.io/akorn_project_page/
Authors: Shiyu Wang, Jiawei Li, Xiaoming Shi, Zhou Ye, Baichuan Mo, Wenze Lin, Shengtong Ju, Zhixuan Chu, Ming Jin
Abstract: Time series analysis plays a critical role in numerous applications, supporting tasks such as forecasting, classification, anomaly detection, and imputation. In this work, we present the time series pattern machine (TSPM), a model designed to excel in a broad range of time series tasks through powerful representation and pattern extraction capabilities. Traditional time series models often struggle to capture universal patterns, limiting their effectiveness across diverse tasks. To address this, we define multiple scales in the time domain and various resolutions in the frequency domain, employing various mixing strategies to extract intricate, task-adaptive time series patterns. Specifically, we introduce a general-purpose TSPM that processes multi-scale time series using (1) multi-resolution time imaging (MRTI), (2) time image decomposition (TID), (3) multi-scale mixing (MCM), and (4) multi-resolution mixing (MRM) to extract comprehensive temporal patterns. MRTI transforms multi-scale time series into multi-resolution time images, capturing patterns across both temporal and frequency domains. TID leverages dual-axis attention to extract seasonal and trend patterns, while MCM hierarchically aggregates these patterns across scales. MRM adaptively integrates all representations across resolutions. This method achieves state-of-the-art performance across 8 time series analytical tasks, consistently surpassing both general-purpose and task-specific models. Our work marks a promising step toward the next generation of TSPMs, paving the way for further advancements in time series analysis.
Authors: Cheng-Han Yeh, Kuanchun Yu, Chun-Shien Lu
Abstract: Deep learning models are known to be vulnerable to adversarial attacks by injecting sophisticated designed perturbations to input data. Training-time defenses still exhibit a significant performance gap between natural accuracy and robust accuracy. In this paper, we investigate a new test-time adversarial defense method via diffusion-based recovery along opposite adversarial paths (OAPs). We present a purifier that can be plugged into a pre-trained model to resist adversarial attacks. Different from prior arts, the key idea is excessive denoising or purification by integrating the opposite adversarial direction with reverse diffusion to push the input image further toward the opposite adversarial direction. For the first time, we also exemplify the pitfall of conducting AutoAttack (Rand) for diffusion-based defense methods. Through the lens of time complexity, we examine the trade-off between the effectiveness of adaptive attack and its computation complexity against our defense. Experimental evaluation along with time cost analysis verifies the effectiveness of the proposed method.
Authors: Xiaohuan Bi, Xi Li
Abstract: Federated Learning (FL) enables decentralized model training while preserving privacy. Recently, the integration of Foundation Models (FMs) into FL has enhanced performance but introduced a novel backdoor attack mechanism. Attackers can exploit FM vulnerabilities to embed backdoors into synthetic data generated by FMs. During global model fusion, these backdoors are transferred to the global model through compromised synthetic data, subsequently infecting all client models. Existing FL backdoor defenses are ineffective against this novel attack due to its fundamentally different mechanism compared to classic ones. In this work, we propose a novel data-free defense strategy that addresses both classic and novel backdoor attacks in FL. The shared attack pattern lies in the abnormal activations within the hidden feature space during model aggregation. Hence, we propose to constrain internal activations to remain within reasonable ranges, effectively mitigating attacks while preserving model functionality. The activation constraints are optimized using synthetic data alongside FL training. Extensive experiments demonstrate its effectiveness against both novel and classic backdoor attacks, outperforming existing defenses.
Authors: Hemanth Saratchandran, Jianqiao Zheng, Yiping Ji, Wenbo Zhang, Simon Lucey
Abstract: This paper questions whether the strong performance of softmax attention in transformers stems from producing a probability distribution over inputs. Instead, we argue that softmax's effectiveness lies in its implicit regularization of the Frobenius norm of the attention matrix, which stabilizes training. Motivated by this, we explore alternative activations, specifically polynomials, that achieve a similar regularization effect. Our theoretical analysis shows that certain polynomials can serve as effective substitutes for softmax, achieving strong performance across transformer applications despite violating softmax's typical properties of positivity, normalization, and sparsity. Extensive experiments support these findings, offering a new perspective on attention mechanisms.
Authors: Kaiyan Zhao, Yiming Wang, Yuyang Chen, Yan Li, Leong Hou U, Xiaoguang Niu
Abstract: Experience replay is widely used to improve learning efficiency in reinforcement learning by leveraging past experiences. However, existing experience replay methods, whether based on uniform or prioritized sampling, often suffer from low efficiency, particularly in real-world scenarios with high-dimensional state spaces. To address this limitation, we propose a novel approach, Efficient Diversity-based Experience Replay (EDER). EDER employs a determinantal point process to model the diversity between samples and prioritizes replay based on the diversity between samples. To further enhance learning efficiency, we incorporate Cholesky decomposition for handling large state spaces in realistic environments. Additionally, rejection sampling is applied to select samples with higher diversity, thereby improving overall learning efficacy. Extensive experiments are conducted on robotic manipulation tasks in MuJoCo, Atari games, and realistic indoor environments in Habitat. The results demonstrate that our approach not only significantly improves learning efficiency but also achieves superior performance in high-dimensional, realistic environments.
Authors: Luis Augenstein, No\'emie Jaquier, Tamim Asfour, Leonel Rozo
Abstract: Probabilistic Latent Variable Models (LVMs) excel at modeling complex, high-dimensional data through lower-dimensional representations. Recent advances show that equipping these latent representations with a Riemannian metric unlocks geometry-aware distances and shortest paths that comply with the underlying data structure. This paper focuses on hyperbolic embeddings, a particularly suitable choice for modeling hierarchical relationships. Previous approaches relying on hyperbolic geodesics for interpolating the latent space often generate paths crossing low-data regions, leading to highly uncertain predictions. Instead, we propose augmenting the hyperbolic manifold with a pullback metric to account for distortions introduced by the LVM's nonlinear mapping and provide a complete development for pullback metrics of Gaussian Process LVMs (GPLVMs). Our experiments demonstrate that geodesics on the pullback metric not only respect the geometry of the hyperbolic latent space but also align with the underlying data distribution, significantly reducing uncertainty in predictions.
Authors: Matthew Landers, Taylor W. Killian, Hugo Barnes, Thomas Hartvigsen, Afsaneh Doryab
Abstract: Offline reinforcement learning in high-dimensional, discrete action spaces is challenging due to the exponential scaling of the joint action space with the number of sub-actions and the complexity of modeling sub-action dependencies. Existing methods either exhaustively evaluate the action space, making them computationally infeasible, or factorize Q-values, failing to represent joint sub-action effects. We propose Branch Value Estimation (BraVE), a value-based method that uses tree-structured action traversal to evaluate a linear number of joint actions while preserving dependency structure. BraVE outperforms prior offline RL methods by up to $20\times$ in environments with over four million actions.
Authors: Doron Haviv, Aram-Alexandre Pooladian, Dana Pe'er, Brandon Amos
Abstract: Generative modeling typically concerns transporting a single source distribution to a target distribution via simple probability flows. However, in fields like computer graphics and single-cell genomics, samples themselves can be viewed as distributions, where standard flow matching ignores their inherent geometry. We propose Wasserstein flow matching (WFM), which lifts flow matching onto families of distributions using the Wasserstein geometry. Notably, WFM is the first algorithm capable of generating distributions in high dimensions, whether represented analytically (as Gaussians) or empirically (as point-clouds). Our theoretical analysis establishes that Wasserstein geodesics constitute proper conditional flows over the space of distributions, making for a valid FM objective. Our algorithm leverages optimal transport theory and the attention mechanism, demonstrating versatility across computational regimes: exploiting closed-form optimal transport paths for Gaussian families, while using entropic estimates on point-clouds for general distributions. WFM successfully generates both 2D & 3D shapes and high-dimensional cellular microenvironments from spatial transcriptomics data. Code is available at https://github.com/DoronHav/WassersteinFlowMatching .
Authors: Md Abrar Jahin, Md. Akmol Masud, M. F. Mridha, Nilanjan Dey, Zeyar Aung
Abstract: In high-energy physics, particle jet tagging plays a pivotal role in distinguishing quark from gluon jets using data from collider experiments. While graph-based deep learning methods have advanced this task beyond traditional feature-engineered approaches, the complex data structure and limited labeled samples present ongoing challenges. However, existing contrastive learning (CL) frameworks struggle to leverage rationale-aware augmentations effectively, often lacking supervision signals that guide the extraction of salient features and facing computational efficiency issues such as high parameter counts. In this study, we demonstrate that integrating a quantum rationale generator (QRG) within our proposed Quantum Rationale-aware Graph Contrastive Learning (QRGCL) framework significantly enhances jet discrimination performance, reducing reliance on labeled data and capturing discriminative features. Evaluated on the quark-gluon jet dataset, QRGCL achieves an AUC score of $77.53\%$ while maintaining a compact architecture of only 45 QRG parameters, outperforming classical, quantum, and hybrid GCL and GNN benchmarks. These results highlight QRGCL's potential to advance jet tagging and other complex classification tasks in high-energy physics, where computational efficiency and feature extraction limitations persist.
Authors: Congyuan Duan, Wanteng Ma, Jiashuo Jiang, Dong Xia
Abstract: This paper investigates regret minimization, statistical inference, and their interplay in high-dimensional online decision-making based on the sparse linear context bandit model. We integrate the $\varepsilon$-greedy bandit algorithm for decision-making with a hard thresholding algorithm for estimating sparse bandit parameters and introduce an inference framework based on a debiasing method using inverse propensity weighting. Under a margin condition, our method achieves either $O(T^{1/2})$ regret or classical $O(T^{1/2})$-consistent inference, indicating an unavoidable trade-off between exploration and exploitation. If a diverse covariate condition holds, we demonstrate that a pure-greedy bandit algorithm, i.e., exploration-free, combined with a debiased estimator based on average weighting can simultaneously achieve optimal $O(\log T)$ regret and $O(T^{1/2})$-consistent inference. We also show that a simple sample mean estimator can provide valid inference for the optimal policy's value. Numerical simulations and experiments on Warfarin dosing data validate the effectiveness of our methods.
Authors: Yadi Cao, Yuxuan Liu, Liu Yang, Rose Yu, Hayden Schaeffer, Stanley Osher
Abstract: In-Context Operator Networks (ICONs) have demonstrated the ability to learn operators across diverse partial differential equations using few-shot, in-context learning. However, existing ICONs process each spatial point as an individual token, severely limiting computational efficiency when handling dense data in higher spatial dimensions. We propose Vision In-Context Operator Networks (VICON), which integrates vision transformer architectures to efficiently process 2D data through patch-wise operations while preserving ICON's adaptability to multiphysics systems and varying timesteps. Evaluated across three fluid dynamics benchmarks, VICON significantly outperforms state-of-the-art baselines: DPOT and MPP, reducing the averaged last-step rollout error by 37.9% compared to DPOT and 44.7% compared to MPP, while requiring only 72.5% and 34.8% of their respective inference times. VICON naturally supports flexible rollout strategies with varying timestep strides, enabling immediate deployment in imperfect measurement systems where sampling frequencies may differ or frames might be dropped - common challenges in real-world settings - without requiring retraining or interpolation. In these realistic scenarios, VICON exhibits remarkable robustness, experiencing only 24.41% relative performance degradation compared to 71.37%-74.49% degradation in baseline methods, demonstrating its versatility for deploying in realistic applications. Our scripts for processing datasets and code are publicly available at https://github.com/Eydcao/VICON.
Authors: Zheng Li, Feng Xie, Xichen Guo, Yan Zeng, Hao Zhang, Zhi Geng
Abstract: Estimating causal effects from nonexperimental data is a fundamental problem in many fields of science. A key component of this task is selecting an appropriate set of covariates for confounding adjustment to avoid bias. Most existing methods for covariate selection often assume the absence of latent variables and rely on learning the global network structure among variables. However, identifying the global structure can be unnecessary and inefficient, especially when our primary interest lies in estimating the effect of a treatment variable on an outcome variable. To address this limitation, we propose a novel local learning approach for covariate selection in nonparametric causal effect estimation, which accounts for the presence of latent variables. Our approach leverages testable independence and dependence relationships among observed variables to identify a valid adjustment set for a target causal relationship, ensuring both soundness and completeness under standard assumptions. We validate the effectiveness of our algorithm through extensive experiments on both synthetic and real-world data.
Authors: Yicheng Zhang, Zhen Qin, Zhaomin Wu, Jian Hou, Shuiguang Deng
Abstract: Large-scale instruction data is essential for aligning pretrained Large Language Models (LLMs) with human instructions, but may contain sensitive information that hinders its public sharing. Federated Learning (FL) enables collaborative fine-tuning of LLMs without accessing raw data. However, existing approaches to federated LLM fine-tuning usually adopt a uniform model architecture, making it hard to fit highly heterogeneous client-side data in varying domains and formats. To address this, we propose FedAMoLE, a lightweight personalized FL framework that enables data-driven heterogeneous model architectures. This framework features a heterogeneous mixture of LoRA experts module for aggregating architecturally heterogeneous models and a reverse selection-based expert assignment strategy that optimizes model architectures based on data distributions. Experiments across five scenarios show that FedAMoLE improves client-side performance by an average of 5.14% compared to existing approaches while maintaining scalability.
Authors: Michael Tschannen, Andr\'e Susano Pinto, Alexander Kolesnikov
Abstract: Removing modeling constraints and unifying architectures across domains has been a key driver of the recent progress in training large multimodal models. However, most of these models still rely on many separately trained components such as modality-specific encoders and decoders. In this work, we further streamline joint generative modeling of images and text. We propose an autoregressive decoder-only transformer - JetFormer - which is trained to directly maximize the likelihood of raw data, without relying on any separately pretrained components, and can understand and generate both text and images. Specifically, we leverage a normalizing flow model to obtain a soft-token image representation that is jointly trained with an autoregressive multimodal transformer. The normalizing flow model serves as both an image encoder for perception tasks and an image decoder for image generation tasks during inference. JetFormer achieves text-to-image generation quality competitive with recent VQ-VAE- and VAE-based baselines. These baselines rely on pretrained image autoencoders, which are trained with a complex mixture of losses, including perceptual ones. At the same time, JetFormer demonstrates robust image understanding capabilities. To the best of our knowledge, JetFormer is the first model that is capable of generating high-fidelity images and producing strong log-likelihood bounds.
Authors: Wei Lin, Qingyu Song, Hong Xu
Abstract: Tuning step sizes is crucial for the stability and efficiency of optimization algorithms. While adaptive coordinate-wise step sizes have been shown to outperform scalar step size in first-order methods, their use in second-order methods is still under-explored and more challenging. Current approaches, including hypergradient descent and cutting plane methods, offer limited improvements or encounter difficulties in second-order contexts. To address these limitations, we first conduct a theoretical analysis within the Broyden-Fletcher-Goldfarb-Shanno (BFGS) framework, a prominent quasi-Newton method, and derive sufficient conditions for coordinate-wise step sizes that ensure convergence and stability. Building on this theoretical foundation, we introduce a novel learn-to-optimize (L2O) method that employs LSTM-based networks to learn optimal step sizes by leveraging insights from past optimization trajectories, while inherently respecting the derived theoretical guarantees. Extensive experiments demonstrate that our approach achieves substantial improvements over scalar step size methods and hypergradient descent-based method, offering up to 4$\times$ faster convergence across diverse optimization tasks.
Authors: Yizuo Chen, Amit Bhatia
Abstract: We introduce a causal modeling framework that captures the input-output behavior of predictive models (e.g., machine learning models). The framework enables us to identify features that directly cause the predictions, which has broad implications for data collection and model evaluation. We then present sound and complete algorithms for discovering direct causes (from data) under some assumptions. Furthermore, we propose a novel independence rule that can be integrated with the algorithms to accelerate the discovery process, as we demonstrate both theoretically and empirically.
Authors: Pritam Anand, Tathagata Bandyopadhyay, Suresh Chandra
Abstract: This paper proposes a novel loss function, called 'Tube Loss', for simultaneous estimation of bounds of a Prediction Interval (PI) in the regression setup. The PIs obtained by minimizing the empirical risk based on the Tube Loss are shown to be of better quality than the PIs obtained by the existing methods in the following sense. First, it yields intervals that attain the prespecified confidence level t $\in$ (0,1) asymptotically. A theoretical proof of this fact is given. Secondly, the user is allowed to move the interval up or down by controlling the value of a parameter. This helps the user to choose a PI capturing denser regions of the probability distribution of the response variable inside the interval, and thus, sharpening its width. This is shown to be especially useful when the conditional distribution of the response variable is skewed. Further, the Tube Loss based PI estimation method can trade-off between the coverage and the average width by solving a single optimization problem. It enables further reduction of the average width of PI through re-calibration. Also, unlike a few existing PI estimation methods the gradient descent (GD) method can be used for minimization of empirical risk. Through extensive experiments, we demonstrate the effectiveness of Tube Loss-based PI estimation in both kernel machines and neural networks. Additionally, we show that Tube Loss-based deep probabilistic forecasting models achieve superior performance compared to existing probabilistic forecasting techniques across several benchmark and wind datasets. Finally, we empirically validate the advantages of the Tube loss approach within the conformal prediction framework. Codes are available at https://github.com/ltpritamanand/Tube$\_$loss.
Authors: Alberto Silvio Chiappa, Briti Gangopadhyay, Zhao Wang, Shingo Takamatsu
Abstract: Online advertising has become one of the most successful business models of the internet era. Impression opportunities are typically allocated through real-time auctions, where advertisers bid to secure advertisement slots. Deciding the best bid for an impression opportunity is challenging, due to the stochastic nature of user behavior and the variability of advertisement traffic over time. In this work, we propose a framework for training auto-bidding agents in multi-slot second-price auctions to maximize acquisitions (e.g., clicks, conversions) while adhering to budget and cost-per-acquisition (CPA) constraints. We exploit the insight that, after an advertisement campaign concludes, determining the optimal bids for each impression opportunity can be framed as a multiple-choice knapsack problem (MCKP) with a nonlinear objective. We propose an "oracle" algorithm that identifies a near-optimal combination of impression opportunities and advertisement slots, considering both past and future advertisement traffic data. This oracle solution serves as a training target for a student network which bids having access only to real-time information, a method we term Oracle Imitation Learning (OIL). Through numerical experiments, we demonstrate that OIL achieves superior performance compared to both online and offline reinforcement learning algorithms, offering improved sample efficiency. Notably, OIL shifts the complexity of training auto-bidding agents from crafting sophisticated learning algorithms to solving a nonlinear constrained optimization problem efficiently.
Authors: Honggu Kang, Seohyeon Cha, Joonhyuk Kang
Abstract: Federated learning (FL) is a distributed training paradigm that enables collaborative learning across clients without sharing local data, thereby preserving privacy. However, the increasing scale and complexity of modern deep models often exceed the computational or memory capabilities of edge devices. Furthermore, clients may be constrained to use heterogeneous model architectures due to hardware variability (e.g., ASICs, FPGAs) or proprietary requirements that prevent the disclosure or modification of local model structures. These practical considerations motivate the need for model-heterogeneous FL, where clients participate using distinct model architectures. In this work, we propose Generative Model-Aided Federated Learning (GeFL), a framework that enables cross-client knowledge sharing via a generative model trained in a federated manner. This generative model captures global data semantics and facilitates local training without requiring model homogeneity across clients. While GeFL achieves strong performance, empirical analysis reveals limitations in scalability and potential privacy leakage due to generative sample memorization. To address these concerns, we propose GeFL-F, which utilizes feature-level generative modeling. This approach enhances scalability to large client populations and mitigates privacy risks. Extensive experiments across image classification tasks demonstrate that both GeFL and GeFL-F offer competitive performance in heterogeneous settings. Code is available at [1].
Authors: Mehrdad Moghimi, Hyejin Ku
Abstract: In domains such as finance, healthcare, and robotics, managing worst-case scenarios is critical, as failure to do so can lead to catastrophic outcomes. Distributional Reinforcement Learning (DRL) provides a natural framework to incorporate risk sensitivity into decision-making processes. However, existing approaches face two key limitations: (1) the use of fixed risk measures at each decision step often results in overly conservative policies, and (2) the interpretation and theoretical properties of the learned policies remain unclear. While optimizing a static risk measure addresses these issues, its use in the DRL framework has been limited to the simple static CVaR risk measure. In this paper, we present a novel DRL algorithm with convergence guarantees that optimizes for a broader class of static Spectral Risk Measures (SRM). Additionally, we provide a clear interpretation of the learned policy by leveraging the distribution of returns in DRL and the decomposition of static coherent risk measures. Extensive experiments demonstrate that our model learns policies aligned with the SRM objective, and outperforms existing risk-neutral and risk-sensitive DRL models in various settings.
Authors: Lucas Prieto, Melih Barsbey, Pedro A. M. Mediano, Tolga Birdal
Abstract: Grokking, the sudden generalization that occurs after prolonged overfitting, is a surprising phenomenon challenging our understanding of deep learning. Although significant progress has been made in understanding grokking, the reasons behind the delayed generalization and its dependence on regularization remain unclear. In this work, we argue that without regularization, grokking tasks push models to the edge of numerical stability, introducing floating point errors in the Softmax function, which we refer to as Softmax Collapse (SC). We demonstrate that SC prevents grokking and that mitigating SC enables grokking without regularization. Investigating the root cause of SC, we find that beyond the point of overfitting, the gradients strongly align with what we call the na\"ive loss minimization (NLM) direction. This component of the gradient does not alter the model's predictions but decreases the loss by scaling the logits, typically by scaling the weights along their current direction. We show that this scaling of the logits explains the delay in generalization characteristic of grokking and eventually leads to SC, halting further learning. To validate our hypotheses, we introduce two key contributions that address the challenges in grokking tasks: StableMax, a new activation function that prevents SC and enables grokking without regularization, and $\perp$Grad, a training algorithm that promotes quick generalization in grokking tasks by preventing NLM altogether. These contributions provide new insights into grokking, elucidating its delayed generalization, reliance on regularization, and the effectiveness of existing grokking-inducing methods. Code for this paper is available at https://github.com/LucasPrietoAl/grokking-at-the-edge-of-numerical-stability.
URLs: https://github.com/LucasPrietoAl/grokking-at-the-edge-of-numerical-stability.
Authors: Shiyun Xu, Zhiqi Bu, Yiliang Zhang, Ian Barnett
Abstract: Differential learning rate (DLR), a technique that applies different learning rates to different model parameters, has been widely used in deep learning and achieved empirical success via its various forms. For example, parameter-efficient fine-tuning (PEFT) applies zero learning rates to most parameters so as to significantly save the computational cost. At the core, DLR leverages the observation that different parameters can have different loss curvature, which is hard to characterize in general. We propose the Hessian-informed differential learning rate (Hi-DLR), an efficient approach that solves the hyperparameter optimization (HPO) of learning rates and captures the loss curvature for any model and optimizer adaptively. Given a proper grouping of parameters, we empirically demonstrate that Hi-DLR can improve the convergence by dynamically determining the learning rates during the training.
Authors: Xueqing Liu, Qinwei Yang, Zhaoqing Tian, Ruocheng Guo, Peng Wu
Abstract: Transfer learning of prediction models has been extensively studied, while the corresponding policy learning approaches are rarely discussed. In this paper, we propose principled approaches for learning the optimal policy in the target domain by leveraging two datasets: one with full information from the source domain and the other from the target domain with only covariates. First, under the setting of covariate shift, we formulate the problem from a perspective of causality and present the identifiability assumptions for the reward induced by a given policy. Then, we derive the efficient influence function and the semiparametric efficiency bound for the reward. Based on this, we construct a doubly robust and semiparametric efficient estimator for the reward and then learn the optimal policy by optimizing the estimated reward. Moreover, we theoretically analyze the bias and the generalization error bound for the learned policy. Extensive experiments demonstrate that the approach not only estimates the reward more accurately but also yields a policy that closely approximates the theoretically optimal policy.
Authors: Menghao Huo, Kuan Lu, Yuxiao Li, Qiang Zhu, Zhenrui Chen
Abstract: Accurately predicting renewable energy output is crucial for the efficient integration of solar and wind power into modern energy systems. This study develops and evaluates an advanced deep learning model, Channel-Time Patch Time-Series Transformer (CT-PatchTST), to forecast the power output of photovoltaic and wind energy systems using annual offshore wind power, onshore wind power, and solar power generation data from Denmark. While the original Patch Time-Series Transformer(PatchTST) model employs a channel-independent (CI) approach, it tends to overlook inter-channel relationships during training, potentially leading to a loss of critical information. To address this limitation and further leverage the benefits of increased data granularity brought by CI, we propose CT-PatchTST. This enhanced model improves the processing of inter-channel information while maintaining the advantages of the channel-independent approach. The predictive performance of CT-PatchTST is rigorously analyzed, demonstrating its ability to provide precise and reliable energy forecasts. This work contributes to improving the predictability of renewable energy systems, supporting their broader adoption and integration into energy grids.
Authors: Yuhan Kang, Yang Shi, Mei We, Jun He, Jianchao Yang, Zeyu Xue, Jing Feng, Xinwang Liu
Abstract: Post-training pruning, as one of the key techniques for compressing large language models, plays a vital role in lightweight model deployment and model sparsity. However, current mainstream pruning methods dependent on the Hessian matrix face significant limitations in both pruning speed and practical effectiveness due to the computationally intensive nature of second-order derivative calculations. This paper presents SwiftPrune, a novel Hessian-free weight pruning method that achieves hardware-efficient model compression through two key innovations: 1) SwiftPrune eliminates the need for computationally intensive Hessian matrix calculations by introducing a contribution-based weight metric, which evaluates the importance of weights without relying on second-order derivatives. 2) we employ the Exponentially Weighted Moving Average (EWMA) technique to bypass weight sorting, enabling the selection of weights that contribute most to LLM accuracy and further reducing time complexity. Our approach is extended to support structured sparsity pruning, facilitating efficient execution on modern hardware accelerators. We validate the SwiftPrune on three LLMs (namely LLaMA2, LLaMA3, and Pythia), demonstrating that it significantly enhances compression performance. The experimental findings reveal that SwiftPrune completes the pruning process within seconds, achieving an average speedup of 12.29x (up to 56.02x) over existing SOTA approaches.
Authors: Qingyu Song, Wei Lin, Hong Xu
Abstract: Learn to Optimize (L2O) trains deep neural network based solvers for optimization, achieving success in accelerating convex problems and improving non-convex solutions. However, L2O lacks rigorous theoretical backing for its own training convergence, as existing analyses often use unrealistic assumptions -- a gap this work highlights empirically. We bridge this gap by proving the training convergence of L2O models that learn Gradient Descent (GD) hyperparameters for quadratic programming, leveraging the Neural Tangent Kernel (NTK) theory. We propose a deterministic initialization strategy to support our theoretical results and promote stable training over extended optimization horizons by mitigating gradient explosion. Our L2O framework demonstrates over 50\% better optimality against GD and superior robustness over state-of-the-art L2O methods on synthetic datasets.
Authors: Tyler Ingebrand, Adam J. Thorpe, Ufuk Topcu
Abstract: A central challenge in transfer learning is designing algorithms that can quickly adapt and generalize to new tasks without retraining. Yet, the conditions of when and how algorithms can effectively transfer to new tasks is poorly characterized. We introduce a geometric characterization of transfer in Hilbert spaces and define three types of inductive transfer: interpolation within the convex hull, extrapolation to the linear span, and extrapolation outside the span. We propose a method grounded in the theory of function encoders to achieve all three types of transfer. Specifically, we introduce a novel training scheme for function encoders using least-squares optimization, prove a universal approximation theorem for function encoders, and provide a comprehensive comparison with existing approaches such as transformers and meta-learning on four diverse benchmarks. Our experiments demonstrate that the function encoder outperforms state-of-the-art methods on four benchmark tasks and on all three types of transfer.
Authors: Keqin Wang, Yulong Yang, Ishan Saha, Christine Allen-Blanchette
Abstract: While graph neural networks (GNNs) have allowed researchers to successfully apply neural networks to non-Euclidean domains, deep GNNs often exhibit lower predictive performance than their shallow counterparts. This phenomena has been attributed in part to oversmoothing, the tendency of node representations to become increasingly similar with network depth. In this paper we introduce an analogy between oversmoothing in GNNs and consensus (i.e., perfect agreement) in the opinion dynamics literature. We show that the message passing algorithms of several GNN models are equivalent to linear opinion dynamics models which have been shown to converge to consensus for all inputs regardless of the graph structure. This new perspective on oversmoothing motivates the use of nonlinear opinion dynamics as an inductive bias in GNN models. In our Behavior-Inspired Message Passing (BIMP) GNN, we leverage the nonlinear opinion dynamics model which is more general than the linear opinion dynamics model, and can be designed to converge to dissensus for general inputs. Through extensive experiments we show that BIMP resists oversmoothing beyond 100 time steps and consistently outperforms existing architectures even when those architectures are amended with oversmoothing mitigation techniques. We also show that BIMP has several desirable properties including well behaved gradients and adaptability to homophilic and heterophilic datasets.
Authors: Hong Huang, Hai Yang, Yuan Chen, Jiaxun Ye, Dapeng Wu
Abstract: Federated Learning (FL) enables collaborative model training across distributed clients without data sharing, but its high computational and communication demands strain resource-constrained devices. While existing methods use dynamic pruning to improve efficiency by periodically adjusting sparse model topologies while maintaining sparsity, these approaches suffer from issues such as greedy adjustments, unstable topologies, and communication inefficiency, resulting in less robust models and suboptimal performance under data heterogeneity and partial client availability. To address these challenges, we propose Federated Robust pruning via combinatorial Thompson Sampling (FedRTS), a novel framework designed to develop robust sparse models. FedRTS enhances robustness and performance through its Thompson Sampling-based Adjustment (TSAdj) mechanism, which uses probabilistic decisions informed by stable, farsighted information instead of deterministic decisions reliant on unstable and myopic information in previous methods. Extensive experiments demonstrate that FedRTS achieves state-of-the-art performance in computer vision and natural language processing tasks while reducing communication costs, particularly excelling in scenarios with heterogeneous data distributions and partial client participation. Our codes are available at: https://github.com/Little0o0/FedRTS
Authors: Morgan Thomas, Albert Bou, Gianni De Fabritiis
Abstract: Chemical Language Models (CLMs) leveraging reinforcement learning (RL) have shown promise in de novo molecular design, yet often suffer from mode collapse, limiting their exploration capabilities. Inspired by Test-Time Training (TTT) in large language models, we propose scaling TTT for CLMs to enhance chemical space exploration. We introduce MolExp, a novel benchmark emphasizing the discovery of structurally diverse molecules with similar bioactivity, simulating real-world drug design challenges. Our results demonstrate that scaling TTT by increasing the number of independent RL agents follows a log-linear scaling law, significantly improving exploration efficiency as measured by MolExp. In contrast, increasing TTT training time yields diminishing returns, even with exploration bonuses. We further evaluate cooperative RL strategies to enhance exploration efficiency. These findings provide a scalable framework for generative molecular design, offering insights into optimizing AI-driven drug discovery.
Authors: Wenzhi Fang, Dong-Jun Han, Liangqi Yuan, Seyyedali Hosseinalipour, Christopher G. Brinton
Abstract: Fine-tuning large language models (LLMs) on devices remains a challenging problem. Recent works have fused low-rank adaptation (LoRA) techniques with federated fine-tuning to mitigate challenges associated with device model sizes and data scarcity. Still, the heterogeneity of resources remains a critical bottleneck: while higher-rank modules generally enhance performance, varying device capabilities constrain LoRA's feasible rank range. Existing approaches attempting to resolve this issue either lack analytical justification or impose additional computational overhead, leaving a wide gap for efficient and theoretically-grounded solutions. To address these challenges, we propose federated sketching LoRA (FSLoRA), which leverages a sketching mechanism to enable devices to selectively update submatrices of global LoRA modules maintained by the server. By adjusting the sketching ratios, which determine the ranks of the submatrices on the devices, FSLoRA flexibly adapts to device-specific communication and computational constraints. We provide a rigorous convergence analysis of FSLoRA that characterizes how the sketching ratios affect the convergence rate. Through comprehensive experiments on multiple datasets and LLM models, we demonstrate FSLoRA's performance improvements compared to various baselines. The code is available at https://github.com/wenzhifang/Federated-Sketching-LoRA-Implementation.
URLs: https://github.com/wenzhifang/Federated-Sketching-LoRA-Implementation.
Authors: Zhiliang Chen, Gregory Kang Ruey Lau, Chuan-Sheng Foo, Bryan Kian Hsiang Low
Abstract: The performance of an LLM depends heavily on the relevance of its training data to the downstream evaluation task. However, in practice, the data involved in an unseen evaluation task is often unknown (e.g., conversations between an LLM and a user are end-to-end encrypted). Hence, it is unclear what data are relevant for fine-tuning the LLM to maximize its performance on the specific unseen evaluation task. Instead, one can only deploy the LLM on the unseen task to gather multiple rounds of feedback on how well the model performs (e.g., user ratings). This novel setting offers a refreshing perspective towards optimizing training data mixtures via feedback from an unseen evaluation task, which prior data mixing and selection works do not consider. Our paper presents DUET, a novel global-to-local algorithm that interleaves influence function as a data selection method with Bayesian optimization to optimize data mixture via feedback from a specific unseen evaluation task. By analyzing DUET's cumulative regret, we theoretically show that DUET converges to the optimal training data mixture for an unseen task even without any data knowledge of the task. Finally, our experiments across a variety of language tasks demonstrate that DUET outperforms existing data selection and mixing methods in the unseen-task setting.
Authors: Haixu Wu, Yuezhou Ma, Hang Zhou, Huikun Weng, Jianmin Wang, Mingsheng Long
Abstract: Physics-informed neural networks (PINNs) have earned high expectations in solving partial differential equations (PDEs), but their optimization usually faces thorny challenges due to the unique derivative-dependent loss function. By analyzing the loss distribution, previous research observed the propagation failure phenomenon of PINNs, intuitively described as the correct supervision for model outputs cannot ''propagate'' from initial states or boundaries to the interior domain. Going beyond intuitive understanding, this paper provides a formal and in-depth study of propagation failure and its root cause. Based on a detailed comparison with classical finite element methods, we ascribe the failure to the conventional single-point-processing architecture of PINNs and further prove that propagation failure is essentially caused by the lower gradient correlation of PINN models on nearby collocation points. Compared to superficial loss maps, this new perspective provides a more precise quantitative criterion to identify where and why PINN fails. The theoretical finding also inspires us to present a new PINN architecture, named ProPINN, which can effectively unite the gradients of region points for better propagation. ProPINN can reliably resolve PINN failure modes and significantly surpass advanced Transformer-based models with 46% relative promotion.
Authors: Jianfeng Cai, Jinhua Zhu, Ruopei Sun, Yue Wang, Li Li, Wengang Zhou, Houqiang Li
Abstract: Reinforcement Learning from Human Feedback (RLHF) has achieved considerable success in aligning large language models (LLMs) by modeling human preferences with a learnable reward model and employing a reinforcement learning algorithm to maximize the reward model's scores. However, these reward models are susceptible to exploitation through various superficial confounding factors, with length bias emerging as a particularly significant concern. Moreover, while the pronounced impact of length bias on preference modeling suggests that LLMs possess an inherent sensitivity to length perception, our preliminary investigations reveal that fine-tuned LLMs consistently struggle to adhere to explicit length instructions. To address these two limitations, we propose a novel framework wherein the reward model explicitly differentiates between human semantic preferences and response length requirements. Specifically, we introduce a $\textbf{R}$esponse-$\textbf{c}$onditioned $\textbf{B}$radley-$\textbf{T}$erry (Rc-BT) model that enhances the model's capability in length bias mitigating and length instruction following, through training on our augmented dataset. Furthermore, we propose the Rc-RM and Rc-DPO algorithm to leverage the Rc-BT model for reward modeling and direct policy optimization (DPO) of LLMs, simultaneously mitigating length bias and promoting adherence to length instructions. Extensive experiments across various foundational models and datasets demonstrate the effectiveness and generalizability of our approach.
Authors: Milena Rmus, Akshay K. Jagadish, Marvin Mathony, Tobias Ludwig, Eric Schulz
Abstract: Computational cognitive models, which formalize theories of cognition, enable researchers to quantify cognitive processes and arbitrate between competing theories by fitting models to behavioral data. Traditionally, these models are handcrafted, which requires significant domain knowledge, coding expertise, and time investment. However, recent advances in machine learning offer solutions to these challenges. In particular, Large Language Models (LLMs) have demonstrated remarkable capabilities for in-context pattern recognition, leveraging knowledge from diverse domains to solve complex problems, and generating executable code that can be used to facilitate the generation of cognitive models. Building on this potential, we introduce a pipeline for Guided generation of Computational Cognitive Models (GeCCo). Given task instructions, participant data, and a template function, GeCCo prompts an LLM to propose candidate models, fits proposals to held-out data, and iteratively refines them based on feedback constructed from their predictive performance. We benchmark this approach across four different cognitive domains -- decision making, learning, planning, and memory -- using three open-source LLMs, spanning different model sizes, capacities, and families. On four human behavioral data sets, the LLM generated models that consistently matched or outperformed the best domain-specific models from the cognitive science literature. Taken together, our results suggest that LLMs can generate cognitive models with conceptually plausible theories that rival -- or even surpass -- the best models from the literature across diverse task domains.
Authors: Yunchuan Guan, Yu Liu, Ke Zhou, Zhiqi Shen, Jenq-Neng Hwang, Lei Li
Abstract: Diffusion-based algorithms have emerged as promising techniques for weight generation. However, existing solutions are limited by two challenges: generalizability and local target assignment. The former arises from the inherent lack of cross-task transferability in existing single-level optimization methods, limiting the model's performance on new tasks. The latter lies in existing research modeling only global optimal weights, neglecting the supervision signals in local target weights. Moreover, naively assigning local target weights causes local-global inconsistency. To address these issues, we propose Mc-Di, which integrates the diffusion algorithm with meta-learning for better generalizability. Furthermore, we extend the vanilla diffusion into a local consistency diffusion algorithm. Our theory and experiments demonstrate that it can learn from local targets while maintaining consistency with the global optima. We validate Mc-Di's superior accuracy and inference efficiency in tasks that require frequent weight updates, including transfer learning, few-shot learning, domain generalization, and large language model adaptation.
Authors: Arya Honarpisheh, Mustafa Bozdag, Octavia Camps, Mario Sznaier
Abstract: State-space models (SSMs) have recently emerged as a compelling alternative to Transformers for sequence modeling tasks. This paper presents a theoretical generalization analysis of selective SSMs, the core architectural component behind the Mamba model. We derive a novel covering number-based generalization bound for selective SSMs, building upon recent theoretical advances in the analysis of Transformer models. Using this result, we analyze how the spectral abscissa of the continuous-time state matrix governs the model's training dynamics and its ability to generalize across sequence lengths. We empirically validate our findings on a synthetic majority task and the IMDb sentiment classification benchmark, illustrating how our theoretical insights translate into practical model behavior.
Authors: Jingzhe Shi, Qinwei Ma, Hongyi Liu, Hang Zhao, Jeng-Neng Hwang, Lei Li
Abstract: Long Context Language Models have drawn great attention in the past few years. There has been work discussing the impact of long context on Language Model performance: some find that long irrelevant context could harm performance, while some experimentally summarize loss reduction by relevant long context as Scaling Laws. This calls for a more thorough understanding on how long context impacts Language Modeling. In this work, we (1) propose a clean and effective theoretical framework for explaining the impact of context length on Language Modeling, from an Intrinsic Space perspective; and (2) conduct experiments on natural language and synthetic data, validating our proposed theoretical assumptions and deductions. Our theoretical framework can provide practical insights such as establishing that training dataset size dictates an optimal context length and bounds context length scaling for certain cases. We hope our work may inspire new long context Language Models, as well as future work studying Physics for Language Models. Code for our experiments is available at: https://github.com/JingzheShi/NLPCtlScalingAndBounds.
Authors: Peter Chen, Yue Xie, Qingpeng Zhang
Abstract: Optimal transport (OT) aims to find a map $T$ that transports mass from one probability measure to another while minimizing a cost function. Recently, neural OT solvers have gained popularity in high dimensional biological applications such as drug perturbation, due to their superior computational and memory efficiency compared to traditional exact Sinkhorn solvers. However, the overly complex high dimensional maps learned by neural OT solvers often suffer from poor interpretability. Prior work addressed this issue in the context of exact OT solvers by introducing \emph{displacement-sparse maps} via designed elastic cost, but such method failed to be applied to neural OT settings. In this work, we propose an intuitive and theoretically grounded approach to learning \emph{displacement-sparse maps} within neural OT solvers. Building on our new formulation, we introduce a novel smoothed $\ell_0$ regularizer that outperforms the $\ell_1$ based alternative from prior work. Leveraging Input Convex Neural Network's flexibility, we further develop a heuristic framework for adaptively controlling sparsity intensity, an approach uniquely enabled by the neural OT paradigm. We demonstrate the necessity of this adaptive framework in large-scale, high-dimensional training, showing not only improved accuracy but also practical ease of use for downstream applications.
Authors: Evan Chen, Shiqiang Wang, Jianing Zhang, Dong-Jun Han, Chaoyue Liu, Christopher Brinton
Abstract: In federated learning (FL), model training performance is strongly impacted by data heterogeneity across clients. Client-drift compensation methods have recently emerged as a solution to this issue, introducing correction terms into local model updates. To date, these methods have only been considered under stochastic gradient descent (SGD)-based model training, while modern FL frameworks also employ adaptive optimizers (e.g., Adam) for improved convergence. However, due to the complex interplay between first and second moments found in most adaptive optimization methods, naively injecting correction terms can lead to performance degradation in heterogeneous settings. In this work, we propose {\tt FAdamGC}, the first algorithm to integrate drift compensation into adaptive federated optimization. The key idea of {\tt FAdamGC} is injecting a pre-estimation correction term that aligns with the moment structure of adaptive methods. We provide a rigorous convergence analysis of our algorithm under non-convex settings, showing that {\tt FAdamGC} results in better rate and milder assumptions than naively porting SGD-based correction algorithms into adaptive optimizers. Our experimental results demonstrate that {\tt FAdamGC} consistently outperform existing methods in total communication and computation cost across varying levels of data heterogeneity, showing the efficacy of correcting gradient information in federated adaptive optimization.
Authors: Ram\'on Calvo Gonz\'alez, Daniele Paliotta, Matteo Pagliardini, Martin Jaggi, Fran\c{c}ois Fleuret
Abstract: Large Language Models (LLMs) demonstrate remarkable capabilities at the cost of high compute requirements. Recent studies have demonstrated that intermediate layers in LLMs can be removed or reordered without substantial accuracy loss; however, this insight has not yet been exploited to improve inference efficiency. Leveraging observed layer independence, we propose a novel method that groups consecutive layers into pairs evaluated in parallel, effectively restructuring the computational graph to enhance parallelism. Without requiring retraining or fine-tuning, this approach achieves an inference throughput improvement of 1.05x-1.20x on standard benchmarks, retaining 95\%-99\% of the original model accuracy. Empirical results demonstrate the practicality of this method in significantly reducing inference cost for large-scale LLM deployment. Additionally, we demonstrate that modest performance degradation can be substantially mitigated through lightweight fine-tuning, further enhancing the method's applicability.
Authors: Runqi Lin, Bo Han, Fengwang Li, Tongling Liu
Abstract: Jailbreaking attacks can effectively manipulate open-source large language models (LLMs) to produce harmful responses. However, these attacks exhibit limited transferability, failing to disrupt proprietary LLMs consistently. To reliably identify vulnerabilities in proprietary LLMs, this work investigates the transferability of jailbreaking attacks by analysing their impact on the model's intent perception. By incorporating adversarial sequences, these attacks can redirect the source LLM's focus away from malicious-intent tokens in the original input, thereby obstructing the model's intent recognition and eliciting harmful responses. Nevertheless, these adversarial sequences fail to mislead the target LLM's intent perception, allowing the target LLM to refocus on malicious-intent tokens and abstain from responding. Our analysis further reveals the inherent distributional dependency within the generated adversarial sequences, whose effectiveness stems from overfitting the source LLM's parameters, resulting in limited transferability to target LLMs. To this end, we propose the Perceived-importance Flatten (PiF) method, which uniformly disperses the model's focus across neutral-intent tokens in the original input, thus obscuring malicious-intent tokens without relying on overfitted adversarial sequences. Extensive experiments demonstrate that PiF provides an effective and efficient red-teaming evaluation for proprietary LLMs.
Authors: Mohamad Louai Shehab, Antoine Aspeel, Necmiye Ozay
Abstract: Inverse reinforcement learning is the problem of inferring a reward function from an optimal policy or demonstrations by an expert. In this work, it is assumed that the reward is expressed as a reward machine whose transitions depend on atomic propositions associated with the state of a Markov Decision Process (MDP). Our goal is to identify the true reward machine using finite information. To this end, we first introduce the notion of a prefix tree policy which associates a distribution of actions to each state of the MDP and each attainable finite sequence of atomic propositions. Then, we characterize an equivalence class of reward machines that can be identified given the prefix tree policy. Finally, we propose a SAT-based algorithm that uses information extracted from the prefix tree policy to solve for a reward machine. It is proved that if the prefix tree policy is known up to a sufficient (but finite) depth, our algorithm recovers the exact reward machine up to the equivalence class. This sufficient depth is derived as a function of the number of MDP states and (an upper bound on) the number of states of the reward machine. These results are further extended to the case where we only have access to demonstrations from an optimal policy. Several examples, including discrete grid and block worlds, a continuous state-space robotic arm, and real data from experiments with mice, are used to demonstrate the effectiveness and generality of the approach.
Authors: Piersilvio De Bartolomeis, Javier Abad, Guanbo Wang, Konstantin Donhauser, Raymond M. Duch, Fanny Yang, Issa J. Dahabreh
Abstract: Randomized experiments are the preferred approach for evaluating the effects of interventions, but they are costly and often yield estimates with substantial uncertainty. On the other hand, in silico experiments leveraging foundation models offer a cost-effective alternative that can potentially attain higher statistical precision. However, the benefits of in silico experiments come with a significant risk: statistical inferences are not valid if the models fail to accurately predict experimental responses to interventions. In this paper, we propose a novel approach that integrates the predictions from multiple foundation models with experimental data while preserving valid statistical inference. Our estimator is consistent and asymptotically normal, with asymptotic variance no larger than the standard estimator based on experimental data alone. Importantly, these statistical properties hold even when model predictions are arbitrarily biased. Empirical results across several randomized experiments show that our estimator offers substantial precision gains, equivalent to a reduction of up to 20% in the sample size needed to match the same precision as the standard estimator based on experimental data alone.
Authors: Chen Fan, Mark Schmidt, Christos Thrampoulidis
Abstract: Different gradient-based methods for optimizing overparameterized models can all achieve zero training error yet converge to distinctly different solutions inducing different generalization properties. We provide the first complete characterization of implicit optimization bias for p-norm normalized steepest descent (NSD) and momentum steepest descent (NMD) algorithms in multi-class linear classification with cross-entropy loss. Our key theoretical contribution is proving that these algorithms converge to solutions maximizing the margin with respect to the classifier matrix's p-norm, with established convergence rates. These results encompass important special cases including Spectral Descent and Muon, which we show converge to max-margin solutions with respect to the spectral norm. A key insight of our contribution is that the analysis of general entry-wise and Schatten p-norms can be reduced to the analysis of NSD/NMD with max-norm by exploiting a natural ordering property between all p-norms relative to the max-norm and its dual sum-norm. For the specific case of descent with respect to the max-norm, we further extend our analysis to include preconditioning, showing that Adam converges to the matrix's max-norm solution. Our results demonstrate that the multi-class linear setting, which is inherently richer than the binary counterpart, provides the most transparent framework for studying implicit biases of matrix-parameter optimization algorithms.
Authors: Zinan Lin, Tadas Baltrusaitis, Wenyu Wang, Sergey Yekhanin
Abstract: Differentially private (DP) synthetic data, which closely resembles the original private data while maintaining strong privacy guarantees, has become a key tool for unlocking the value of private data without compromising privacy. Recently, Private Evolution (PE) has emerged as a promising method for generating DP synthetic data. Unlike other training-based approaches, PE only requires access to inference APIs from foundation models, enabling it to harness the power of state-of-the-art (SoTA) models. However, a suitable foundation model for a specific private data domain is not always available. In this paper, we discover that the PE framework is sufficiently general to allow APIs beyond foundation models. In particular, we demonstrate that many SoTA data synthesizers that do not rely on neural networks--such as computer graphics-based image generators, which we refer to as simulators--can be effectively integrated into PE. This insight significantly broadens PE's applicability and unlocks the potential of powerful simulators for DP data synthesis. We explore this approach, named Sim-PE, in the context of image synthesis. Across four diverse simulators, Sim-PE performs well, improving the downstream classification accuracy of PE by up to 3x, reducing FID by up to 80%, and offering much greater efficiency. We also show that simulators and foundation models can be easily leveraged together within PE to achieve further improvements. The code is open-sourced in the Private Evolution Python library: https://github.com/microsoft/DPSDA.
Authors: Xiangcheng Zhang, Yige Hong, Weina Wang
Abstract: Heterogeneity poses a fundamental challenge for many real-world large-scale decision-making problems but remains largely understudied. In this paper, we study the fully heterogeneous setting of a prominent class of such problems, known as weakly-coupled Markov decision processes (WCMDPs). Each WCMDP consists of $N$ arms (or subproblems), which have distinct model parameters in the fully heterogeneous setting, leading to the curse of dimensionality when $N$ is large. We show that, under mild assumptions, an efficiently computable policy achieves an $O(1/\sqrt{N})$ optimality gap in the long-run average reward per arm for fully heterogeneous WCMDPs as $N$ becomes large. This is the first asymptotic optimality result for fully heterogeneous average-reward WCMDPs. Our main technical innovation is the construction of projection-based Lyapunov functions that certify the convergence of rewards and costs to an optimal region, even under full heterogeneity.
Authors: Guanglong Sun, Hongwei Yan, Liyuan Wang, Qian Li, Bo Lei, Yi Zhong
Abstract: Knowledge distillation (KD) is a powerful strategy for training deep neural networks (DNNs). Although it was originally proposed to train a more compact "student" model from a large "teacher" model, many recent efforts have focused on adapting it to promote generalization of the model itself, such as online KD and self KD. Here, we propose an accessible and compatible strategy named Spaced KD to improve the effectiveness of both online KD and self KD, in which the student model distills knowledge from a teacher model trained with a space interval ahead. This strategy is inspired by a prominent theory named spacing effect in biological learning and memory, positing that appropriate intervals between learning trials can significantly enhance learning performance. With both theoretical and empirical analyses, we demonstrate that the benefits of the proposed Spaced KD stem from convergence to a flatter loss landscape during stochastic gradient descent (SGD). We perform extensive experiments to validate the effectiveness of Spaced KD in improving the learning performance of DNNs (e.g., the performance gain is up to 2.31% and 3.34% on Tiny-ImageNet over online KD and self KD, respectively). Our codes have been released on github https://github.com/SunGL001/Spaced-KD.
Authors: Won-Jun Jang, Hyeon-Seo Park, Si-Hyeon Lee
Abstract: Federated ensemble distillation addresses client heterogeneity by generating pseudo-labels for an unlabeled server dataset based on client predictions and training the server model using the pseudo-labeled dataset. The unlabeled server dataset can either be pre-existing or generated through a data-free approach. The effectiveness of this approach critically depends on the method of assigning weights to client predictions when creating pseudo-labels, especially in highly heterogeneous settings. Inspired by theoretical results from GANs, we propose a provably near-optimal weighting method that leverages client discriminators trained with a server-distributed generator and local datasets. Our experiments on various image classification tasks demonstrate that the proposed method significantly outperforms baselines. Furthermore, we show that the additional communication cost, client-side privacy leakage, and client-side computational overhead introduced by our method are negligible, both in scenarios with and without a pre-existing server dataset.
Authors: Qian Shao, Bang Du, Zepeng Li, Qiyuan Chen, Hongxia Xu, Jimeng Sun, Jian Wu, Jintai Chen
Abstract: Clinical trials remain critical in cardiac drug development but face high failure rates due to efficacy limitations and safety risks, incurring substantial costs. In-silico trial methodologies, particularly generative models simulating drug-induced electrocardiogram (ECG) alterations, offer a potential solution to mitigate these challenges. While existing models show progress in ECG synthesis, their constrained fidelity and inability to characterize individual-specific pharmacological response patterns fundamentally limit clinical translatability. To address these issues, we propose a novel Drug-Aware Diffusion Model (DADM). Specifically, we construct a set of ordinary differential equations to provide external physical knowledge (EPK) of the realistic ECG morphology. The EPK is used to adaptively constrain the morphology of the generated ECGs through a dynamic cross-attention (DCA) mechanism. Furthermore, we propose an extension of ControlNet to incorporate demographic and drug data, simulating individual drug reactions. Compared to the other eight state-of-the-art (SOTA) ECG generative models: 1) Quantitative and expert evaluation demonstrate that DADM generates ECGs with superior fidelity; 2) Comparative results on two real-world databases covering 8 types of drug regimens verify that DADM can more accurately simulate drug-induced changes in ECGs, improving the accuracy by at least 5.79% and recall by 8%. In addition, the ECGs generated by DADM can also enhance model performance in downstream drug-effect classification tasks.
Authors: Marten Lienen, Marcel Kollovieh, Stephan G\"unnemann
Abstract: We derive a novel generative model from iterative Gaussian posterior inference. By treating the generated sample as an unknown variable, we can formulate the sampling process in the language of Bayesian probability. Our model uses a sequence of prediction and posterior update steps to iteratively narrow down the unknown sample starting from a broad initial belief. In addition to a rigorous theoretical analysis, we establish a connection between our model and diffusion models and show that it includes Bayesian Flow Networks (BFNs) as a special case. In our experiments, we demonstrate that our model improves sample quality on ImageNet32 over both BFNs and the closely related Variational Diffusion Models, while achieving equivalent log-likelihoods on ImageNet32 and CIFAR10. Find our code at https://github.com/martenlienen/bsi.
Authors: Xiaoyu Yang, Jie Lu, En Yu
Abstract: The remarkable success of large-scale contrastive pre-training, fueled by vast and curated datasets, is encountering new frontiers as the scaling paradigm evolves. A critical emerging challenge is the effective pre-training of models on dynamic data streams characterized by concept drift, unpredictable changes in the underlying data distribution. This paper undertakes a foundational investigation of this issue. We first reveal that conventional contrastive pre-training methods are notably vulnerable to concept drift, leading to significant biases in the learned feature space of pre-trained models. To systematically analyze these effects, we construct a structural causal model that elucidates how drift acts as a confounder, distorting learned representations. Based on these causal insights, we propose Resilient Contrastive Pre-training (RCP), a novel method incorporating causal intervention. RCP introduces a causally-informed objective designed to mitigate drift-induced biases by leveraging targeted interventions. RCP is designed for simple and scalable implementation and exhibits notable adaptability, promoting robust pre-training on evolving data. Comprehensive experiments across diverse downstream tasks compellingly demonstrate that RCP effectively alleviates the detrimental impact of concept drift, yielding more resilient and generalizable representations.
Authors: Abdellah Zakaria Sellam, Ilyes Benaissa, Abdelmalik Taleb-Ahmed, Luigi Patrono, Cosimo Distante
Abstract: Anomaly detection in time series is essential for industrial monitoring and environmental sensing, yet distinguishing anomalies from complex patterns remains challenging. Existing methods like the Anomaly Transformer and DCdetector have progressed, but they face limitations such as sensitivity to short-term contexts and inefficiency in noisy, non-stationary environments. To overcome these issues, we introduce MAAT, an improved architecture that enhances association discrepancy modeling and reconstruction quality. MAAT features Sparse Attention, efficiently capturing long-range dependencies by focusing on relevant time steps, thereby reducing computational redundancy. Additionally, a Mamba-Selective State Space Model is incorporated into the reconstruction module, utilizing a skip connection and Gated Attention to improve anomaly localization and detection performance. Extensive experiments show that MAAT significantly outperforms previous methods, achieving better anomaly distinguishability and generalization across various time series applications, setting a new standard for unsupervised time series anomaly detection in real-world scenarios.
Authors: Zander W. Blasingame, Chen Liu
Abstract: Training-free guided generation is a widely used and powerful technique that allows the end user to exert further control over the generative process of flow/diffusion models. Generally speaking, two families of techniques have emerged for solving this problem for gradient-based guidance: namely, posterior guidance (i.e., guidance via projecting the current sample to the target distribution via the target prediction model) and end-to-end guidance (i.e., guidance by performing backpropagation throughout the entire ODE solve). In this work, we show that these two seemingly separate families can actually be unified by looking at posterior guidance as a greedy strategy of end-to-end guidance. We explore the theoretical connections between these two families and provide an in-depth theoretical of these two techniques relative to the continuous ideal gradients. Motivated by this analysis we then show a method for interpolating between these two families enabling a trade-off between compute and accuracy of the guidance gradients. We then validate this work on several inverse image problems and property-guided molecular generation.
Authors: Sungyoung Lee, Ziyi Wang, Seunggeun Kim, Taekyun Lee, Yao Lai, David Z. Pan
Abstract: Pretraining models with unsupervised graph representation learning has led to significant advancements in domains such as social network analysis, molecular design, and electronic design automation (EDA). However, prior work in EDA has mainly focused on pretraining models for digital circuits, overlooking analog and mixed-signal circuits. To bridge this gap, we introduce DICE, a Device-level Integrated Circuits Encoder, which is the first graph neural network (GNN) pretrained via self-supervised learning specifically tailored for graph-level prediction tasks in both analog and digital circuits. DICE adopts a simulation-free pretraining approach based on graph contrastive learning, leveraging two novel graph augmentation techniques. Experimental results demonstrate substantial performance improvements across three downstream tasks, highlighting the effectiveness of DICE for both analog and digital circuits. The code is available at github.com/brianlsy98/DICE.
Authors: Harsh Poonia, Felix Divo, Kristian Kersting, Devendra Singh Dhami
Abstract: Causality in time series can be difficult to determine, especially in the presence of non-linear dependencies. The concept of Granger causality helps analyze potential relationships between variables, thereby offering a method to determine whether one time series can predict - Granger cause - future values of another. Although successful, Granger causal methods still struggle with capturing long-range relations between variables. To this end, we leverage the recently successful Extended Long Short-Term Memory (xLSTM) architecture and propose Granger causal xLSTMs (GC-xLSTM). It first enforces sparsity between the time series components by using a novel dynamic loss penalty on the initial projection. Specifically, we adaptively improve the model and identify sparsity candidates. Our joint optimization procedure then ensures that the Granger causal relations are recovered robustly. Our experimental evaluation on six diverse datasets demonstrates the overall efficacy of our proposed GC-xLSTM model.
Authors: Zhi Sheng, Yuan Yuan, Yudi Zhang, Depeng Jin, Yong Li
Abstract: Probabilistic forecasting is crucial for real-world spatiotemporal systems, such as climate, energy, and urban environments, where quantifying uncertainty is essential for informed, risk-aware decision-making. While diffusion models have shown promise in capturing complex data distributions, their application to spatiotemporal forecasting remains limited due to complex spatiotemporal dynamics and high computational demands. In this work, we propose CoST, a novel framework that collaborates deterministic and diffusion models for spatiotemporal forecasting. CoST formulates a mean-residual decomposition strategy: it leverages a powerful deterministic model to capture the conditional mean and a lightweight diffusion model to learn residual uncertainties. This collaborative formulation simplifies learning objectives, enhances forecasting accuracy, enables uncertainty quantification, and significantly improves computational efficiency. To address spatial heterogeneity, we further design a scale-aware diffusion mechanism to guide the diffusion process. Extensive experiments across ten real-world datasets from climate, energy, communication, and urban systems show that CoST achieves 25% performance gains over state-of-the-art baselines, while significantly reducing computational cost.
Authors: William Toner, Thomas L. Lee, Artjom Joosen, Rajkarn Singh, Martin Asenov
Abstract: Time series foundation models (FMs) have emerged as a popular paradigm for zero-shot multi-domain forecasting. FMs are trained on numerous diverse datasets and claim to be effective forecasters across multiple different time series domains, including cloud data. In this work we investigate this claim, exploring the effectiveness of FMs on cloud data. We demonstrate that many well-known FMs fail to generate meaningful or accurate zero-shot forecasts in this setting. We support this claim empirically, showing that FMs are outperformed consistently by simple linear baselines. We also illustrate a number of interesting pathologies, including instances where FMs suddenly output seemingly erratic, random-looking forecasts. Our results suggest a widespread failure of FMs to model cloud data.
Authors: D\'avid Terj\'ek
Abstract: We propose a homogeneous multilayer perceptron parameterization with polynomial hidden layer width pattern and analyze its training dynamics under stochastic gradient descent with depthwise gradient scaling in a general supervised learning scenario. We obtain formulas for the first three Taylor coefficients of the minibatch loss during training that illuminate the connection between sharpness and feature learning, providing in particular a soft rank variant that quantifies the quality of learned hidden layer features. Based on our theory, we design a gradient scaling scheme that in tandem with a quadratic width pattern enables training beyond the edge of stability without loss explosions or numerical errors, resulting in improved feature learning and implicit sharpness regularization as demonstrated empirically.
Authors: Sangkyu Lee, Janghoon Han, Hosung Song, Stanley Jungkyu Choi, Honglak Lee, Youngjae Yu
Abstract: Direct Preference Optimization (DPO) demonstrates the advantage of aligning a large language model with human preference using only an offline dataset. However, DPO has the limitation that the KL penalty, which prevents excessive deviation from the reference model, is static throughout the training process. Several methods claim to change this static KL penalty of DPO into a dynamic one, but no approach can adaptively assign different KL penalties for each preference pair. In this paper, we propose $\varepsilon$-Direct Preference Optimization ($\varepsilon$-DPO), which allows adaptive control of the KL penalty strength $\beta$ for each preference pair. Specifically, $\varepsilon$-DPO adaptively controls $\beta$ for each preference pair based on the monotonicity of logits as a preference model under the perturbation of $\beta$ during training. This is equivalent to adjusting the KL penalty by checking whether the change in training-time temperature can lead to better preference confidence as preference models by simply reusing the logit of the current policy and the reference policy. Experimental results show that the simple criterion of $\varepsilon$-DPO for KL penalty relaxation significantly improves DPO compared to most existing direct alignment algorithms on general chatbot benchmarks and reveal that this KL penalty control criterion can reflect confusion as a preference model and provide an efficient KL trade-off, highlighting the significance of instance-level adaptive KL penalty control in DPO.
Authors: Adrien Aumon, Shuang Ni, Myriam Lizotte, Guy Wolf, Kevin R. Moon, Jake S. Rhodes
Abstract: Extensive research has produced robust methods for unsupervised data visualization. Yet supervised visualization$\unicode{x2013}$where expert labels guide representations$\unicode{x2013}$remains underexplored, as most supervised approaches prioritize classification over visualization. Recently, RF-PHATE, a diffusion-based manifold learning method leveraging random forests and information geometry, marked significant progress in supervised visualization. However, its lack of an explicit mapping function limits scalability and its application to unseen data, posing challenges for large datasets and label-scarce scenarios. To overcome these limitations, we introduce Random Forest Autoencoders (RF-AE), a neural network-based framework for out-of-sample kernel extension that combines the flexibility of autoencoders with the supervised learning strengths of random forests and the geometry captured by RF-PHATE. RF-AE enables efficient out-of-sample supervised visualization and outperforms existing methods, including RF-PHATE's standard kernel extension, in both accuracy and interpretability. Additionally, RF-AE is robust to the choice of hyperparameters and generalizes to any kernel-based dimensionality reduction method.
Authors: Denis Tarasov, Alexander Nikulin, Ilya Zisman, Albina Klepach, Andrei Polubarov, Nikita Lyubaykin, Alexander Derevyagin, Igor Kiselev, Vladislav Kurenkov
Abstract: Existing offline in-context reinforcement learning (ICRL) methods have predominantly relied on supervised training objectives, which are known to have limitations in offline RL settings. In this study, we explore the integration of RL objectives within an offline ICRL framework. Through experiments on more than 150 GridWorld and MuJoCo environment-derived datasets, we demonstrate that optimizing RL objectives directly improves performance by approximately 30% on average compared to widely adopted Algorithm Distillation (AD), across various dataset coverages, structures, expertise levels, and environmental complexities. Furthermore, in the challenging XLand-MiniGrid environment, RL objectives doubled the performance of AD. Our results also reveal that the addition of conservatism during value learning brings additional improvements in almost all settings tested. Our findings emphasize the importance of aligning ICRL learning objectives with the RL reward-maximization goal, and demonstrate that offline RL is a promising direction for advancing ICRL.
Authors: Shuheng Chen, Junyi Fan, Elham Pishgar, Kamiar Alaei, Greg Placencia, Maryam Pishgar
Abstract: Background: Sepsis-Associated Acute Kidney Injury (SA-AKI) leads to high mortality in intensive care. This study develops machine learning models using the Medical Information Mart for Intensive Care IV (MIMIC-IV) database to predict Intensive Care Unit (ICU) mortality in SA-AKI patients. External validation is conducted using the eICU Collaborative Research Database. Methods: For 9,474 identified SA-AKI patients in MIMIC-IV, key features like lab results, vital signs, and comorbidities were selected using Variance Inflation Factor (VIF), Recursive Feature Elimination (RFE), and expert input, narrowing to 24 predictive variables. An Extreme Gradient Boosting (XGBoost) model was built for in-hospital mortality prediction, with hyperparameters optimized using GridSearch. Model interpretability was enhanced with SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME). External validation was conducted using the eICU database. Results: The proposed XGBoost model achieved an internal Area Under the Receiver Operating Characteristic curve (AUROC) of 0.878 (95% Confidence Interval: 0.859-0.897). SHAP identified Sequential Organ Failure Assessment (SOFA), serum lactate, and respiratory rate as key mortality predictors. LIME highlighted serum lactate, Acute Physiology and Chronic Health Evaluation II (APACHE II) score, total urine output, and serum calcium as critical features. Conclusions: The integration of advanced techniques with the XGBoost algorithm yielded a highly accurate and interpretable model for predicting SA-AKI mortality across diverse populations. It supports early identification of high-risk patients, enhancing clinical decision-making in intensive care. Future work needs to focus on enhancing adaptability, versatility, and real-world applications.
Authors: Jiawei Huang, Bingcong Li, Christoph Dann, Niao He
Abstract: Sample efficiency is critical for online Reinforcement Learning from Human Feedback (RLHF). While existing works investigate sample-efficient online exploration strategies, the potential of utilizing misspecified yet relevant reward models to accelerate learning remains underexplored. This paper studies how to transfer knowledge from those imperfect reward models in online RLHF. We start by identifying a novel property due to KL-regularization in the RLHF objective: \emph{a policy's coverability of the optimal policy is captured by its sub-optimality}. Building on this insight, we propose novel transfer learning principles and a theoretical algorithm -- \emph{\textbf{T}ransfer \textbf{P}olicy \textbf{O}ptimization (\textbf{TPO})} -- with provable benefits compared to standard online learning. Empirically, inspired by our theoretical findings, we develop a win-rate-based transfer policy selection strategy with improved computational efficiency. Moreover, our empirical transfer learning technique is modular and can be integrated with various policy optimization methods, such as DPO, IPO and XPO, to further enhance their performance. We validate the effectiveness of our method through experiments on summarization tasks.
Authors: Yi Zhao, Aidan Scannell, Wenshuai Zhao, Yuxin Hou, Tianyu Cui, Le Chen, Dieter B\"uchler, Arno Solin, Juho Kannala, Joni Pajarinen
Abstract: Leveraging offline data is a promising way to improve the sample efficiency of online reinforcement learning (RL). This paper expands the pool of usable data for offline-to-online RL by leveraging abundant non-curated data that is reward-free, of mixed quality, and collected across multiple embodiments. Although learning a world model appears promising for utilizing such data, we find that naive fine-tuning fails to accelerate RL training on many tasks. Through careful investigation, we attribute this failure to the distributional shift between offline and online data during fine-tuning. To address this issue and effectively use the offline data, we propose two essential techniques: \emph{i)} experience rehearsal and \emph{ii)} execution guidance. With these modifications, the non-curated offline data substantially improves RL's sample efficiency. Under limited sample budgets, our method achieves a 102.8\% relative improvement in aggregate score over learning-from-scratch baselines across 72 visuomotor tasks spanning 6 embodiments. On challenging tasks such as locomotion and robotic manipulation, it outperforms prior methods that utilize offline data by a decent margin.
Authors: Jianing Zhang, Evan Chen, Chaoyue Liu, Christopher G. Brinton
Abstract: Vertical Federated Learning (VFL) enables collaborative training with feature-partitioned data, yet remains vulnerable to privacy leakage through gradient transmissions. Standard differential privacy (DP) techniques such as DP-SGD are difficult to apply in this setting due to VFL's distributed nature and the high variance incurred by vector-valued noise. On the other hand, zeroth-order (ZO) optimization techniques can avoid explicit gradient exposure but lack formal privacy guarantees. In this work, we propose DPZV, the first ZO optimization framework for VFL that achieves tunable DP with performance guarantees. DPZV overcomes these limitations by injecting low-variance scalar noise at the server, enabling controllable privacy with reduced memory overhead. We conduct a comprehensive theoretical analysis showing that DPZV matches the convergence rate of first-order optimization methods while satisfying formal ($\epsilon, \delta$)-DP guarantees. Experiments on image and language benchmarks demonstrate that DPZV outperforms several baselines in terms of accuracy under a wide range of privacy constraints ($\epsilon \le 10$), thereby elevating the privacy-utility tradeoff in VFL.
Authors: Hajar Kazemi Naeini, Roya Shomali, Abolhassan Pishahang, Hamidreza Hasanzadeh, Mahdieh Mohammadi, Saeed Asadi, Abbas Varmaghani, Ahmad Gholizadeh Lonbar
Abstract: The advancement of smart grid technologies necessitates the integration of cutting-edge computational methods to enhance predictive energy optimization. This study proposes a multi-faceted approach by incorporating (1) Deep Reinforcement Learning (DRL) agents trained using data from Digital Twins (DTs) to optimize energy consumption in real time, (2) Physics-Informed Neural Networks (PINNs) to seamlessly embed physical laws within the optimization process, ensuring model accuracy and interpretability, and (3) Blockchain (BC) technology to facilitate secure and transparent communication across the smart grid infrastructure. The model was trained and validated using comprehensive datasets, including smart meter energy consumption data, renewable energy outputs, dynamic pricing, and user preferences collected from IoT devices. The proposed framework achieved superior predictive performance with a Mean Absolute Error (MAE) of 0.237 kWh, Root Mean Square Error (RMSE) of 0.298 kWh, and an R-squared (R2) value of 0.978, indicating a 97.8% explanation of data variance. Classification metrics further demonstrated the model's robustness, achieving 97.7% accuracy, 97.8% precision, 97.6% recall, and an F1 Score of 97.7%. Comparative analysis with traditional models like Linear Regression, Random Forest, SVM, LSTM, and XGBoost revealed the superior accuracy and real-time adaptability of the proposed method. In addition to enhancing energy efficiency, the model reduced energy costs by 35%, maintained a 96% user comfort index, and increased renewable energy utilization to 40%. This study demonstrates the transformative potential of integrating PINNs, DT, and Blockchain technologies to optimize energy consumption in smart grids, paving the way for sustainable, secure, and efficient energy management systems.
Authors: Daniil Sherki, Ivan Oseledets, Ekaterina Muravleva
Abstract: The efficient resolution of Bayesian inverse problems remains challenging due to the high computational cost of traditional sampling methods. In this paper, we propose a novel framework that integrates Conditional Flow Matching (CFM) with a transformer-based architecture to enable fast and flexible sampling from complex posterior distributions. The proposed methodology involves the direct learning of conditional probability trajectories from the data, leveraging CFM's ability to bypass iterative simulation and transformers' capacity to process arbitrary numbers of observations. The efficacy of the proposed framework is demonstrated through its application to three problems: a simple nonlinear model, a disease dynamics framework, and a two-dimensional Darcy flow Partial Differential Equation. The primary outcomes demonstrate that the relative errors in parameters recovery are as low as 1.5%, and that the inference time is reduced by up to 2000 times on CPU in comparison with the Monte Carlo Markov Chain. This framework facilitates the expeditious resolution of Bayesian problems through the utilisation of sampling from the learned conditional distribution.
Authors: Tiansheng Wen, Yifei Wang, Zequn Zeng, Zhong Peng, Yudi Su, Xinyang Liu, Bo Chen, Hongwei Liu, Stefanie Jegelka, Chenyu You
Abstract: Many large-scale systems rely on high-quality deep representations (embeddings) to facilitate tasks like retrieval, search, and generative modeling. Matryoshka Representation Learning (MRL) recently emerged as a solution for adaptive embedding lengths, but it requires full model retraining and suffers from noticeable performance degradations at short lengths. In this paper, we show that sparse coding offers a compelling alternative for achieving adaptive representation with minimal overhead and higher fidelity. We propose Contrastive Sparse Representation (CSR), a method that sparsifies pre-trained embeddings into a high-dimensional but selectively activated feature space. By leveraging lightweight autoencoding and task-aware contrastive objectives, CSR preserves semantic quality while allowing flexible, cost-effective inference at different sparsity levels. Extensive experiments on image, text, and multimodal benchmarks demonstrate that CSR consistently outperforms MRL in terms of both accuracy and retrieval speed-often by large margins-while also cutting training time to a fraction of that required by MRL. Our results establish sparse coding as a powerful paradigm for adaptive representation learning in real-world applications where efficiency and fidelity are both paramount. Code is available at https://github.com/neilwen987/CSR_Adaptive_Rep
Authors: Aleksandrs Slivkins, Yunzong Xu, Shiliang Zuo
Abstract: We study the greedy (exploitation-only) algorithm in bandit problems with a known reward structure. We allow arbitrary finite reward structures, while prior work focused on a few specific ones. We fully characterize when the greedy algorithm asymptotically succeeds or fails, in the sense of sublinear vs. linear regret as a function of time. Our characterization identifies a partial identifiability property of the problem instance as the necessary and sufficient condition for the asymptotic success. Notably, once this property holds, the problem becomes easy -- any algorithm will succeed (in the same sense as above), provided it satisfies a mild non-degeneracy condition. Our characterization extends to contextual bandits and interactive decision-making with arbitrary feedback. Examples demonstrating broad applicability and extensions to infinite reward structures are provided.
Authors: Jonas Elsborg, Luca Thiede, Al\'an Aspuru-Guzik, Tejs Vegge, Arghya Bhowmik
Abstract: We present the Electronic Tensor Reconstruction Algorithm (ELECTRA) - an equivariant model for predicting electronic charge densities using floating orbitals. Floating orbitals are a long-standing concept in the quantum chemistry community that promises more compact and accurate representations by placing orbitals freely in space, as opposed to centering all orbitals at the position of atoms. Finding the ideal placement of these orbitals requires extensive domain knowledge, though, which thus far has prevented widespread adoption. We solve this in a data-driven manner by training a Cartesian tensor network to predict the orbital positions along with orbital coefficients. This is made possible through a symmetry-breaking mechanism that is used to learn position displacements with lower symmetry than the input molecule while preserving the rotation equivariance of the charge density itself. Inspired by recent successes of Gaussian Splatting in representing densities in space, we are using Gaussian orbitals and predicting their weights and covariance matrices. Our method achieves a state-of-the-art balance between computational efficiency and predictive accuracy on established benchmarks.
Authors: Yaorui Shi, Jiaqi Yang, Changhao Nai, Sihang Li, Junfeng Fang, Xiang Wang, Zhiyuan Liu, Yang Zhang
Abstract: Single-cell RNA sequencing (scRNA-seq) offers detailed insights into cellular heterogeneity. Recent advancements leverage single-cell large language models (scLLMs) for effective representation learning. These models focus exclusively on transcriptomic data, neglecting complementary biological knowledge from textual descriptions. To overcome this limitation, we propose scMMGPT, a novel multimodal framework designed for language-enhanced representation learning in single-cell transcriptomics. Unlike existing methods, scMMGPT employs robust cell representation extraction, preserving quantitative gene expression data, and introduces an innovative two-stage pre-training strategy combining discriminative precision with generative flexibility. Extensive experiments demonstrate that scMMGPT significantly outperforms unimodal and multimodal baselines across key downstream tasks, including cell annotation and clustering, and exhibits superior generalization in out-of-distribution scenarios.
Authors: Marianne Arriola, Aaron Gokaslan, Justin T. Chiu, Zhihan Yang, Zhixuan Qi, Jiaqi Han, Subham Sekhar Sahoo, Volodymyr Kuleshov
Abstract: Diffusion language models offer unique benefits over autoregressive models due to their potential for parallelized generation and controllability, yet they lag in likelihood modeling and are limited to fixed-length generation. In this work, we introduce a class of block diffusion language models that interpolate between discrete denoising diffusion and autoregressive models. Block diffusion overcomes key limitations of both approaches by supporting flexible-length generation and improving inference efficiency with KV caching and parallel token sampling. We propose a recipe for building effective block diffusion models that includes an efficient training algorithm, estimators of gradient variance, and data-driven noise schedules to minimize the variance. Block diffusion sets a new state-of-the-art performance among diffusion models on language modeling benchmarks and enables generation of arbitrary-length sequences. We provide the code, along with the model weights and blog post on the project page: https://m-arriola.com/bd3lms
Authors: Yuanmin Huang, Mi Zhang, Zhaoxiang Wang, Wenxuan Li, Min Yang
Abstract: Time series classification (TSC) is a cornerstone of modern web applications, powering tasks such as financial data analysis, network traffic monitoring, and user behavior analysis. In recent years, deep neural networks (DNNs) have greatly enhanced the performance of TSC models in these critical domains. However, DNNs are vulnerable to backdoor attacks, where attackers can covertly implant triggers into models to induce malicious outcomes. Existing backdoor attacks targeting DNN-based TSC models remain elementary. In particular, early methods borrow trigger designs from computer vision, which are ineffective for time series data. More recent approaches utilize generative models for trigger generation, but at the cost of significant computational complexity. In this work, we analyze the limitations of existing attacks and introduce an enhanced method, FreqBack. Drawing inspiration from the fact that DNN models inherently capture frequency domain features in time series data, we identify that improper perturbations in the frequency domain are the root cause of ineffective attacks. To address this, we propose to generate triggers both effectively and efficiently, guided by frequency analysis. FreqBack exhibits substantial performance across five models and eight datasets, achieving an impressive attack success rate of over 90%, while maintaining less than a 3% drop in model accuracy on clean data.
Authors: Luyuan Xie, Tianyu Luan, Wenyuan Cai, Guochen Yan, Zhaoyu Chen, Nan Xi, Yuejian Fang, Qingni Shen, Zhonghai Wu, Junsong Yuan
Abstract: Federated learning has wide applications in the medical field. It enables knowledge sharing among different healthcare institutes while protecting patients' privacy. However, existing federated learning systems are typically centralized, requiring clients to upload client-specific knowledge to a central server for aggregation. This centralized approach would integrate the knowledge from each client into a centralized server, and the knowledge would be already undermined during the centralized integration before it reaches back to each client. Besides, the centralized approach also creates a dependency on the central server, which may affect training stability if the server malfunctions or connections are unstable. To address these issues, we propose a decentralized federated learning framework named dFLMoE. In our framework, clients directly exchange lightweight head models with each other. After exchanging, each client treats both local and received head models as individual experts, and utilizes a client-specific Mixture of Experts (MoE) approach to make collective decisions. This design not only reduces the knowledge damage with client-specific aggregations but also removes the dependency on the central server to enhance the robustness of the framework. We validate our framework on multiple medical tasks, demonstrating that our method evidently outperforms state-of-the-art approaches under both model homogeneity and heterogeneity settings.
Authors: Afrar Jahin, Arif Hassan Zidan, Wei Zhang, Yu Bao, Tianming Liu
Abstract: With the rapid advancement of Artificial Intelligence (AI), Large Language Models (LLMs) have significantly impacted a wide array of domains, including healthcare, engineering, science, education, and mathematical reasoning. Among these, mathematical reasoning remains a particularly challenging capability, often requiring multi-step logic and abstract generalization. While prior work has explored LLM performance on reasoning tasks, comprehensive evaluations that span both depth and breadth across model families remain limited. In this study, we present a systematic evaluation of mathematical reasoning abilities across eight leading LLMs, including two recent DeepSeek models, using three independent benchmark datasets. Our analyses reveal several key findings: (1) DeepSeek-R1 performs competitively with o1 across most domains and achieves the highest accuracy on the MMLU Formal Logic benchmark; (2) distilled variants, such as DeepSeek-1.5B, exhibit substantial performance degradation; and (3) Gemini 2.0 Flash achieves the lowest response latency. Beyond quantitative metrics, we explore how architectural choices, training paradigms, and optimization strategies contribute to variation in reasoning performance. These findings provide new insights into the capabilities and limitations of current LLMs in mathematical domains, and offer guidance for the development of future models better aligned with rigorous reasoning demands.
Authors: Jia Zhang, Chen-Xi Zhang, Yao Liu, Yi-Xuan Jin, Xiao-Wen Yang, Bo Zheng, Yi Liu, Lan-Zhe Guo
Abstract: Recent advancements in instruction tuning for large language models (LLMs) suggest that a small, high-quality dataset can significantly equip LLMs with instruction-following capabilities, outperforming large datasets often burdened by quality and redundancy issues. However, the challenge lies in automatically identifying valuable subsets from large datasets to boost both the effectiveness and efficiency of instruction tuning. In this paper, we first establish data selection criteria based on three distinct aspects of data value: diversity, difficulty, and dependability, and then propose the D3 method comprising two key steps of scoring and selection. Specifically, in the scoring step, we define the diversity function to measure sample distinctiveness and introduce the uncertainty-based prediction difficulty to evaluate sample difficulty by mitigating the interference of context-oriented generation diversity. Additionally, we integrate an external LLM for dependability assessment. In the selection step, we formulate the D3 weighted coreset objective, which jointly optimizes three aspects of data value to solve for the most valuable subset. The two steps of D3 can iterate multiple rounds, incorporating feedback to refine the selection focus adaptively. Experiments on both public datasets and the real-world Taobao Live application demonstrate the effectiveness of D3 in endowing LLMs with competitive or even superior instruction-following capabilities using less than 10\% of the entire dataset.
Authors: Ziyad Sheebaelhamd, Michael Tschannen, Michael Muehlebach, Claire Vernade
Abstract: Current transformer-based imitation learning approaches introduce discrete action representations and train an autoregressive transformer decoder on the resulting latent code. However, the initial quantization breaks the continuous structure of the action space thereby limiting the capabilities of the generative model. We propose a quantization-free method instead that leverages Generative Infinite-Vocabulary Transformers (GIVT) as a direct, continuous policy parametrization for autoregressive transformers. This simplifies the imitation learning pipeline while achieving state-of-the-art performance on a variety of popular simulated robotics tasks. We enhance our policy roll-outs by carefully studying sampling algorithms, further improving the results.
Authors: Daniel Herbst, Stefanie Jegelka
Abstract: Graph limit models, like graphons for limits of dense graphs, have recently been used to study size transferability of graph neural networks (GNNs). While most literature focuses on message passing GNNs (MPNNs), in this work we attend to the more powerful higher-order GNNs. First, we extend the $k$-WL test for graphons (B\"oker, 2023) to the graphon-signal space and introduce signal-weighted homomorphism densities as a key tool. As an exemplary focus, we generalize Invariant Graph Networks (IGNs) to graphons, proposing Invariant Graphon Networks (IWNs) defined via a subset of the IGN basis corresponding to bounded linear operators. Even with this restricted basis, we show that IWNs of order $k$ are at least as powerful as the $k$-WL test, and we establish universal approximation results for graphon-signals in $L^p$ distances. This significantly extends the prior work of Cai & Wang (2022), showing that IWNs--a subset of their IGN-small--retain effectively the same expressivity as the full IGN basis in the limit. In contrast to their approach, our blueprint of IWNs also aligns better with the geometry of graphon space, for example facilitating comparability to MPNNs. We highlight that, while typical higher-order GNNs are discontinuous w.r.t. cut distance--which causes their lack of convergence and is inherently tied to the definition of $k$-WL--transferability remains achievable.
Authors: Zihao Guo, Shuqing Shi, Richard Willis, Tristan Tomilin, Joel Z. Leibo, Yali Du
Abstract: Sequential social dilemmas pose a significant challenge in the field of multi-agent reinforcement learning (MARL), requiring environments that accurately reflect the tension between individual and collective interests. Previous benchmarks and environments, such as Melting Pot, provide an evaluation protocol that measures generalization to new social partners in various test scenarios. However, running reinforcement learning algorithms in traditional environments requires substantial computational resources. In this paper, we introduce SocialJax, a suite of sequential social dilemma environments and algorithms implemented in JAX. JAX is a high-performance numerical computing library for Python that enables significant improvements in operational efficiency. Our experiments demonstrate that the SocialJax training pipeline achieves at least 50\texttimes{} speed-up in real-time performance compared to Melting Pot RLlib baselines. Additionally, we validate the effectiveness of baseline algorithms within SocialJax environments. Finally, we use Schelling diagrams to verify the social dilemma properties of these environments, ensuring that they accurately capture the dynamics of social dilemmas.
Authors: Nima Negarandeh, Carlos Mora, Ramin Bostanabad
Abstract: Gaussian processes (GPs) are powerful probabilistic models that define flexible priors over functions, offering strong interpretability and uncertainty quantification. However, GP models often rely on simple, stationary kernels which can lead to suboptimal predictions and miscalibrated uncertainty estimates, especially in nonstationary real-world applications. In this paper, we introduce SEEK, a novel class of learnable kernels to model complex, nonstationary functions via GPs. Inspired by artificial neurons, SEEK is derived from first principles to ensure symmetry and positive semi-definiteness, key properties of valid kernels. The proposed method achieves flexible and adaptive nonstationarity by learning a mapping from a set of base kernels. Compared to existing techniques, our approach is more interpretable and much less prone to overfitting. We conduct comprehensive sensitivity analyses and comparative studies to demonstrate that our approach is not only robust to many of its design choices, but also outperforms existing stationary/nonstationary kernels in both mean prediction accuracy and uncertainty quantification.
Authors: Xiaohao Liu, Xiaobo Xia, See-Kiong Ng, Tat-Seng Chua
Abstract: Multimodal contrastive learning (MCL) advances in aligning different modalities and generating multimodal representations in a joint space. By leveraging contrastive learning across diverse modalities, large-scale multimodal data enhances representational quality. However, a critical yet often overlooked challenge remains: multimodal data is rarely collected in a single process, and training from scratch is computationally expensive. Instead, emergent multimodal data can be used to optimize existing models gradually, \textit{i.e.}, models are trained on a sequence of modality pair data. We define this problem as Continual Multimodal Contrastive Learning (CMCL), an underexplored yet crucial research direction at the intersection of multimodal and continual learning. In this paper, we formulate CMCL through two specialized principles of stability and plasticity. We theoretically derive a novel optimization-based method, which projects updated gradients from dual sides onto subspaces where any gradient is prevented from interfering with the previously learned knowledge. Two upper bounds provide theoretical insights on both stability and plasticity in our solution. Beyond our theoretical contributions, we conduct experiments on multiple datasets by comparing our method against advanced continual learning baselines. The empirical results further support our claims and demonstrate the efficacy of our method. The code will be publicly available.
Authors: Massimo Bini, Leander Girrbach, Zeynep Akata
Abstract: Parameter-Efficient FineTuning (PEFT) methods have recently gained significant popularity thanks to the widespread availability of large-scale pretrained models. These methods allow for quick adaptation to downstream tasks with minimal computational cost. However, popular finetuning methods such as LoRA exhibit limited robustness when it comes to hyperparameter choices or extended training regimes, preventing optimal out-of-the-box performance. In contrast, bounded approaches, such as ETHER, provide greater robustness but are limited to extremely low-rank adaptations and fixed-strength transformations, reducing their adaptation expressive power. In this work, we propose Decoupled Low-rank Adaptation (DeLoRA), a novel finetuning method that normalizes and scales learnable low-rank matrices. By bounding the distance of the transformation, DeLoRA effectively decouples the angular learning from the adaptation strength, enhancing robustness without compromising performance. Through evaluations on subject-driven image generation, natural language understanding, and instruction tuning, we show that DeLoRA matches or surpasses performance of competing PEFT methods, while exhibiting stronger robustness. Code is available at https://github.com/ExplainableML/DeLoRA.
Authors: Varun Mulchandani, Jung-Eun Kim
Abstract: Deep neural networks have been shown to learn and rely on spurious correlations present in the data that they are trained on. Reliance on such correlations can cause these networks to malfunction when deployed in the real world, where these correlations may no longer hold. To overcome the learning of and reliance on such correlations, recent studies propose approaches that yield promising results. These works, however, study settings where the strength of the spurious signal is significantly greater than that of the core, invariant signal, making it easier to detect the presence of spurious features in individual training samples and allow for further processing. In this paper, we identify new settings where the strength of the spurious signal is relatively weaker, making it difficult to detect any spurious information while continuing to have catastrophic consequences. We also discover that spurious correlations are learned primarily due to only a handful of all the samples containing the spurious feature and develop a novel data pruning technique that identifies and prunes small subsets of the training data that contain these samples. Our proposed technique does not require inferred domain knowledge, information regarding the sample-wise presence or nature of spurious information, or human intervention. Finally, we show that such data pruning attains state-of-the-art performance on previously studied settings where spurious information is identifiable.
Authors: Zihao Zheng, Xiuping Cui, Size Zheng, Maoliang Li, Jiayu Chen, Yun Liang, Xiang Chen
Abstract: With the advances in artificial intelligence, Mix-of-Experts (MoE) has become the main form of Large Language Models (LLMs), and its demand for model compression is increasing. Quantization is an effective method that not only compresses the models but also significantly accelerates their performance. Existing quantization methods have gradually shifted the focus from parameter scaling to the analysis of data distributions. However, their analysis is designed for dense LLMs, which are suboptimal for MoE quantization, due to MoEs' complex data-model distribution. To address this problem, we decouple the complexity of MoEs' data-model distribution into a multi-stage analysis and reveal MoEs' inherent dynamics. The analysis results show that the expert performance of MoE varies dynamically both within and across data distributions. Based on these, we design two quantization strategies with data-model distribution awareness and integrate them into an end-to-end framework for MoE quantization, which is named MoQa. MoQa uses an expert-level mix-precision base quantization with distribution awareness. Moreover, MoQa uses a channel-level quantization adjustment to dynamically adjust expert performance to adapt to novel distributions. Experiments show that MoQa's base quantization achieves a 0.49~8.51 PPL decrease on known distributions. With the adjustments, MoQa achieves a 2.74~6.44 PPL decrease and 1.85%~3.77% average accuracy improvements on novel distributions. We believe MoQa will play a role in future MoE construction, optimization, and compression.
Authors: Antonis Matakos, Martino Ciaperoni, Heikki Mannila
Abstract: The Min-Max Fair PCA problem seeks a low-rank representation of multi-group data such that the the approximation error is as balanced as possible across groups. Existing approaches to this problem return a rank-$d$ fair subspace, but lack the fundamental containment property of standard PCA: each rank-$d$ PCA subspace should contain all lower-rank PCA subspaces. To fill this gap, we define fair principal components as directions that minimize the maximum group-wise reconstruction error, subject to orthogonality with previously selected components, and we introduce an iterative method to compute them. This approach preserves the containment property of standard PCA, and reduces to standard \pca for data with a single group. We analyze the theoretical properties of our method and show empirically that it outperforms existing approaches to Min-Max Fair PCA.
Authors: Tong Wu, Chong Xiang, Jiachen T. Wang, G. Edward Suh, Prateek Mittal
Abstract: Reasoning-enhanced large language models (LLMs) explicitly generate intermediate reasoning steps prior to generating final answers, helping the model excel in complex problem-solving. In this paper, we demonstrate that this emerging generation framework offers a unique opportunity for more fine-grained control over model behavior. We propose Thinking Intervention, a novel paradigm designed to explicitly guide the internal reasoning processes of LLMs by strategically inserting or revising specific thinking tokens. We find that the Thinking Intervention paradigm enhances the capabilities of reasoning models across a wide range of tasks, including instruction following on IFEval, instruction hierarchy on SEP, and safety alignment on XSTest and SorryBench. Our results demonstrate that Thinking Intervention significantly outperforms baseline prompting approaches, achieving up to 6.7% accuracy gains in instruction-following scenarios, 15.4% improvements in reasoning about instruction hierarchies, and a 40.0% increase in refusal rates for unsafe prompts using open-source DeepSeek R1 models. Overall, our work opens a promising new research avenue for controlling reasoning LLMs.
Authors: Canlin Zhang, Xiuwen Liu
Abstract: Deep neural networks trained through end-to-end learning have achieved remarkable success across various domains in the past decade. However, the end-to-end learning strategy faces two fundamental limitations: the struggle to form explainable representations in a self-supervised manner, and the inability to compress information rigorously following the Minimum Description Length (MDL) principle. In this paper, we establish a novel theory connecting these two challenges. We design the Spectrum VAE, a novel deep learning architecture whose minimum description length (MDL) can be rigorously evaluated. Then, we introduce the concept of latent dimension combinations, or what we term spiking patterns, and demonstrate that the observed spiking patterns should be as few as possible based on the training data in order for the Spectrum VAE to achieve the MDL. Finally, our theory demonstrates that when the MDL is achieved with respect to the given data distribution, the model will naturally produce explainable latent representations of the data. That is, explainable representations of the data, or understanding the data, can be achieved in a self-supervised manner simply by making the deep neural network obey the MDL principle. In our opinion, this reveals an even more profound principle: Understanding means to represent the acquired information by as small an amount of information as possible. This work is entirely theoretical and aims at inspiring future research to realize self-supervised explainable AI simply by obeying the MDL principle.
Authors: Haoyuan Li, Mathias Funk, Jindong Wang, Aaqib Saeed
Abstract: Federated Active Learning (FAL) has emerged as a promising framework to leverage large quantities of unlabeled data across distributed clients while preserving data privacy. However, real-world deployments remain limited by high annotation costs and communication-intensive sampling processes, particularly in a cross-silo setting, when clients possess substantial local datasets. This paper addresses the crucial question: What is the best practice to reduce communication costs in human-in-the-loop learning with minimal annotator effort? Existing FAL methods typically rely on iterative annotation processes that separate active sampling from federated updates, leading to multiple rounds of expensive communication and annotation. In response, we introduce FAST, a two-pass FAL framework that harnesses foundation models for weak labeling in a preliminary pass, followed by a refinement pass focused exclusively on the most uncertain samples. By leveraging representation knowledge from foundation models and integrating refinement steps into a streamlined workflow, FAST substantially reduces the overhead incurred by iterative active sampling. Extensive experiments on diverse medical and natural image benchmarks demonstrate that FAST outperforms existing FAL methods by an average of 4.36% while reducing communication rounds eightfold under a limited 5% labeling budget.
Authors: Shiguang Sun, Hanbo Zhang, Zeyang Liu, Xinrui Yang, Lipeng Wan, Bing Yan, Xingyu Chen, Xuguang Lan
Abstract: Existing visual model-based reinforcement learning (MBRL) algorithms with observation reconstruction often suffer from information conflicts, making it difficult to learn compact representations and hence result in less robust policies, especially in the presence of task-irrelevant visual distractions. In this paper, we first reveal that the information conflicts in current visual MBRL algorithms stem from visual representation learning and latent dynamics modeling with an information-theoretic perspective. Based on this finding, we present a new algorithm to resolve information conflicts for visual MBRL, named MInCo, which mitigates information conflicts by leveraging negative-free contrastive learning, aiding in learning invariant representation and robust policies despite noisy observations. To prevent the dominance of visual representation learning, we introduce time-varying reweighting to bias the learning towards dynamics modeling as training proceeds. We evaluate our method on several robotic control tasks with dynamic background distractions. Our experiments demonstrate that MInCo learns invariant representations against background noise and consistently outperforms current state-of-the-art visual MBRL methods. Code is available at https://github.com/ShiguangSun/minco.
Authors: Qingyang Zhang, Haitao Wu, Changqing Zhang, Peilin Zhao, Yatao Bian
Abstract: Existing methods to enhance the reasoning capability of large language models predominantly rely on supervised fine-tuning (SFT) followed by reinforcement learning (RL) on reasoning-specific data. These approaches critically depend on external supervisions--such as labeled reasoning traces, verified golden answers, or pre-trained reward models. In this work, we propose Entropy Minimized Policy Optimization (\ours), which makes an early attempt at fully unsupervised LLM reasoning incentivization. By continuously minimizing the predictive entropy of LLMs on unlabeled questions in a latent semantic space, \ours achieves competitive performance compared to supervised counterparts on both mathematical and free-form natural reasoning tasks. Specifically, without any supervised signals, \ours boosts the accuracy of Qwen2.5-Math-7B Base from 30.7\% to 48.1\% on mathematical benchmarks and improves the accuracy of Qwen2.5-7B Base from 32.1\% to 50.1\% on MMLU-Pro. Primary experiments and analysis are also provided to interpret the effectiveness of \ours. Code is available at https://github.com/QingyangZhang/EMPO.
Authors: Rui Pan, Yinwei Dai, Zhihao Zhang, Gabriele Oliaro, Zhihao Jia, Ravi Netravali
Abstract: Recent advances in inference-time compute have significantly improved performance on complex tasks by generating long chains of thought (CoTs) using Large Reasoning Models (LRMs). However, this improved accuracy comes at the cost of high inference latency due to the length of generated reasoning sequences and the autoregressive nature of decoding. Our key insight in tackling these overheads is that LRM inference, and the reasoning that it embeds, is highly tolerant of approximations: complex tasks are typically broken down into simpler steps, each of which brings utility based on the semantic insight it provides for downstream steps rather than the exact tokens it generates. Accordingly, we introduce SpecReason, a system that automatically accelerates LRM inference by using a lightweight model to (speculatively) carry out simpler intermediate reasoning steps and reserving the costly base model only to assess (and potentially correct) the speculated outputs. Importantly, SpecReason's focus on exploiting the semantic flexibility of thinking tokens in preserving final-answer accuracy is complementary to prior speculation techniques, most notably speculative decoding, which demands token-level equivalence at each step. Across a variety of reasoning benchmarks, SpecReason achieves $1.4-3.0\times$ speedup over vanilla LRM inference while improving accuracy by $0.4-9.0\%$. Compared to speculative decoding without SpecReason, their combination yields an additional $8.8-58.0\%$ latency reduction. We open-source SpecReason at https://github.com/ruipeterpan/specreason.
Authors: Sahil Sethi, David Chen, Thomas Statchen, Michael C. Burkhart, Nipun Bhandari, Bashar Ramadan, Brett Beaulieu-Jones
Abstract: Deep learning-based electrocardiogram (ECG) classification has shown impressive performance but clinical adoption has been slowed by the lack of transparent and faithful explanations. Post hoc methods such as saliency maps may fail to reflect a model's true decision process. Prototype-based reasoning offers a more transparent alternative by grounding decisions in similarity to learned representations of real ECG segments, enabling faithful, case-based explanations. We introduce ProtoECGNet, a prototype-based deep learning model for interpretable, multi-label ECG classification. ProtoECGNet employs a structured, multi-branch architecture that reflects clinical interpretation workflows: it integrates a 1D CNN with global prototypes for rhythm classification, a 2D CNN with time-localized prototypes for morphology-based reasoning, and a 2D CNN with global prototypes for diffuse abnormalities. Each branch is trained with a prototype loss designed for multi-label learning, combining clustering, separation, diversity, and a novel contrastive loss that encourages appropriate separation between prototypes of unrelated classes while allowing clustering for frequently co-occurring diagnoses. We evaluate ProtoECGNet on all 71 diagnostic labels from the PTB-XL dataset, demonstrating competitive performance relative to state-of-the-art black-box models while providing structured, case-based explanations. To assess prototype quality, we conduct a structured clinician review of the final model's projected prototypes, finding that they are rated as representative and clear. ProtoECGNet shows that prototype learning can be effectively scaled to complex, multi-label time-series classification, offering a practical path toward transparent and trustworthy deep learning models for clinical decision support.
Authors: Yuchu Jiang, Jiale Fu, Chenduo Hao, Xinting Hu, Yingzhe Peng, Xin Geng, Xu Yang
Abstract: Recently, In-context Learning (ICL) has become a significant inference paradigm in Large Multimodal Models (LMMs), utilizing a few in-context demonstrations (ICDs) to prompt LMMs for new tasks. However, the synergistic effects in multimodal data increase the sensitivity of ICL performance to the configurations of ICDs, stimulating the need for a more stable and general mapping function. Mathematically, in Transformer-based models, ICDs act as "shift vectors" added to the hidden states of query tokens. Inspired by this, we introduce Mimic In-Context Learning (MimIC) to learn stable and generalizable shift effects from ICDs. Specifically, compared with some previous shift vector-based methods, MimIC more strictly approximates the shift effects by integrating lightweight learnable modules into LMMs with four key enhancements: 1) inserting shift vectors after attention layers, 2) assigning a shift vector to each attention head, 3) making shift magnitude query-dependent, and 4) employing a layer-wise alignment loss. Extensive experiments on two LMMs (Idefics-9b and Idefics2-8b-base) across three multimodal tasks (VQAv2, OK-VQA, Captioning) demonstrate that MimIC outperforms existing shift vector-based methods. The code is available at https://github.com/Kamichanw/MimIC.
Authors: Yichao Cai, Yuhang Liu, Erdun Gao, Tianjiao Jiang, Zhen Zhang, Anton van den Hengel, Javen Qinfeng Shi
Abstract: Multimodal representation learning, exemplified by multimodal contrastive learning (MMCL) using image-text pairs, aims to learn powerful representations by aligning cues across modalities. This approach relies on the core assumption that the exemplar image-text pairs constitute two representations of an identical concept. However, recent research has revealed that real-world datasets often exhibit cross-modal misalignment. There are two distinct viewpoints on how to address this issue: one suggests mitigating the misalignment, and the other leveraging it. We seek here to reconcile these seemingly opposing perspectives, and to provide a practical guide for practitioners. Using latent variable models we thus formalize cross-modal misalignment by introducing two specific mechanisms: Selection bias, where some semantic variables are absent in the text, and perturbation bias, where semantic variables are altered -- both leading to misalignment in data pairs. Our theoretical analysis demonstrates that, under mild assumptions, the representations learned by MMCL capture exactly the information related to the subset of the semantic variables invariant to selection and perturbation biases. This provides a unified perspective for understanding misalignment. Based on this, we further offer actionable insights into how misalignment should inform the design of real-world ML systems. We validate our theoretical findings via extensive empirical studies on both synthetic data and real image-text datasets, shedding light on the nuanced impact of cross-modal misalignment on multimodal representation learning.
Authors: You Rim Choi, Subeom Park, Seojun Heo, Eunchung Noh, Hyung-Sin Kim
Abstract: Open-Set Semi-Supervised Learning (OSSL) tackles the practical challenge of learning from unlabeled data that may include both in-distribution (ID) and unknown out-of-distribution (OOD) classes. However, existing OSSL methods form suboptimal feature spaces by either excluding OOD samples, interfering with them, or overtrusting their information during training. In this work, we introduce MagMatch, a novel framework that naturally isolates OOD samples through a prototype-based contrastive learning paradigm. Unlike conventional methods, MagMatch does not assign any prototypes to OOD samples; instead, it selectively aligns ID samples with class prototypes using an ID-Selective Magnetic (ISM) module, while allowing OOD samples - the "others" - to remain unaligned in the feature space. To support this process, we propose Selective Magnetic Alignment (SMA) loss for unlabeled data, which dynamically adjusts alignment based on sample confidence. Extensive experiments on diverse datasets demonstrate that MagMatch significantly outperforms existing methods in both closed-set classification accuracy and OOD detection AUROC, especially in generalizing to unseen OOD data.
Authors: Hao Xu, Xiangru Jian, Xinjian Zhao, Wei Pang, Chao Zhang, Suyuchen Wang, Qixin Zhang, Zhengyuan Dong, Joao Monteiro, Bang Liu, Qiuzhuang Sun, Tianshu Yu
Abstract: This paper introduces GraphOmni, a comprehensive benchmark designed to evaluate the reasoning capabilities of LLMs on graph-theoretic tasks articulated in natural language. GraphOmni encompasses diverse graph types, serialization formats, and prompting schemes, significantly exceeding prior efforts in both scope and depth. Through extensive systematic evaluation, we identify critical interactions among these dimensions, demonstrating their substantial impact on model performance. Our experiments reveal that state-of-the-art models like Claude-3.5 and o4-mini consistently outperform other models, yet even these leading models exhibit substantial room for improvement. Performance variability is evident depending on the specific combinations of factors we considered, underscoring the necessity of comprehensive evaluations across these interconnected dimensions. Additionally, we observe distinct impacts of serialization and prompting strategies between open-source and closed-source models, encouraging the development of tailored approaches. Motivated by the findings, we also propose a reinforcement learning-inspired framework that adaptively selects the optimal factors influencing LLM reasoning capabilities. This flexible and extendable benchmark not only deepens our understanding of LLM performance on structured tasks but also provides a robust foundation for advancing research in LLM-based graph reasoning.
Authors: Helen Jin, Anton Xue, Weiqiu You, Surbhi Goel, Eric Wong
Abstract: Stability guarantees have emerged as a principled way to evaluate feature attributions, but existing certification methods rely on heavily smoothed classifiers and often produce conservative guarantees. To address these limitations, we introduce soft stability and propose a simple, model-agnostic, sample-efficient stability certification algorithm (SCA) that yields non-trivial and interpretable guarantees for any attribution method. Moreover, we show that mild smoothing achieves a more favorable trade-off between accuracy and stability, avoiding the aggressive compromises made in prior certification methods. To explain this behavior, we use Boolean function analysis to derive a novel characterization of stability under smoothing. We evaluate SCA on vision and language tasks and demonstrate the effectiveness of soft stability in measuring the robustness of explanation methods.
Authors: Zhanglin Wu, Tengfei Song, Ning Xie, Mengli Zhu, Weidong Zhang, Shuang Wu, Pengfei Li, Chong Li, Junhao Zhu, Hao Yang, Shiliang Sun
Abstract: The rapid advancement of large vision-language models (LVLMs) has significantly propelled applications in document understanding, particularly in optical character recognition (OCR) and multilingual translation. However, current evaluations of LVLMs, like the widely used OCRBench, mainly focus on verifying the correctness of their short-text responses and long-text responses with simple layout, while the evaluation of their ability to understand long texts with complex layout design is highly significant but largely overlooked. In this paper, we propose Menu OCR and Translation Benchmark (MOTBench), a specialized evaluation framework emphasizing the pivotal role of menu translation in cross-cultural communication. MOTBench requires LVLMs to accurately recognize and translate each dish, along with its price and unit items on a menu, providing a comprehensive assessment of their visual understanding and language processing capabilities. Our benchmark is comprised of a collection of Chinese and English menus, characterized by intricate layouts, a variety of fonts, and culturally specific elements across different languages, along with precise human annotations. Experiments show that our automatic evaluation results are highly consistent with professional human evaluation. We evaluate a range of publicly available state-of-the-art LVLMs, and through analyzing their output to identify the strengths and weaknesses in their performance, offering valuable insights to guide future advancements in LVLM development. MOTBench is available at https://github.com/gitwzl/MOTBench.
Authors: Enrico Parisini, Tapabrata Chakraborti, Chris Harbron, Ben D. MacArthur, Christopher R. S. Banerji
Abstract: Concept Bottleneck Models aim to improve interpretability by predicting high-level intermediate concepts, representing a promising approach for deployment in high-risk scenarios. However, they are known to suffer from information leakage, whereby models exploit unintended information encoded within the learned concepts. We introduce an information-theoretic framework to rigorously characterise and quantify leakage, and define two complementary measures: the concepts-task leakage (CTL) and interconcept leakage (ICL) scores. We show that these measures are strongly predictive of model behaviour under interventions and outperform existing alternatives in robustness and reliability. Using this framework, we identify the primary causes of leakage and provide strong evidence that Concept Embedding Models exhibit substantial leakage regardless of the hyperparameters choice. Finally, we propose practical guidelines for designing concept-based models to reduce leakage and ensure interpretability.
Authors: Xinye Chen
Abstract: This paper presents a novel reinforcement learning (RL) framework for dynamically optimizing numerical precision in the preconditioned conjugate gradient (CG) method. By modeling precision selection as a Markov Decision Process (MDP), we employ Q-learning to adaptively assign precision levels to key operations, striking an optimal balance between computational efficiency and numerical accuracy, while ensuring stability through double-precision scalar computations and residual computing. In practice, the algorithm is trained on a set of data and subsequently performs inference for precision selection on out-of-sample data, without requiring re-analysis or retraining for new datasets. This enables the method to adapt seamlessly to new problem instances without the computational overhead of recalibration. Our results demonstrate the effectiveness of RL in enhancing solver's performance, marking the first application of RL to mixed-precision numerical methods. The findings highlight the approach's practical advantages, robustness, and scalability, providing valuable insights into its integration with iterative solvers and paving the way for AI-driven advancements in scientific computing.
Authors: Assaf Gerner, Netta Madvil, Nadav Barak, Alex Zaikman, Jonatan Liberman, Liron Hamra, Rotem Brazilay, Shay Tsadok, Yaron Friedman, Neal Harow, Noam Bresler, Shir Chorev, Philip Tannor
Abstract: Despite advancements in grounded content generation, production Large Language Models (LLMs) based applications still suffer from hallucinated answers. We present "Grounded in Context" - Deepchecks' hallucination detection framework, designed for production-scale long-context data and tailored to diverse use cases, including summarization, data extraction, and RAG. Inspired by RAG architecture, our method integrates retrieval and Natural Language Inference (NLI) models to predict factual consistency between premises and hypotheses using an encoder-based model with only a 512-token context window. Our framework identifies unsupported claims with an F1 score of 0.83 in RAGTruth's response-level classification task, matching methods that trained on the dataset, and outperforming all comparable frameworks using similar-sized models.
Authors: Muhammad Khalifa, Rishabh Agarwal, Lajanugen Logeswaran, Jaekyeom Kim, Hao Peng, Moontae Lee, Honglak Lee, Lu Wang
Abstract: Step-by-step verifiers -- also known as process reward models (PRMs) -- are a key ingredient for test-time scaling. PRMs require step-level supervision, making them expensive to train. This work aims to build data-efficient PRMs as verbalized step-wise reward models that verify every step in the solution by generating a verification chain-of-thought (CoT). We propose ThinkPRM, a long CoT verifier fine-tuned on orders of magnitude fewer process labels than those required by discriminative PRMs. Our approach capitalizes on the inherent reasoning abilities of long CoT models, and outperforms LLM-as-a-Judge and discriminative verifiers -- using only 1% of the process labels in PRM800K -- across several challenging benchmarks. Specifically, ThinkPRM beats the baselines on ProcessBench, MATH-500, and AIME '24 under best-of-N selection and reward-guided search. In an out-of-domain evaluation on a subset of GPQA-Diamond and LiveCodeBench, our PRM surpasses discriminative verifiers trained on the full PRM800K by 8% and 4.5%, respectively. Lastly, under the same token budget, ThinkPRM scales up verification compute more effectively compared to LLM-as-a-Judge, outperforming it by 7.2% on a subset of ProcessBench. Our work highlights the value of generative, long CoT PRMs that can scale test-time compute for verification while requiring minimal supervision for training. Our code, data, and models will be released at https://github.com/mukhal/thinkprm.
Authors: Vasily Ilin, Peter Sushko, Jingwei Hu
Abstract: We propose a deterministic sampling framework using Score-Based Transport Modeling for sampling an unnormalized target density $\pi$ given only its score $\nabla \log \pi$. Our method approximates the Wasserstein gradient flow on $\mathrm{KL}(f_t\|\pi)$ by learning the time-varying score $\nabla \log f_t$ on the fly using score matching. While having the same marginal distribution as Langevin dynamics, our method produces smooth deterministic trajectories, resulting in monotone noise-free convergence. We prove that our method dissipates relative entropy at the same rate as the exact gradient flow, provided sufficient training. Numerical experiments validate our theoretical findings: our method converges at the optimal rate, has smooth trajectories, and is usually more sample efficient than its stochastic counterpart. Experiments on high dimensional image data show that our method produces high quality generations in as few as 15 steps and exhibits natural exploratory behavior. The memory and runtime scale linearly in the sample size.
Authors: Haoyan Xu, Zhengtao Yao, Xuzhi Zhang, Ziyi Wang, Langzhou He, Yushun Dong, Philip S. Yu, Mengyuan Li, Yue Zhao
Abstract: Out-of-distribution (OOD) detection is critical for ensuring the safety and reliability of machine learning systems, particularly in dynamic and open-world environments. In the vision and text domains, zero-shot OOD detection - which requires no training on in-distribution (ID) data - has advanced significantly through the use of large-scale pretrained models, such as vision-language models (VLMs) and large language models (LLMs). However, zero-shot OOD detection in graph-structured data remains largely unexplored, primarily due to the challenges posed by complex relational structures and the absence of powerful, large-scale pretrained models for graphs. In this work, we take the first step toward enabling zero-shot graph OOD detection by leveraging a graph foundation model (GFM). Our experiments show that, when provided only with class label names for both ID and OOD categories, the GFM can effectively perform OOD detection - often surpassing existing "supervised" OOD detection methods that rely on extensive labeled node data. We further address the practical scenario in which OOD label names are not available in real-world settings by introducing GLIP-OOD, a framework that uses LLMs to generate semantically informative pseudo-OOD labels from unlabeled data. These generated OOD labels allow the GFM to better separate ID and OOD classes, facilitating more precise OOD detection - all without any labeled nodes (only ID label names). To our knowledge, this is the first approach to achieve node-level graph OOD detection in a fully zero-shot setting, and it attains performance comparable to state-of-the-art supervised methods on four benchmark text-attributed graph datasets.
Authors: Haoyan Xu, Zhengtao Yao, Ziyi Wang, Zhan Cheng, Xiyang Hu, Mengyuan Li, Yue Zhao
Abstract: Out-of-distribution (OOD) detection in graphs is critical for ensuring model robustness in open-world and safety-sensitive applications. Existing graph OOD detection approaches typically train an in-distribution (ID) classifier on ID data alone, then apply post-hoc scoring to detect OOD instances. While OOD exposure - adding auxiliary OOD samples during training - can improve detection, current graph-based methods often assume access to real OOD nodes, which is often impractical or costly. In this paper, we present GOE-LLM, a framework that leverages Large Language Models (LLMs) to achieve OOD exposure on text-attributed graphs without using any real OOD nodes. GOE-LLM introduces two pipelines: (1) identifying pseudo-OOD nodes from the initially unlabeled graph using zero-shot LLM annotations, and (2) generating semantically informative synthetic OOD nodes via LLM-prompted text generation. These pseudo-OOD nodes are then used to regularize ID classifier training and enhance OOD detection awareness. Empirical results on multiple benchmarks show that GOE-LLM substantially outperforms state-of-the-art methods without OOD exposure, achieving up to a 23.5% improvement in AUROC for OOD detection, and attains performance on par with those relying on real OOD labels for exposure.
Authors: Vishnu Sarukkai, Zhiqiang Xie, Kayvon Fatahalian
Abstract: Improving Large Language Model (LLM) agents for sequential decision-making tasks typically requires extensive task-specific knowledge engineering--custom prompts, curated examples, and specialized observation/action spaces. We investigate a different approach where agents automatically improve by learning from their own successful experiences without human intervention. Our method constructs and refines a database of self-generated trajectories that serve as in-context examples for future tasks. Even naive accumulation of successful trajectories yields substantial performance gains across three diverse benchmarks: ALFWorld (73% to 89%), Wordcraft (55% to 64%), and InterCode-SQL (75% to 79%). These improvements exceed those achieved by upgrading from gpt-4o-mini to gpt-4o and match the performance of allowing multiple attempts per task. We further enhance this approach with two innovations: database-level curation using population-based training to propagate high-performing example collections, and exemplar-level curation that selectively retains trajectories based on their empirical utility as in-context examples. With these enhancements, our method achieves 93% success on ALFWorld--surpassing approaches that use more powerful LLMs and hand-crafted components. Our trajectory bootstrapping technique demonstrates that agents can autonomously improve through experience, offering a scalable alternative to labor-intensive knowledge engineering.
Authors: Weiwei Ye, Zhuopeng Xu, Ning Gui
Abstract: Due to the dynamics of underlying physics and external influences, the uncertainty of time series often varies over time. However, existing Denoising Diffusion Probabilistic Models (DDPMs) often fail to capture this non-stationary nature, constrained by their constant variance assumption from the additive noise model (ANM). In this paper, we innovatively utilize the Location-Scale Noise Model (LSNM) to relax the fixed uncertainty assumption of ANM. A diffusion-based probabilistic forecasting framework, termed Non-stationary Diffusion (NsDiff), is designed based on LSNM that is capable of modeling the changing pattern of uncertainty. Specifically, NsDiff combines a denoising diffusion-based conditional generative model with a pre-trained conditional mean and variance estimator, enabling adaptive endpoint distribution modeling. Furthermore, we propose an uncertainty-aware noise schedule, which dynamically adjusts the noise levels to accurately reflect the data uncertainty at each step and integrates the time-varying variances into the diffusion process. Extensive experiments conducted on nine real-world and synthetic datasets demonstrate the superior performance of NsDiff compared to existing approaches. Code is available at https://github.com/wwy155/NsDiff.
Authors: Shai Feldman, Stephen Bates, Yaniv Romano
Abstract: We introduce a framework for robust uncertainty quantification in situations where labeled training data are corrupted, through noisy or missing labels. We build on conformal prediction, a statistical tool for generating prediction sets that cover the test label with a pre-specified probability. The validity of conformal prediction, however, holds under the i.i.d assumption, which does not hold in our setting due to the corruptions in the data. To account for this distribution shift, the privileged conformal prediction (PCP) method proposed leveraging privileged information (PI) -- additional features available only during training -- to re-weight the data distribution, yielding valid prediction sets under the assumption that the weights are accurate. In this work, we analyze the robustness of PCP to inaccuracies in the weights. Our analysis indicates that PCP can still yield valid uncertainty estimates even when the weights are poorly estimated. Furthermore, we introduce uncertain imputation (UI), a new conformal method that does not rely on weight estimation. Instead, we impute corrupted labels in a way that preserves their uncertainty. Our approach is supported by theoretical guarantees and validated empirically on both synthetic and real benchmarks. Finally, we show that these techniques can be integrated into a triply robust framework, ensuring statistically valid predictions as long as at least one underlying method is valid.
Authors: Wei Chen, Shigui Li, Jiacheng Li, Junmei Yang, John Paisley, Delu Zeng
Abstract: Density ratio estimation is fundamental to tasks involving $f$-divergences, yet existing methods often fail under significantly different distributions or inadequately overlap supports, suffering from the density-chasm and the support-chasm problems. Additionally, prior approaches yield divergent time scores near boundaries, leading to instability. We design $\textbf{D}^3\textbf{RE}$, a unified framework for robust, stable and efficient density ratio estimation. We propose the dequantified diffusion bridge interpolant (DDBI), which expands support coverage and stabilizes time scores via diffusion bridges and Gaussian dequantization. Building on DDBI, the proposed dequantified Schr{\"o}dinger bridge interpolant (DSBI) incorporates optimal transport to solve the Schr{\"o}dinger bridge problem, enhancing accuracy and efficiency. Our method offers uniform approximation and bounded time scores in theory, and outperforms baselines empirically in mutual information and density estimation tasks.
Authors: Winston Chen, Trenton Chang, Jenna Wiens
Abstract: Estimates of heterogeneous treatment assignment effects can inform treatment decisions. Under the presence of non-adherence (e.g., patients do not adhere to their assigned treatment), both the standard backdoor adjustment (SBD) and the conditional front-door adjustment (CFD) can recover unbiased estimates of the treatment assignment effects. However, the estimation variance of these approaches may vary widely across settings, which remains underexplored in the literature. In this work, we demonstrate theoretically and empirically that CFD yields lower-variance estimates than SBD when the true effect of treatment assignment is small (i.e., assigning an intervention leads to small changes in patients' future outcome). Additionally, since CFD requires estimating multiple nuisance parameters, we introduce LobsterNet, a multi-task neural network that implements CFD with joint modeling of the nuisance parameters. Empirically, LobsterNet reduces estimation error across several semi-synthetic and real-world datasets compared to baselines. Our findings suggest CFD with shared nuisance parameter modeling can improve treatment assignment effect estimation under non-adherence.
Authors: Hang Gao, Chenhao Zhang, Tie Wang, Junsuo Zhao, Fengge Wu, Changwen Zheng, Huaping Liu
Abstract: Large Language Models (LLMs) have achieved remarkable success across various domains. However, they still face significant challenges, including high computational costs for training and limitations in solving complex reasoning problems. Although existing methods have extended the reasoning capabilities of LLMs through structured paradigms, these approaches often rely on task-specific prompts and predefined reasoning processes, which constrain their flexibility and generalizability. To address these limitations, we propose a novel framework that leverages graph learning to enable more flexible and adaptive reasoning capabilities for LLMs. Specifically, this approach models the reasoning process of a problem as a graph and employs LLM-based graph learning to guide the adaptive generation of each reasoning step. To further enhance the adaptability of the model, we introduce a Graph Neural Network (GNN) module to perform representation learning on the generated reasoning process, enabling real-time adjustments to both the model and the prompt. Experimental results demonstrate that this method significantly improves reasoning performance across multiple tasks without requiring additional training or task-specific prompt design. Code can be found in https://github.com/zch65458525/L2T.
Authors: Minting Pan, Yitao Zheng, Jiajian Li, Yunbo Wang, Xiaokang Yang
Abstract: Offline reinforcement learning (RL) enables policy optimization using static datasets, avoiding the risks and costs of extensive real-world exploration. However, it struggles with suboptimal offline behaviors and inaccurate value estimation due to the lack of environmental interaction. We present Video-Enhanced Offline RL (VeoRL), a model-based method that constructs an interactive world model from diverse, unlabeled video data readily available online. Leveraging model-based behavior guidance, our approach transfers commonsense knowledge of control policy and physical dynamics from natural videos to the RL agent within the target domain. VeoRL achieves substantial performance gains (over 100% in some cases) across visual control tasks in robotic manipulation, autonomous driving, and open-world video games.
Authors: Xiyuan Wei, Ming Lin, Fanjiang Ye, Fengguang Song, Liangliang Cao, My T. Thai, Tianbao Yang
Abstract: This paper formalizes an emerging learning paradigm that uses a trained model as a reference to guide and enhance the training of a target model through strategic data selection or weighting, named $\textbf{model steering}$. While ad-hoc methods have been used in various contexts, including the training of large foundation models, its underlying principles remain insufficiently understood, leading to sub-optimal performance. In this work, we propose a theory-driven framework for model steering called $\textbf{DRRho risk minimization}$, which is rooted in Distributionally Robust Optimization (DRO). Through a generalization analysis, we provide theoretical insights into why this approach improves generalization and data efficiency compared to training without a reference model. To the best of our knowledge, this is the first time such theoretical insights are provided for the new learning paradigm, which significantly enhance our understanding and practice of model steering. Building on these insights and the connection between contrastive learning and DRO, we introduce a novel method for Contrastive Language-Image Pretraining (CLIP) with a reference model, termed DRRho-CLIP. Extensive experiments validate the theoretical insights, reveal a superior scaling law compared to CLIP without a reference model, and demonstrate its strength over existing heuristic approaches.
Authors: Tianyu Zhang, Shen Dong, O. Deniz Kose, Yanning Shen, Yupeng Zhang
Abstract: With the rise of machine learning techniques, ensuring the fairness of decisions made by machine learning algorithms has become of great importance in critical applications. However, measuring fairness often requires full access to the model parameters, which compromises the confidentiality of the models. In this paper, we propose a solution using zero-knowledge proofs, which allows the model owner to convince the public that a machine learning model is fair while preserving the secrecy of the model. To circumvent the efficiency barrier of naively proving machine learning inferences in zero-knowledge, our key innovation is a new approach to measure fairness only with model parameters and some aggregated information of the input, but not on any specific dataset. To achieve this goal, we derive new bounds for the fairness of logistic regression and deep neural network models that are tighter and better reflecting the fairness compared to prior work. Moreover, we develop efficient zero-knowledge proof protocols for common computations involved in measuring fairness, including the spectral norm of matrices, maximum, absolute value, and fixed-point arithmetic. We have fully implemented our system, FairZK, that proves machine learning fairness in zero-knowledge. Experimental results show that FairZK is significantly faster than the naive approach and an existing scheme that use zero-knowledge inferences as a subroutine. The prover time is improved by 3.1x--1789x depending on the size of the model and the dataset. FairZK can scale to a large model with 47 million parameters for the first time, and generates a proof for its fairness in 343 seconds. This is estimated to be 4 orders of magnitude faster than existing schemes, which only scale to small models with hundreds to thousands of parameters.
Authors: Lu Dai, Wenxuan Zhu, Xuehui Quan, Renzi Meng, Sheng Chai, Yichen Wang
Abstract: To improve the identification of potential anomaly patterns in complex user behavior, this paper proposes an anomaly detection method based on a deep mixture density network. The method constructs a Gaussian mixture model parameterized by a neural network, enabling conditional probability modeling of user behavior. It effectively captures the multimodal distribution characteristics commonly present in behavioral data. Unlike traditional classifiers that rely on fixed thresholds or a single decision boundary, this approach defines an anomaly scoring function based on probability density using negative log-likelihood. This significantly enhances the model's ability to detect rare and unstructured behaviors. Experiments are conducted on the real-world network user dataset UNSW-NB15. A series of performance comparisons and stability validation experiments are designed. These cover multiple evaluation aspects, including Accuracy, F1- score, AUC, and loss fluctuation. The results show that the proposed method outperforms several advanced neural network architectures in both performance and training stability. This study provides a more expressive and discriminative solution for user behavior modeling and anomaly detection. It strongly promotes the application of deep probabilistic modeling techniques in the fields of network security and intelligent risk control.
Authors: Raghav Garg, Kapil Sharma, Karan Gupta
Abstract: Large Language Models (LLMs) hold immense potential for revolutionizing Customer Experience Management (CXM), particularly in contact center operations. However, evaluating their practical utility in complex operational environments is hindered by data scarcity (due to privacy concerns) and the limitations of current benchmarks. Existing benchmarks often lack realism, failing to incorporate deep knowledge base (KB) integration, real-world noise, or critical operational tasks beyond conversational fluency. To bridge this gap, we introduce CXMArena, a novel, large-scale synthetic benchmark dataset specifically designed for evaluating AI in operational CXM contexts. Given the diversity in possible contact center features, we have developed a scalable LLM-powered pipeline that simulates the brand's CXM entities that form the foundation of our datasets-such as knowledge articles including product specifications, issue taxonomies, and contact center conversations. The entities closely represent real-world distribution because of controlled noise injection (informed by domain experts) and rigorous automated validation. Building on this, we release CXMArena, which provides dedicated benchmarks targeting five important operational tasks: Knowledge Base Refinement, Intent Prediction, Agent Quality Adherence, Article Search, and Multi-turn RAG with Integrated Tools. Our baseline experiments underscore the benchmark's difficulty: even state of the art embedding and generation models achieve only 68% accuracy on article search, while standard embedding methods yield a low F1 score of 0.3 for knowledge base refinement, highlighting significant challenges for current models necessitating complex pipelines and solutions over conventional techniques.
Authors: JieHao Wu, Ziwei Wang, Junjie Sheng, Wenhao Li, Xiangfeng Wang, Jun Luo
Abstract: In cloud services, virtual machine (VM) scheduling is a typical Online Dynamic Multidimensional Bin Packing (ODMBP) problem, characterized by large-scale complexity and fluctuating demands. Traditional optimization methods struggle to adapt to real-time changes, domain-expert-designed heuristic approaches suffer from rigid strategies, and existing learning-based methods often lack generalizability and interpretability. To address these limitations, this paper proposes a hierarchical language agent framework named MiCo, which provides a large language model (LLM)-driven heuristic design paradigm for solving ODMBP. Specifically, ODMBP is formulated as a Semi-Markov Decision Process with Options (SMDP-Option), enabling dynamic scheduling through a two-stage architecture, i.e., Option Miner and Option Composer. Option Miner utilizes LLMs to discover diverse and useful non-context-aware strategies by interacting with constructed environments. Option Composer employs LLMs to discover a composing strategy that integrates the non-context-aware strategies with the contextual ones. Extensive experiments on real-world enterprise datasets demonstrate that MiCo achieves a 96.9\% competitive ratio in large-scale scenarios involving more than 10,000 virtual machines. It maintains high performance even under nonstationary request flows and diverse configurations, thus validating its effectiveness in complex and large-scale cloud environments.
Authors: Wujun Zhou, Shu Ding, ZeLin Li, Wei Wang
Abstract: Federated learning enables the clients to collaboratively train a global model, which is aggregated from local models. Due to the heterogeneous data distributions over clients and data privacy in federated learning, it is difficult to train local models to achieve a well-performed global model. In this paper, we introduce the adaptability of local models, i.e., the average performance of local models on data distributions over clients, and enhance the performance of the global model by improving the adaptability of local models. Since each client does not know the data distributions over other clients, the adaptability of the local model cannot be directly optimized. First, we provide the property of an appropriate local model which has good adaptability on the data distributions over clients. Then, we formalize the property into the local training objective with a constraint and propose a feasible solution to train the local model. Extensive experiments on federated learning benchmarks demonstrate that our method significantly improves the adaptability of local models and achieves a well-performed global model that consistently outperforms the baseline methods.
Authors: Xiangwen Zhuge, Xu Shen, Zeyu Wang, Fan Dang, Xuan Ding, Danyang Li, Yahui Han, Tianxiang Hao, Zheng Yang
Abstract: Efficient LLM inference on resource-constrained devices presents significant challenges in compute and memory utilization. Due to limited GPU memory, existing systems offload model weights to CPU memory, incurring substantial I/O overhead between the CPU and GPU. This leads to two major inefficiencies: (1) GPU cores are underutilized, often remaining idle while waiting for data to be loaded; and (2) GPU memory has low impact on performance, as reducing its capacity has minimal effect on overall throughput.In this paper, we propose SpecOffload, a high-throughput inference engine that embeds speculative decoding into offloading. Our key idea is to unlock latent GPU resources for storing and executing a draft model used for speculative decoding, thus accelerating inference at near-zero additional cost. To support this, we carefully orchestrate the interleaved execution of target and draft models in speculative decoding within the offloading pipeline, and propose a planner to manage tensor placement and select optimal parameters. Compared to the best baseline, SpecOffload improves GPU core utilization by 4.49x and boosts inference throughput by 2.54x. Our code is available at https://github.com/MobiSense/SpecOffload .
Authors: Jingyao Wang, Wenwen Qiang, Zeen Song, Changwen Zheng, Hui Xiong
Abstract: Large language models (LLMs) excel at complex tasks thanks to advances in reasoning abilities. However, existing methods overlook the trade-off between reasoning effectiveness and computational efficiency, often encouraging unnecessarily long reasoning chains and wasting tokens. To address this, we propose Learning to Think (L2T), an information-theoretic reinforcement fine-tuning framework for LLMs to make the models achieve optimal reasoning with fewer tokens. Specifically, L2T treats each query-response interaction as a hierarchical session of multiple episodes and proposes a universal dense process reward, i.e., quantifies the episode-wise information gain in parameters, requiring no extra annotations or task-specific evaluators. We propose a method to quickly estimate this reward based on PAC-Bayes bounds and the Fisher information matrix. Theoretical analyses show that it significantly reduces computational complexity with high estimation accuracy. By immediately rewarding each episode's contribution and penalizing excessive updates, L2T optimizes the model via reinforcement learning to maximize the use of each episode and achieve effective updates. Empirical results on various reasoning benchmarks and base models demonstrate the advantage of L2T across different tasks, boosting both reasoning effectiveness and efficiency.
Authors: Yizhou Liu, Ziming Liu, Jeff Gore
Abstract: The success of today's large language models (LLMs) depends on the observation that larger models perform better. However, the origin of this neural scaling law -- the finding that loss decreases as a power law with model size -- remains unclear. Starting from two empirical principles -- that LLMs represent more things than the model dimensions (widths) they have (i.e., representations are superposed), and that words or concepts in language occur with varying frequencies -- we constructed a toy model to study the loss scaling with model size. We found that when superposition is weak, meaning only the most frequent features are represented without interference, the scaling of loss with model size depends on the underlying feature frequency; if feature frequencies follow a power law, so does the loss. In contrast, under strong superposition, where all features are represented but overlap with each other, the loss becomes inversely proportional to the model dimension across a wide range of feature frequency distributions. This robust scaling behavior is explained geometrically: when many more vectors are packed into a lower dimensional space, the interference (squared overlaps) between vectors scales inversely with that dimension. We then analyzed four families of open-sourced LLMs and found that they exhibit strong superposition and quantitatively match the predictions of our toy model. The Chinchilla scaling law turned out to also agree with our results. We conclude that representation superposition is an important mechanism underlying the observed neural scaling laws. We anticipate that these insights will inspire new training strategies and model architectures to achieve better performance with less computation and fewer parameters.
Authors: Mugilan Ganesan, Shane Segal, Ankur Aggarwal, Nish Sinnadurai, Sean Lie, Vithursan Thangarasa
Abstract: Speculative decoding significantly accelerates language model inference by enabling a lightweight draft model to propose multiple tokens that a larger target model verifies simultaneously. However, applying this technique to vision-language models (VLMs) presents two fundamental challenges: small language models that could serve as efficient drafters lack the architectural components to process visual inputs, and their token predictions fail to match those of VLM target models that consider visual context. We introduce Multimodal Adaptation and Self-Data Distillation for Speculative Decoding of Vision-Language Models (MASSV), which transforms existing small language models into effective multimodal drafters through a two-phase approach. MASSV first connects the target VLM's vision encoder to the draft model via a lightweight trainable projector, then applies self-distilled visual instruction tuning using responses generated by the target VLM to align token predictions. Comprehensive experiments across the Qwen2.5-VL and Gemma3 model families demonstrate that MASSV increases accepted length by up to 30% and delivers end-to-end inference speedups of up to 1.46x on visually-grounded tasks. MASSV provides a scalable, architecture-compatible method for accelerating both current and future VLMs.
Authors: Jiazheng Zhang, Wenqing Jing, Zizhuo Zhang, Zhiheng Xi, Shihan Dou, Rongxiang Weng, Jiahuan Li, Jingang Wang, Mingxu Chai, Shibo Hong, Tao Gui, Qi Zhang
Abstract: Reward models (RMs) play a pivotal role in aligning large language models (LLMs) with human values. However, noisy preferences in human feedback can lead to reward misgeneralization - a phenomenon where reward models learn spurious correlations or overfit to noisy preferences, which poses important challenges to the generalization of RMs. This paper systematically analyzes the characteristics of preference pairs and aims to identify how noisy preferences differ from human-aligned preferences in reward modeling. Our analysis reveals that noisy preferences are difficult for RMs to fit, as they cause sharp training fluctuations and irregular gradient updates. These distinctive dynamics suggest the feasibility of identifying and excluding such noisy preferences. Empirical studies demonstrate that policy LLM optimized with a reward model trained on the full preference dataset, which includes substantial noise, performs worse than the one trained on a subset of exclusively high quality preferences. To address this challenge, we propose an online Collaborative Reward Modeling (CRM) framework to achieve robust preference learning through peer review and curriculum learning. In particular, CRM maintains two RMs that collaboratively filter potential noisy preferences by peer-reviewing each other's data selections. Curriculum learning synchronizes the capabilities of two models, mitigating excessive disparities to promote the utility of peer review. Extensive experiments demonstrate that CRM significantly enhances RM generalization, with up to 9.94 points improvement on RewardBench under an extreme 40\% noise. Moreover, CRM can seamlessly extend to implicit-reward alignment methods, offering a robust and versatile alignment strategy.
Authors: Justin Sanders, Melih Yilmaz, Jacob H. Russell, Wout Bittremieux, William E. Fondrie, Nicholas M. Riley, Sewoong Oh, William Stafford Noble
Abstract: Mass spectrometry is the dominant technology in the field of proteomics, enabling high-throughput analysis of the protein content of complex biological samples. Due to the complexity of the instrumentation and resulting data, sophisticated computational methods are required for the processing and interpretation of acquired mass spectra. Machine learning has shown great promise to improve the analysis of mass spectrometry data, with numerous purpose-built methods for improving specific steps in the data acquisition and analysis pipeline reaching widespread adoption. Here, we propose unifying various spectrum prediction tasks under a single foundation model for mass spectra. To this end, we pre-train a spectrum encoder using de novo sequencing as a pre-training task. We then show that using these pre-trained spectrum representations improves our performance on the four downstream tasks of spectrum quality prediction, chimericity prediction, phosphorylation prediction, and glycosylation status prediction. Finally, we perform multi-task fine-tuning and find that this approach improves the performance on each task individually. Overall, our work demonstrates that a foundation model for tandem mass spectrometry proteomics trained on de novo sequencing learns generalizable representations of spectra, improves performance on downstream tasks where training data is limited, and can ultimately enhance data acquisition and analysis in proteomics experiments.
Authors: Kehan Long, Jorge Cort\'es, Nikolay Atanasov
Abstract: We study the problem of certifying the stability of closed-loop systems under control policies derived from optimal control or reinforcement learning (RL). Classical Lyapunov methods require a strict step-wise decrease in the Lyapunov function but such a certificate is difficult to construct for a learned control policy. The value function associated with an RL policy is a natural Lyapunov function candidate but it is not clear how it should be modified. To gain intuition, we first study the linear quadratic regulator (LQR) problem and make two key observations. First, a Lyapunov function can be obtained from the value function of an LQR policy by augmenting it with a residual term related to the system dynamics and stage cost. Second, the classical Lyapunov decrease requirement can be relaxed to a generalized Lyapunov condition requiring only decrease on average over multiple time steps. Using this intuition, we consider the nonlinear setting and formulate an approach to learn generalized Lyapunov functions by augmenting RL value functions with neural network residual terms. Our approach successfully certifies the stability of RL policies trained on Gymnasium and DeepMind Control benchmarks. We also extend our method to jointly train neural controllers and stability certificates using a multi-step Lyapunov loss, resulting in larger certified inner approximations of the region of attraction compared to the classical Lyapunov approach. Overall, our formulation enables stability certification for a broad class of systems with learned policies by making certificates easier to construct, thereby bridging classical control theory and modern learning-based methods.
Authors: Qiuyu Zhu, Liang Zhang, Qianxiong Xu, Cheng Long, Jie Zhang
Abstract: Graph-based retrieval-augmented generation (RAG) enables large language models (LLMs) to incorporate structured knowledge via graph retrieval as contextual input, enhancing more accurate and context-aware reasoning. We observe that for different queries, it could retrieve similar subgraphs as prompts, and thus we propose SubGCache, which aims to reduce inference latency by reusing computation across queries with similar structural prompts (i.e., subgraphs). Specifically, SubGCache clusters queries based on subgraph embeddings, constructs a representative subgraph for each cluster, and pre-computes the key-value (KV) cache of the representative subgraph. For each query with its retrieved subgraph within a cluster, it reuses the pre-computed KV cache of the representative subgraph of the cluster without computing the KV tensors again for saving computation. Experiments on two new datasets across multiple LLM backbones and graph-based RAG frameworks demonstrate that SubGCache consistently reduces inference latency with comparable and even improved generation quality, achieving up to 6.68$\times$ reduction in time-to-first-token (TTFT).
Authors: Chenhong Zhou, Jie Chen, Zaifeng Yang, Ching Eng Png
Abstract: Physics-informed neural networks (PINNs) have emerged as a new learning paradigm for solving partial differential equations (PDEs) by enforcing the constraints of physical equations, boundary conditions (BCs), and initial conditions (ICs) into the loss function. Despite their successes, vanilla PINNs still suffer from poor accuracy and slow convergence due to the intractable multi-objective optimization issue. In this paper, we propose a novel Dual-Balanced PINN (DB-PINN), which dynamically adjusts loss weights by integrating inter-balancing and intra-balancing to alleviate two imbalance issues in PINNs. Inter-balancing aims to mitigate the gradient imbalance between PDE residual loss and condition-fitting losses by determining an aggregated weight that offsets their gradient distribution discrepancies. Intra-balancing acts on condition-fitting losses to tackle the imbalance in fitting difficulty across diverse conditions. By evaluating the fitting difficulty based on the loss records, intra-balancing can allocate the aggregated weight proportionally to each condition loss according to its fitting difficulty level. We further introduce a robust weight update strategy to prevent abrupt spikes and arithmetic overflow in instantaneous weight values caused by large loss variances, enabling smooth weight updating and stable training. Extensive experiments demonstrate that DB-PINN achieves significantly superior performance than those popular gradient-based weighting methods in terms of convergence speed and prediction accuracy. Our code and supplementary material are available at https://github.com/chenhong-zhou/DualBalanced-PINNs.
Authors: Simone Azeglio, Arianna Di Bernardo
Abstract: Recent advances in diffusion models have demonstrated their remarkable ability to capture complex image distributions, but the geometric properties of the learned data manifold remain poorly understood. We address this gap by introducing a score-based Riemannian metric that leverages the Stein score function from diffusion models to characterize the intrinsic geometry of the data manifold without requiring explicit parameterization. Our approach defines a metric tensor in the ambient space that stretches distances perpendicular to the manifold while preserving them along tangential directions, effectively creating a geometry where geodesics naturally follow the manifold's contours. We develop efficient algorithms for computing these geodesics and demonstrate their utility for both interpolation between data points and extrapolation beyond the observed data distribution. Through experiments on synthetic data with known geometry, Rotated MNIST, and complex natural images via Stable Diffusion, we show that our score-based geodesics capture meaningful transformations that respect the underlying data distribution. Our method consistently outperforms baseline approaches on perceptual metrics (LPIPS) and distribution-level metrics (FID, KID), producing smoother, more realistic image transitions. These results reveal the implicit geometric structure learned by diffusion models and provide a principled way to navigate the manifold of natural images through the lens of Riemannian geometry.
Authors: Wilson Wongso, Hao Xue, Flora D. Salim
Abstract: Understanding human mobility through Point-of-Interest (POI) recommendation is increasingly important for applications such as urban planning, personalized services, and generative agent simulation. However, progress in this field is hindered by two key challenges: the over-reliance on older datasets from 2012-2013 and the lack of reproducible, city-level check-in datasets that reflect diverse global regions. To address these gaps, we present Massive-STEPS (Massive Semantic Trajectories for Understanding POI Check-ins), a large-scale, publicly available benchmark dataset built upon the Semantic Trails dataset and enriched with semantic POI metadata. Massive-STEPS spans 12 geographically and culturally diverse cities and features more recent (2017-2018) and longer-duration (24 months) check-in data than prior datasets. We benchmarked a wide range of POI recommendation models on Massive-STEPS using both supervised and zero-shot approaches, and evaluated their performance across multiple urban contexts. By releasing Massive-STEPS, we aim to facilitate reproducible and equitable research in human mobility and POI recommendation. The dataset and benchmarking code are available at: https://github.com/cruiseresearchgroup/Massive-STEPS
Authors: Chao Jin, Ziheng Jiang, Zhihao Bai, Zheng Zhong, Juncai Liu, Xiang Li, Ningxin Zheng, Xi Wang, Cong Xie, Qi Huang, Wen Heng, Yiyuan Ma, Wenlei Bao, Size Zheng, Yanghua Peng, Haibin Lin, Xuanzhe Liu, Xin Jin, Xin Liu
Abstract: We present MegaScale-MoE, a production system tailored for the efficient training of large-scale mixture-of-experts (MoE) models. MoE emerges as a promising architecture to scale large language models (LLMs) to unprecedented sizes, thereby enhancing model performance. However, existing MoE training systems experience a degradation in training efficiency, exacerbated by the escalating scale of MoE models and the continuous evolution of hardware. Recognizing the pivotal role of efficient communication in enhancing MoE training, MegaScale-MoE customizes communication-efficient parallelism strategies for attention and FFNs in each MoE layer and adopts a holistic approach to overlap communication with computation at both inter- and intra-operator levels. Additionally, MegaScale-MoE applies communication compression with adjusted communication patterns to lower precision, further improving training efficiency. When training a 352B MoE model on 1,440 NVIDIA Hopper GPUs, MegaScale-MoE achieves a training throughput of 1.41M tokens/s, improving the efficiency by 1.88$\times$ compared to Megatron-LM. We share our operational experience in accelerating MoE training and hope that by offering our insights in system design, this work will motivate future research in MoE systems.
Authors: Riccardo Grazzi, Massimiliano Pontil, Saverio Salzo
Abstract: Bilevel optimization problems are receiving increasing attention in machine learning as they provide a natural framework for hyperparameter optimization and meta-learning. A key step to tackle these problems is the efficient computation of the gradient of the upper-level objective (hypergradient). In this work, we study stochastic approximation schemes for the hypergradient, which are important when the lower-level problem is empirical risk minimization on a large dataset. The method that we propose is a stochastic variant of the approximate implicit differentiation approach in (Pedregosa, 2016). We provide bounds for the mean square error of the hypergradient approximation, under the assumption that the lower-level problem is accessible only through a stochastic mapping which is a contraction in expectation. In particular, our main bound is agnostic to the choice of the two stochastic solvers employed by the procedure. We provide numerical experiments to support our theoretical analysis and to show the advantage of using stochastic hypergradients in practice.
Authors: Fei Wang, Tingting Zhang, Xilei Wu, Xin Wang, Pengcheng Wang, Han Ding, Jingang Shi, Jinsong Han, Dong Huang
Abstract: Hand hygiene is among the most effective daily practices for preventing infectious diseases such as influenza, malaria, and skin infections. While professional guidelines emphasize proper handwashing to reduce the risk of viral infections, surveys reveal that adherence to these recommendations remains low. To address this gap, we propose UWash, a wearable solution leveraging smartwatches to evaluate handwashing procedures, aiming to raise awareness and cultivate high-quality handwashing habits. We frame the task of handwashing assessment as an action segmentation problem, similar to those in computer vision, and introduce a simple yet efficient two-stream UNet-like network to achieve this goal. Experiments involving 51 subjects demonstrate that UWash achieves 92.27% accuracy in handwashing gesture recognition, an error of <0.5 seconds in onset/offset detection, and an error of <5 points in gesture scoring under user-dependent settings. The system also performs robustly in user-independent and user-independent-location-independent evaluations. Remarkably, UWash maintains high performance in real-world tests, including evaluations with 10 random passersby at a hospital 9 months later and 10 passersby in an in-the-wild test conducted 2 years later. UWash is the first system to score handwashing quality based on gesture sequences, offering actionable guidance for improving daily hand hygiene. The code and dataset are publicly available at https://github.com/aiotgroup/UWash
Authors: Andreas Kontogiannis, Dimitrios Kelesis, Vasilis Pollatos, George Giannakopoulos, Georgios Paliouras
Abstract: A focused crawler aims at discovering as many web pages and web sites relevant to a target topic as possible, while avoiding irrelevant ones. Reinforcement Learning (RL) has been a promising direction for optimizing focused crawling, because RL can naturally optimize the long-term profit of discovering relevant web locations within the context of a reward. In this paper, we propose TRES, a novel RL-empowered framework for focused crawling that aims at maximizing both the number of relevant web pages (aka \textit{harvest rate}) and the number of relevant web sites (\textit{domains}). We model the focused crawling problem as a novel Markov Decision Process (MDP), which the RL agent aims to solve by determining an optimal crawling strategy. To overcome the computational infeasibility of exhaustively searching for the best action at each time step, we propose Tree-Frontier, a provably efficient tree-based sampling algorithm that adaptively discretizes the large state and action spaces and evaluates only a few representative actions. Experimentally, utilizing online real-world data, we show that TRES significantly outperforms and Pareto-dominates state-of-the-art methods in terms of harvest rate and the number of retrieved relevant domains, while it provably reduces by orders of magnitude the number of URLs needed to be evaluated at each crawling step.
Authors: Jan Kronqvist, Ruth Misener, Calvin Tsay
Abstract: We develop a class of mixed-integer formulations for disjunctive constraints intermediate to the big-M and convex hull formulations in terms of relaxation strength. The main idea is to capture the best of both the big-M and convex hull formulations: a computationally light formulation with a tight relaxation. The "P-split" formulations are based on a lifted transformation that splits convex additively separable constraints into P partitions and forms the convex hull of the linearized and partitioned disjunction. The "P-split" formulations are derived for disjunctive constraints with convex constraints within each disjunct, and we generalize the results for the case with nonconvex constraints within the disjuncts. We analyze the continuous relaxation of the P-split formulations and show that, under certain assumptions, the formulations form a hierarchy starting from a big-M equivalent and converging to the convex hull. We computationally compare the P-split formulations against big-M and convex hull formulations on 344 test instances. The test problems include K-means clustering, semi-supervised clustering, P_ball problems, and optimization over trained ReLU neural networks. The computational results show promising potential of the P-split formulations. For many of the test problems, P-split formulations are solved with a similar number of explored nodes as the convex hull formulation, while reducing the solution time by an order of magnitude and outperforming big-M both in time and number of explored nodes.
Authors: Max Biggs, Georgia Perakis
Abstract: We focus on modeling the relationship between an input feature vector and the predicted outcome of a trained decision tree using mixed-integer optimization. This can be used in many practical applications where a decision tree or tree ensemble is incorporated into an optimization problem to model the predicted outcomes of a decision. We propose tighter mixed-integer optimization formulations than those previously introduced. Existing formulations can be shown to have linear relaxations that have fractional extreme points, even for the simple case of modeling a single decision tree. A formulation we propose, based on a projected union of polyhedra approach, is ideal for a single decision tree. While the formulation is generally not ideal for tree ensembles or if additional constraints are added, it generally has fewer extreme points, leading to a faster time to solve, particularly if the formulation has relatively few trees. However, previous work has shown that formulations based on a binary representation of the feature vector perform well computationally and hence are attractive for use in practical applications. We present multiple approaches to tighten existing formulations with binary vectors, and show that fractional extreme points are removed when there are multiple splits on the same feature. At an extreme, we prove that this results in ideal formulations for tree ensembles modeling a one-dimensional feature vector. Building on this result, we also show via numerical simulations that these additional constraints result in significantly tighter linear relaxations when the feature vector is low dimensional. We also present instances where the time to solve to optimality is significantly improved using these formulations.
Authors: Jo\~ao Vitorino, Tiago Dias, Tiago Fonseca, Eva Maia, Isabel Pra\c{c}a
Abstract: It is imperative to safeguard computer applications and information systems against the growing number of cyber-attacks. Automated software testing tools can be developed to quickly analyze many lines of code and detect vulnerabilities by generating function-specific testing data. This process draws similarities to the constrained adversarial examples generated by adversarial machine learning methods, so there could be significant benefits to the integration of these methods in testing tools to identify possible attack vectors. Therefore, this literature review is focused on the current state-of-the-art of constrained data generation approaches applied for adversarial learning and software testing, aiming to guide researchers and developers to enhance their software testing tools with adversarial testing methods and improve the resilience and robustness of their information systems. The found approaches were systematized, and the advantages and limitations of those specific for white-box, grey-box, and black-box testing were analyzed, identifying research gaps and opportunities to automate the testing tools with data generated by adversarial attacks.
Authors: Zeyuan Allen-Zhu, Yuanzhi Li
Abstract: Transformer-based language models are effective but complex, and understanding their inner workings and reasoning mechanisms is a significant challenge. Previous research has primarily explored how these models handle simple tasks like name copying or selection, and we extend this by investigating how these models perform recursive language structure reasoning defined by context-free grammars (CFGs). We introduce a family of synthetic CFGs that produce hierarchical rules, capable of generating lengthy sentences (e.g., hundreds of tokens) that are locally ambiguous and require dynamic programming to parse. Despite this complexity, we demonstrate that generative models like GPT can accurately learn and reason over CFG-defined hierarchies and generate sentences based on it. We explore the model's internals, revealing that its hidden states precisely capture the structure of CFGs, and its attention patterns resemble the information passing in a dynamic programming algorithm. This paper also presents several corollaries, including showing why absolute positional embeddings is inferior to relative and rotary embeddings; uniform attention alone is surprisingly effective (motivating our follow-up work on Canon layers); encoder-only models (e.g., BERT, DeBERTa) struggle with deep structure reasoning on CFGs compared to autoregressive models (e.g., GPT); and injecting structural or syntactic noise into pretraining data markedly improves robustness to corrupted language prompts.
Authors: Zinan Lin, Sivakanth Gopi, Janardhan Kulkarni, Harsha Nori, Sergey Yekhanin
Abstract: Generating differentially private (DP) synthetic data that closely resembles the original private data is a scalable way to mitigate privacy concerns in the current data-driven world. In contrast to current practices that train customized models for this task, we aim to generate DP Synthetic Data via APIs (DPSDA), where we treat foundation models as blackboxes and only utilize their inference APIs. Such API-based, training-free approaches are easier to deploy as exemplified by the recent surge in the number of API-based apps. These approaches can also leverage the power of large foundation models which are only accessible via their inference APIs. However, this comes with greater challenges due to strictly more restrictive model access and the need to protect privacy from the API provider. In this paper, we present a new framework called Private Evolution (PE) to solve this problem and show its initial promise on synthetic images. Surprisingly, PE can match or even outperform state-of-the-art (SOTA) methods without any model training. For example, on CIFAR10 (with ImageNet as the public data), we achieve FID <= 7.9 with privacy cost {\epsilon} = 0.67, significantly improving the previous SOTA from {\epsilon} = 32. We further demonstrate the promise of applying PE on large foundation models such as Stable Diffusion to tackle challenging private datasets with a small number of high-resolution images. The code and data are released at https://github.com/microsoft/DPSDA.
Authors: Minglei Yin, Bin Liu, Neil Zhenqiang Gong, Xin Li
Abstract: With rich visual data, such as images, becoming readily associated with items, visually-aware recommendation systems (VARS) have been widely used in different applications. Recent studies have shown that VARS are vulnerable to item-image adversarial attacks, which add human-imperceptible perturbations to the clean images associated with those items. Attacks on VARS pose new security challenges to a wide range of applications such as e-Commerce and social networks where VARS are widely used. How to secure VARS from such adversarial attacks becomes a critical problem. Currently, there is still a lack of systematic study on how to design secure defense strategies against visual attacks on VARS. In this paper, we attempt to fill this gap by proposing an adversarial image reconstruction and detection framework to secure VARS. Our proposed method can simultaneously (1) secure VARS from adversarial attacks characterized by local perturbations by image reconstruction based on global vision transformers; and (2) accurately detect adversarial examples using a novel contrastive learning approach. Meanwhile, our framework is designed to be used as both a filter and a detector so that they can be jointly trained to improve the flexibility of our defense strategy to a variety of attacks and VARS models. We have conducted extensive experimental studies with two popular attack methods (FGSM and PGD). Our experimental results on two real-world datasets show that our defense strategy against visual attacks is effective and outperforms existing methods on different attacks. Moreover, our method can detect adversarial examples with high accuracy.
Authors: Jirui Qi, Raquel Fern\'andez, Arianna Bisazza
Abstract: Multilingual large-scale Pretrained Language Models (PLMs) have been shown to store considerable amounts of factual knowledge, but large variations are observed across languages. With the ultimate goal of ensuring that users with different language backgrounds obtain consistent feedback from the same model, we study the cross-lingual consistency (CLC) of factual knowledge in various multilingual PLMs. To this end, we propose a Ranking-based Consistency (RankC) metric to evaluate knowledge consistency across languages independently from accuracy. Using this metric, we conduct an in-depth analysis of the determining factors for CLC, both at model level and at language-pair level. Among other results, we find that increasing model size leads to higher factual probing accuracy in most languages, but does not improve cross-lingual consistency. Finally, we conduct a case study on CLC when new factual associations are inserted in the PLMs via model editing. Results on a small sample of facts inserted in English reveal a clear pattern whereby the new piece of knowledge transfers only to languages with which English has a high RankC score.
Authors: Chao Chen, Tian Zhou, Yanjun Zhao, Hui Liu, Liang Sun, Rong Jin
Abstract: Spatio-temporal forecasting is crucial in various fields and requires a careful balance between identifying subtle patterns and filtering out noise. Vector quantization (VQ) appears well-suited for this purpose, as it quantizes input vectors into a set of codebook vectors or patterns. Although VQ has shown promise in various computer vision tasks, it surprisingly falls short in enhancing the accuracy of spatio-temporal forecasting. We attribute this to two main issues: inaccurate optimization due to non-differentiability and limited representation power in hard-VQ. To tackle these challenges, we introduce Differentiable Sparse Soft-Vector Quantization (SVQ), the first VQ method to enhance spatio-temporal forecasting. SVQ balances detail preservation with noise reduction, offering full differentiability and a solid foundation in sparse regression. Our approach employs a two-layer MLP and an extensive codebook to streamline the sparse regression process, significantly cutting computational costs while simplifying training and improving performance. Empirical studies on five spatio-temporal benchmark datasets show SVQ achieves state-of-the-art results, including a 7.9% improvement on the WeatherBench-S temperature dataset and an average mean absolute error reduction of 9.4% in video prediction benchmarks (Human3.6M, KTH, and KittiCaltech), along with a 17.3% enhancement in image quality (LPIPS). Code is publicly available at https://github.com/Pachark/SVQ-Forecasting.
Authors: Suhas Thejaswi, Ameet Gadekar, Bruno Ordozgoiti, Aristides Gionis
Abstract: In this work, we study diversity-aware clustering problems where the data points are associated with multiple attributes resulting in intersecting groups. A clustering solution needs to ensure that the number of chosen cluster centers from each group should be within the range defined by a lower and upper bound threshold for each group, while simultaneously minimizing the clustering objective, which can be either $k$-median, $k$-means or $k$-supplier. We study the computational complexity of the proposed problems, offering insights into their NP-hardness, polynomial-time inapproximability, and fixed-parameter intractability. We present parameterized approximation algorithms with approximation ratios $1+ \frac{2}{e} + \epsilon \approx 1.736$, $1+\frac{8}{e} + \epsilon \approx 3.943$, and $5$ for diversity-aware $k$-median, diversity-aware $k$-means and diversity-aware $k$-supplier, respectively. Assuming Gap-ETH, the approximation ratios are tight for the diversity-aware $k$-median and diversity-aware $k$-means problems. Our results imply the same approximation factors for their respective fair variants with disjoint groups -- fair $k$-median, fair $k$-means, and fair $k$-supplier -- with lower bound requirements.
Authors: Branislav Pecher, Ivan Srba, Maria Bielikova
Abstract: When solving NLP tasks with limited labelled data, researchers typically either use a general large language model without further update, or use a small number of labelled samples to tune a specialised smaller model. In this work, we answer an important question -- how many labelled samples are required for the specialised small models to outperform general large models, while taking the performance variance into consideration. By observing the behaviour of fine-tuning, instruction-tuning, prompting and in-context learning on 8 language models, we identify such performance break-even points across 8 representative text classification tasks of varying characteristics. We show that the specialised models often need only few samples (on average $100$) to be on par or better than the general ones. At the same time, the number of required labels strongly depends on the dataset or task characteristics, with fine-tuning on binary datasets requiring significantly more samples. When performance variance is taken into consideration, the number of required labels increases on average by $100 - 200\%$. Finally, larger models do not consistently lead to better performance and lower variance, with 4-bit quantisation having negligible impact.
Authors: Teresa Yeo, Andrei Atanov, Harold Benoit, Aleksandr Alekseev, Ruchira Ray, Pooya Esmaeil Akhoondi, Amir Zamir
Abstract: We present a method to control a text-to-image generative model to produce training data useful for supervised learning. Unlike previous works that employ an open-loop approach and pre-define prompts to generate new data using either a language model or human expertise, we develop an automated closed-loop system which involves two feedback mechanisms. The first mechanism uses feedback from a given supervised model and finds adversarial prompts that result in image generations that maximize the model loss. While these adversarial prompts result in diverse data informed by the model, they are not informed of the target distribution, which can be inefficient. Therefore, we introduce the second feedback mechanism that guides the generation process towards a certain target distribution. We call the method combining these two mechanisms Guided Adversarial Prompts. We perform our evaluations on different tasks, datasets and architectures, with different types of distribution shifts (spuriously correlated data, unseen domains) and demonstrate the efficiency of the proposed feedback mechanisms compared to open-loop approaches.
Authors: Jaeyong Bae, Hawoong Jeong
Abstract: Bridging the gap between the practical performance of deep learning and its theoretical foundations often involves analyzing neural networks through stochastic gradient descent (SGD). Expanding on previous research that focused on modeling structured inputs under a simple Gaussian setting, we analyze the behavior of a deep learning system trained on inputs modeled as Gaussian mixtures to better simulate more general structured inputs. Through empirical analysis and theoretical investigation, we demonstrate that under certain standardization schemes, the deep learning model converges toward Gaussian setting behavior, even when the input data follow more complex or real-world distributions. This finding exhibits a form of universality in which diverse structured distributions yield results consistent with Gaussian assumptions, which can support the theoretical understanding of deep learning models.
Authors: Changliang Zhou, Xi Lin, Zhenkun Wang, Xialiang Tong, Mingxuan Yuan, Qingfu Zhang
Abstract: The neural combinatorial optimization (NCO) method has shown great potential for solving routing problems of intelligent transportation systems without requiring expert knowledge. However, existing constructive NCO methods still struggle to solve large-scale instances, which significantly limits their application prospects. To address these crucial shortcomings, this work proposes a novel Instance-Conditioned Adaptation Model (ICAM) for better large-scale generalization of neural routing solvers. In particular, we design a simple yet efficient instance-conditioned adaptation function to significantly improve the generalization performance of existing NCO models with a small time and memory overhead. In addition, with a systematic investigation on the performance of information incorporation between different attention mechanisms, we further propose a powerful yet low-complexity instance-conditioned adaptation module to generate better solutions for instances across different scales. Extensive experimental results on both synthetic and benchmark instances show that our proposed method is capable of obtaining promising results with a very fast inference time in solving large-scale Traveling Salesman Problems (TSPs), Capacitated Vehicle Routing Problems (CVRPs), and Asymmetric Traveling Salesman Problems (ATSPs). Our code is available at https://github.com/CIAM-Group/ICAM.
Authors: Jonas Katona, Xiuyuan Wang, Andre Wibisono
Abstract: Motivated by understanding the behavior of the Alternating Mirror Descent (AMD) algorithm for bilinear zero-sum games, we study the discretization of continuous-time Hamiltonian flow via the symplectic Euler method. We provide a framework for analysis using results from Hamiltonian dynamics, Lie algebra, and symplectic numerical integrators, with an emphasis on the existence and properties of a conserved quantity, the modified Hamiltonian (MH), for the symplectic Euler method. We compute the MH in closed-form when the original Hamiltonian is a quadratic function, and show that it generally differs from the other conserved quantity known previously in that case. We derive new error bounds on the MH when truncated at orders in the stepsize in terms of the number of iterations, $K$, and use these bounds to show an improved $\mathcal{O}(K^{1/5})$ total regret bound and an $\mathcal{O}(K^{-4/5})$ duality gap of the average iterates for AMD. Finally, we propose a conjecture which, if true, would imply that the total regret for AMD scales as $\mathcal{O}\left(K^{\varepsilon}\right)$ and the duality gap of the average iterates as $\mathcal{O}\left(K^{-1+\varepsilon}\right)$ for any $\varepsilon>0$, and we can take $\varepsilon=0$ upon certain convergence conditions for the MH.
Authors: Christoforos Vasilatos, Dunia J. Mahboobeh, Hithem Lamri, Manaar Alam, Michail Maniatakos
Abstract: Industrial Control Systems (ICS) are extensively used in critical infrastructures ensuring efficient, reliable, and continuous operations. However, their increasing connectivity and addition of advanced features make them vulnerable to cyber threats, potentially leading to severe disruptions in essential services. In this context, honeypots play a vital role by acting as decoy targets within ICS networks, or on the Internet, helping to detect, log, analyze, and develop mitigations for ICS-specific cyber threats. Deploying ICS honeypots, however, is challenging due to the necessity of accurately replicating industrial protocols and device characteristics, a crucial requirement for effectively mimicking the unique operational behavior of different industrial systems. Moreover, this challenge is compounded by the significant manual effort required in also mimicking the control logic the PLC would execute, in order to capture attacker traffic aiming to disrupt critical infrastructure operations. In this paper, we propose LLMPot, a novel approach for designing honeypots in ICS networks harnessing the potency of Large Language Models (LLMs). LLMPot aims to automate and optimize the creation of realistic honeypots with vendor-agnostic configurations, and for any control logic, aiming to eliminate the manual effort and specialized knowledge traditionally required in this domain. We conducted extensive experiments focusing on a wide array of parameters, demonstrating that our LLM-based approach can effectively create honeypot devices implementing different industrial protocols and diverse control logic.
Authors: Salma J. Ahmed, Emad A. Mohammed
Abstract: Developing new drugs is laborious and costly, demanding extensive time investment. In this paper, we introduce a de-novo drug design strategy, which harnesses the capabilities of language models to devise targeted drugs for specific proteins. Employing a Reinforcement Learning (RL) framework utilizing Proximal Policy Optimization (PPO), we refine the model to acquire a policy for generating drugs tailored to protein targets. The proposed method integrates a composite reward function, combining considerations of drug-target interaction and molecular validity. Following RL fine-tuning, the proposed method demonstrates promising outcomes, yielding notable improvements in molecular validity, interaction efficacy, and critical chemical properties, achieving 65.37 for Quantitative Estimation of Drug-likeness (QED), 321.55 for Molecular Weight (MW), and 4.47 for Octanol-Water Partition Coefficient (logP), respectively. Furthermore, out of the generated drugs, only 0.041% do not exhibit novelty.
Authors: Simmaco Di Lillo, Domenico Marinucci, Michele Salvi, Stefano Vigogna
Abstract: It is well-known that randomly initialized, push-forward, fully-connected neural networks weakly converge to isotropic Gaussian processes, in the limit where the width of all layers goes to infinity. In this paper, we propose to use the angular power spectrum of the limiting field to characterize the complexity of the network architecture. In particular, we define sequences of random variables associated with the angular power spectrum, and provide a full characterization of the network complexity in terms of the asymptotic distribution of these sequences as the depth diverges. On this basis, we classify neural networks as low-disorder, sparse, or high-disorder; we show how this classification highlights a number of distinct features for standard activation functions, and in particular, sparsity properties of ReLU networks. Our theoretical results are also validated by numerical simulations.
Authors: Ryan Thompson, Edwin V. Bonilla, Robert Kohn
Abstract: Directed acyclic graph (DAG) learning is a central task in structure discovery and causal inference. Although the field has witnessed remarkable advances over the past few years, it remains statistically and computationally challenging to learn a single (point estimate) DAG from data, let alone provide uncertainty quantification. We address the difficult task of quantifying graph uncertainty by developing a Bayesian variational inference framework based on novel, provably valid distributions that have support directly on the space of sparse DAGs. These distributions, which we use to define our prior and variational posterior, are induced by a projection operation that maps an arbitrary continuous distribution onto the space of sparse weighted acyclic adjacency matrices. While this projection is combinatorial, it can be solved efficiently using recent continuous reformulations of acyclicity constraints. We empirically demonstrate that our method, ProDAG, can outperform state-of-the-art alternatives in both accuracy and uncertainty quantification.
Authors: Neha Kalibhat, Priyatham Kattakinda, Sumit Nawathe, Arman Zarei, Nikita Seleznev, Samuel Sharpe, Senthil Kumar, Soheil Feizi
Abstract: Vision transformers have established a precedent of patchifying images into uniformly-sized chunks before processing. We hypothesize that this design choice may limit models in learning comprehensive and compositional representations from visual data. This paper explores the notion of providing semantically-meaningful visual tokens to transformer encoders within a vision-language pre-training framework. Leveraging off-the-shelf segmentation and scene-graph models, we extract representations of instance segmentation masks (referred to as tangible tokens) and relationships and actions (referred to as intangible tokens). Subsequently, we pre-train a vision-side transformer by incorporating these newly extracted tokens and aligning the resultant embeddings with caption embeddings from a text-side encoder. To capture the structural and semantic relationships among visual tokens, we introduce additive attention weights, which are used to compute self-attention scores. Our experiments on COCO demonstrate notable improvements over ViTs in learned representation quality across text-to-image (+47%) and image-to-text retrieval (+44%) tasks. Furthermore, we showcase the advantages on compositionality benchmarks such as ARO (+18%) and Winoground (+10%).
Authors: Lars Doorenbos, Raphael Sznitman, Pablo M\'arquez-Neila
Abstract: Deep learning models are vulnerable to performance degradation when encountering out-of-distribution (OOD) images, potentially leading to misdiagnoses and compromised patient care. These shortcomings have led to great interest in the field of OOD detection. Existing unsupervised OOD (U-OOD) detection methods typically assume that OOD samples originate from an unconcentrated distribution complementary to the training distribution, neglecting the reality that deployed models passively accumulate task-specific OOD samples over time. To better reflect this real-world scenario, we introduce Iterative Deployment Exposure (IDE), a novel and more realistic setting for U-OOD detection. We propose CSO, a method for IDE that starts from a U-OOD detector that is agnostic to the OOD distribution and slowly refines it during deployment using observed unlabeled data. CSO uses a new U-OOD scoring function that combines the Mahalanobis distance with a nearest-neighbor approach, along with a novel confidence-scaled few-shot OOD detector to effectively learn from limited OOD examples. We validate our approach on a dedicated benchmark, showing that our method greatly improves upon strong baselines on three medical imaging modalities.
Authors: Lecheng Zheng, Zhengzhang Chen, Dongjie Wang, Chengyuan Deng, Reon Matsuoka, Haifeng Chen
Abstract: Root cause analysis (RCA) is crucial for enhancing the reliability and performance of complex systems. However, progress in this field has been hindered by the lack of large-scale, open-source datasets tailored for RCA. To bridge this gap, we introduce LEMMA-RCA, a large dataset designed for diverse RCA tasks across multiple domains and modalities. LEMMA-RCA features various real-world fault scenarios from IT and OT operation systems, encompassing microservices, water distribution, and water treatment systems, with hundreds of system entities involved. We evaluate the quality of LEMMA-RCA by testing the performance of eight baseline methods on this dataset under various settings, including offline and online modes as well as single and multiple modalities. Our experimental results demonstrate the high quality of LEMMA-RCA. The dataset is publicly available at https://lemma-rca.github.io/.
Authors: Patrick Chao, Yan Sun, Edgar Dobriban, Hamed Hassani
Abstract: Recent progress in large language models enables the creation of realistic machine-generated content. Watermarking is a promising approach to distinguish machine-generated text from human text, embedding statistical signals in the output that are ideally undetectable to humans. We propose a watermarking framework that encodes such signals through an error correcting code. Our method, termed robust binary code (RBC) watermark, introduces no noticeable degradation in quality. We evaluate our watermark on base and instruction fine-tuned models and find our watermark is robust to edits, deletions, and translations. We provide an information-theoretic perspective on watermarking, a powerful statistical test for detection and for generating $p$-values, and theoretical guarantees. Our empirical findings suggest our watermark is fast, powerful, and robust, comparing favorably to the state-of-the-art.
Authors: Gurusha Juneja, Gautam Jajoo, Nagarajan Natarajan, Hua Li, Jian Jiao, Amit Sharma
Abstract: Given a task in the form of a basic description and its training examples, prompt optimization is the problem of synthesizing the given information into a text prompt for a large language model. Humans solve this problem by also considering the different facets that define a task (e.g., counter-examples, explanations, analogies) and including them in the prompt. However, it is unclear whether existing algorithmic approaches, based on iteratively editing a given prompt or automatically selecting a few in-context examples, can cover the multiple facets required to solve a complex task. In this work, we view prompt optimization as that of learning multiple facets of a task from a set of training examples. We exploit structure in the prompt optimization problem and break down a prompt into loosely coupled semantic sections. The proposed algorithm, UniPrompt, (1) clusters the input space and uses clustered batches so that each batch likely corresponds to a different facet of the task, and (2) utilizes a feedback mechanism to propose adding, editing or deleting a section, which in turn is aggregated over a batch to capture generalizable facets. Empirical evaluation on multiple datasets and a real-world task shows that prompts generated using \shortname{} obtain higher accuracy than human-tuned prompts and those from state-of-the-art methods. In particular, our algorithm can generate long, complex prompts that existing methods are unable to generate. Code for UniPrompt is available at https://aka.ms/uniprompt.
Authors: Pablo Moriano, Steven C. Hespeler, Mingyan Li, Robert A. Bridges
Abstract: Vehicular controller area networks (CANs) are susceptible to masquerade attacks by malicious adversaries. In masquerade attacks, adversaries silence a targeted ID and then send malicious frames with forged content at the expected timing of benign frames. As masquerade attacks could seriously harm vehicle functionality and are the stealthiest attacks to detect in CAN, recent work has devoted attention to compare frameworks for detecting masquerade attacks in CAN. However, most existing works report offline evaluations using CAN logs already collected using simulations that do not comply with the domain's real-time constraints. Here we contribute to advance the state of the art by introducing a benchmark study of four different non-deep learning (DL)-based unsupervised online intrusion detection systems (IDS) for masquerade attacks in CAN. Our approach differs from existing benchmarks in that we analyze the effect of controlling streaming data conditions in a sliding window setting. In doing so, we use realistic masquerade attacks being replayed from the ROAD dataset. We show that although benchmarked IDS are not effective at detecting every attack type, the method that relies on detecting changes in the hierarchical structure of clusters of time series produces the best results at the expense of higher computational overhead. We discuss limitations, open challenges, and how the benchmarked methods can be used for practical unsupervised online CAN IDS for masquerade attacks.
Authors: Jiajun Hu, Zhen Lu, Yue Yang
Abstract: We propose a geometry-to-flow diffusion model that utilizes obstacle shape as input to predict a flow field around an obstacle. The model is based on a learnable Markov transition kernel to recover the data distribution from the Gaussian distribution. The Markov process is conditioned on the obstacle geometry, estimating the noise to be removed at each step, implemented via a U-Net. A cross-attention mechanism incorporates the geometry as a prompt. We train the geometry-to-flow diffusion model using a dataset of flows around simple obstacles, including circles, ellipses, rectangles, and triangles. For comparison, two CNN-based models and a VAE model are trained on the same dataset. Tests are carried out on flows around obstacles with simple and complex geometries, representing interpolation and generalization on the geometry condition, respectively. To evaluate performance under demanding conditions, the test set incorporates scenarios including crosses and the characters `PKU.' Generated flow fields show that the geometry-to-flow diffusion model is superior to the CNN-based models and the VAE model in predicting instantaneous flow fields and handling complex geometries. Quantitative analysis of the accuracy and divergence demonstrates the model's robustness.
Authors: Zhiqiang Cai, Anastassia Doktorova, Robert D. Falgout, C\'esar Herrera
Abstract: This paper studies the shallow Ritz method for solving one-dimensional diffusion-reaction problems. The method is capable of improving the order of approximation for non-smooth problems. By following a similar approach to the one presented in [9], we present a damped block Newton (dBN) method to achieve nearly optimal order of approximation. The dBN method optimizes the Ritz functional by alternating between the linear and non-linear parameters of the shallow ReLU neural network (NN). For diffusion-reaction problems, new difficulties arise: (1) for the linear parameters, the mass matrix is dense and even more ill-conditioned than the stiffness matrix, and (2) for the non-linear parameters, the Hessian matrix is dense and may be singular. This paper addresses these challenges, resulting in a dBN method with computational cost of ${\cal O}(n)$. The ideas presented for diffusion-reaction problems can also be applied to least-squares approximation problems. For both applications, starting with the non-linear parameters as a uniform partition, numerical experiments show that the dBN method moves the mesh points to nearly optimal locations.
Authors: Fadel M. Megahed, Ying-Ju Chen, Joshua A. Ferris, Cameron Resatar, Kaitlyn Ross, Younghwa Lee, L. Allison Jones-Farmer
Abstract: As generative AI ('GenAI') continues to evolve, educators face the challenge of preparing students for a future where AI-assisted work is integral to professional success. This paper introduces ChatISA, an in-house, multi-model AI chatbot designed to support students and faculty in an Information Systems and Analytics (ISA) department. ChatISA comprises four primary modules: Coding Companion, Project Coach, Exam Ally, and Interview Mentor, each tailored to enhance different aspects of the educational experience. Through iterative development, student feedback, and leveraging open-source frameworks, we created a robust tool that addresses coding inquiries, project management, exam preparation, and interview readiness. The implementation of ChatISA provided valuable insights and highlighted key challenges. Our findings demonstrate the benefits of ChatISA for ISA education while underscoring the need for adaptive pedagogy and proactive engagement with AI tools to fully harness their educational potential. To support broader adoption and innovation, all code for ChatISA is made publicly available on GitHub, enabling other institutions to customize and integrate similar AI-driven educational tools within their curricula.
Authors: Yinghao Zhu, Junyi Gao, Zixiang Wang, Weibin Liao, Xiaochen Zheng, Lifang Liang, Miguel O. Bernabeu, Yasha Wang, Lequan Yu, Chengwei Pan, Ewen M. Harrison, Liantao Ma
Abstract: Large Language Models (LLMs) are increasingly deployed in medicine. However, their utility in non-generative clinical prediction, often presumed inferior to specialized models, remains under-evaluated, leading to ongoing debate within the field and potential for misuse, misunderstanding, or over-reliance due to a lack of systematic benchmarking. Our ClinicRealm study addresses this by benchmarking 9 GPT-based LLMs, 5 BERT-based models, and 7 traditional methods on unstructured clinical notes and structured Electronic Health Records (EHR). Key findings reveal a significant shift: for clinical note predictions, leading LLMs (e.g., DeepSeek R1/V3, GPT o3-mini-high) in zero-shot settings now decisively outperform finetuned BERT models. On structured EHRs, while specialized models excel with ample data, advanced LLMs (e.g., GPT-4o, DeepSeek R1/V3) show potent zero-shot capabilities, often surpassing conventional models in data-scarce settings. Notably, leading open-source LLMs can match or exceed proprietary counterparts. These results establish modern LLMs as powerful non-generative clinical prediction tools, particularly with unstructured text and offering data-efficient structured data options, thus necessitating a re-evaluation of model selection strategies. This research should serve as an important insight for medical informaticists, AI developers, and clinical researchers, potentially prompting a reassessment of current assumptions and inspiring new approaches to LLM application in predictive healthcare.
Authors: Xianzhen Luo, Yixuan Wang, Qingfu Zhu, Zhiming Zhang, Xuanyu Zhang, Qing Yang, Dongliang Xu
Abstract: Massive parameters of LLMs have made inference latency a fundamental bottleneck. Speculative decoding represents a lossless approach to accelerate inference through a guess-and-verify paradigm. Some methods rely on additional architectures to guess draft tokens, which need extra training before use. Alternatively, retrieval-based training-free techniques build libraries from pre-existing corpora or by n-gram generation. However, they face challenges like large storage requirements, time-consuming retrieval, and limited adaptability. Observing that candidate tokens generated during the decoding process are likely to reoccur in future sequences, we propose Token Recycling. It stores candidate tokens in an adjacency matrix and employs a breadth-first-search (BFS)-like algorithm to construct a draft tree, which is then validated through tree attention. New candidate tokens from the decoding process are then used to update the matrix. Token Recycling requires \textless2MB of additional storage and achieves approximately 2x speedup across all sizes of LLMs. It significantly outperforms existing train-free methods by 30\% and even a widely recognized training method by 25\%.
Authors: Aishik Nagar, Viktor Schlegel, Thanh-Tung Nguyen, Hao Li, Yuping Wu, Kuluhan Binici, Stefan Winkler
Abstract: Large Language Models (LLMs) are increasingly adopted for applications in healthcare, reaching the performance of domain experts on tasks such as question answering and document summarisation. Despite their success on these tasks, it is unclear how well LLMs perform on tasks that are traditionally pursued in the biomedical domain, such as structured information extraction. To bridge this gap, in this paper, we systematically benchmark LLM performance in Medical Classification and Named Entity Recognition (NER) tasks. We aim to disentangle the contribution of different factors to the performance, particularly the impact of LLMs' task knowledge and reasoning capabilities, their (parametric) domain knowledge, and addition of external knowledge. To this end, we evaluate various open LLMs - including BioMistral and Llama-2 models - on a diverse set of biomedical datasets, using standard prompting, Chain of-Thought (CoT) and Self Consistency based reasoning as well as Retrieval-Augmented Generation (RAG) with PubMed and Wikipedia corpora. Counter intuitively, our results reveal that standard prompting consistently outperforms more complex techniques across both tasks, laying bare the limitations in the current application of CoT, self-consistency and RAG in the biomedical domain. Our findings suggest that advanced prompting methods developed for knowledge- or reasoning-intensive tasks, such as CoT or RAG, are not easily portable to biomedical tasks where precise structured outputs are required. This highlights the need for more effective integration of external knowledge and reasoning mechanisms in LLMs to enhance their performance in real-world biomedical applications.
Authors: Buxin Su, Jiayao Zhang, Natalie Collina, Yuling Yan, Didong Li, Kyunghyun Cho, Jianqing Fan, Aaron Roth, Weijie Su
Abstract: We conducted an experiment during the review process of the 2023 International Conference on Machine Learning (ICML), asking authors with multiple submissions to rank their papers based on perceived quality. In total, we received 1,342 rankings, each from a different author, covering 2,592 submissions. In this paper, we present an empirical analysis of how author-provided rankings could be leveraged to improve peer review processes at machine learning conferences. We focus on the Isotonic Mechanism, which calibrates raw review scores using the author-provided rankings. Our analysis shows that these ranking-calibrated scores outperform the raw review scores in estimating the ground truth ``expected review scores'' in terms of both squared and absolute error metrics. Furthermore, we propose several cautious, low-risk applications of the Isotonic Mechanism and author-provided rankings in peer review, including supporting senior area chairs in overseeing area chairs' recommendations, assisting in the selection of paper awards, and guiding the recruitment of emergency reviewers.
Authors: Rahul Moorthy, Volkan Isler
Abstract: The ability to capture rich representations of combinatorial structures has enabled the application of machine learning to tasks such as analysis and generation of floorplans, terrains, images, and animations. Recent work has primarily focused on understanding structures with well-defined features, neighborhoods, or underlying distance metrics, while those lacking such characteristics remain largely unstudied. Examples of these combinatorial structures can be found in polygons, where a small change in the vertex locations causes a significant rearrangement of the combinatorial structure, expressed as a visibility or triangulation graphs. Current representation learning approaches fail to capture structures without well-defined features and distance metrics. In this paper, we study the open problem of Visibility Reconstruction: Given a visibility graph $G$, construct a polygon $P$ whose visibility graph is $G$. We introduce VisDiff, a novel diffusion-based approach to generate polygon $P$ from the input visibility graph $G$. The main novelty of our approach is that, rather than generating the polygon's vertex set directly, we first estimate the signed distance function (SDF) associated with the polygon. The SDF is then used to extract the vertex location representing the final polygon. We show that going through the SDF allows VisDiff to learn the visibility relationship much more effectively than generating vertex locations directly. In order to train VisDiff, we create a carefully curated dataset. We use this dataset to benchmark our method and achieve 26% improvement in F1-Score over standard methods as well as state of the art approaches.
Authors: Zonglin Yang, Wanhao Liu, Ben Gao, Tong Xie, Yuqiang Li, Wanli Ouyang, Soujanya Poria, Erik Cambria, Dongzhan Zhou
Abstract: Scientific discovery plays a pivotal role in advancing human society, and recent progress in large language models (LLMs) suggests their potential to accelerate this process. However, it remains unclear whether LLMs can autonomously generate novel and valid hypotheses in chemistry. In this work, we investigate whether LLMs can discover high-quality chemistry hypotheses given only a research background-comprising a question and/or a survey-without restriction on the domain of the question. We begin with the observation that hypothesis discovery is a seemingly intractable task. To address this, we propose a formal mathematical decomposition grounded in a fundamental assumption: that most chemistry hypotheses can be composed from a research background and a set of inspirations. This decomposition leads to three practical subtasks-retrieving inspirations, composing hypotheses with inspirations, and ranking hypotheses - which together constitute a sufficient set of subtasks for the overall scientific discovery task. We further develop an agentic LLM framework, MOOSE-Chem, that is a direct implementation of this mathematical decomposition. To evaluate this framework, we construct a benchmark of 51 high-impact chemistry papers published and online after January 2024, each manually annotated by PhD chemists with background, inspirations, and hypothesis. The framework is able to rediscover many hypotheses with high similarity to the groundtruth, successfully capturing the core innovations-while ensuring no data contamination since it uses an LLM with knowledge cutoff date prior to 2024. Finally, based on LLM's surprisingly high accuracy on inspiration retrieval, a task with inherently out-of-distribution nature, we propose a bold assumption: that LLMs may already encode latent scientific knowledge associations not yet recognized by humans.
Authors: Desai Xie, Zhan Xu, Yicong Hong, Hao Tan, Difan Liu, Feng Liu, Arie Kaufman, Yang Zhou
Abstract: Current frontier video diffusion models have demonstrated remarkable results at generating high-quality videos. However, they can only generate short video clips, normally around 10 seconds or 240 frames, due to computation limitations during training. Existing methods naively achieve autoregressive long video generation by directly placing the ending of the previous clip at the front of the attention window as conditioning, which leads to abrupt scene changes, unnatural motion, and error accumulation. In this work, we introduce a more natural formulation of autoregressive long video generation by revisiting the noise level assumption in video diffusion models. Our key idea is to 1. assign the frames with per-frame, progressively increasing noise levels rather than a single noise level and 2. denoise and shift the frames in small intervals rather than all at once. This allows for smoother attention correspondence among frames with adjacent noise levels, larger overlaps between the attention windows, and better propagation of information from the earlier to the later frames. Video diffusion models equipped with our progressive noise schedule can autoregressively generate long videos with much improved fidelity compared to the baselines and minimal quality degradation over time. We present the first results on text-conditioned 60-second (1440 frames) long video generation at a quality close to frontier models. Code and video results are available at https://desaixie.github.io/pa-vdm/.
Authors: Zhiwei Li, Guodong Long, Jing Jiang, Chengqi Zhang
Abstract: Delivering deeply personalized recommendations necessitates understanding user interactions with diverse multimedia features, but achieving this within the constraints of Federated Recommendation Systems (FedRec) is severely hampered by communication bottlenecks, user heterogeneity, and the complexity of privacy-preserving multimodal fusion. To this end, we propose FedMR, a novel multimodal FedRec framework centered around the Mixing Feature Fusion Module (MFFM). FedMR employs a two-stage process: (1) Server-side centralized multimedia content processing provides rich, shared item context using pre-trained models, mitigating limitations from client sparsity and resource constraints efficiently. (2) Client-Side Personalized Refinement, where the MFFM dynamically adapts these server-provided multimodal representations based on client-specific interaction patterns, effectively tailoring recommendations and resolving heterogeneity in user preferences towards different modalities. Extensive experiments validate that FedMR seamlessly enhances existing ID-based FedRecs, effectively transforming them into high-performing federated multimodal systems.
Authors: Mengdi Zhang, Kai Kiat Goh, Peixin Zhang, Jun Sun, Rose Lin Xin, Hongyu Zhang
Abstract: Despite the success of Large Language Models (LLMs) across various fields, their potential to generate untruthful, biased and harmful responses poses significant risks, particularly in critical applications. This highlights the urgent need for systematic methods to detect and prevent such misbehavior. While existing approaches target specific issues such as harmful responses, this work introduces LLMScan, an innovative LLM monitoring technique based on causality analysis, offering a comprehensive solution. LLMScan systematically monitors the inner workings of an LLM through the lens of causal inference, operating on the premise that the LLM's `brain' behaves differently when misbehaving. By analyzing the causal contributions of the LLM's input tokens and transformer layers, LLMScan effectively detects misbehavior. Extensive experiments across various tasks and models reveal clear distinctions in the causal distributions between normal behavior and misbehavior, enabling the development of accurate, lightweight detectors for a variety of misbehavior detection tasks.
Authors: Mikko Partio, Leila Hieta, Anniina Kokkonen
Abstract: Cloud cover plays a critical role in weather prediction and impacts several sectors, including agriculture, solar power generation, and aviation. Despite advancements in numerical weather prediction (NWP) models, forecasting total cloud cover remains challenging due to the small-scale nature of cloud formation processes. In this study, we introduce CloudCast, a convolutional neural network (CNN) based on the U-Net architecture, designed to predict total cloud cover (TCC) up to five hours ahead. Trained on five years of satellite data, CloudCast significantly outperforms traditional NWP models and optical flow methods. Compared to a reference NWP model, CloudCast achieves a 24% lower mean absolute error and reduces multi-category prediction errors by 46%. The model demonstrates strong performance, particularly in capturing the large-scale structure of cloud cover in the first few forecast hours, though later predictions are subject to blurring and underestimation of cloud formation. An ablation study identified the optimal input features and loss functions, with MAE-based models performing the best. CloudCast has been integrated into the Finnish Meteorological Institute's operational nowcasting system, where it improves cloud cover forecasts used by public and private sector clients. While CloudCast is limited by a relatively short skillful lead time of about three hours, future work aims to extend this through more complex network architectures and higher-resolution data. CloudCast code is available at https://github.com/fmidev/cloudcast.
Authors: He Jiang, Yutong Wang, Rishi Veerapaneni, Tanishq Duhan, Guillaume Sartoretti, Jiaoyang Li
Abstract: Lifelong Multi-Agent Path Finding (LMAPF) repeatedly finds collision-free paths for multiple agents that are continually assigned new goals when they reach current ones. Recently, this field has embraced learning-based methods, which reactively generate single-step actions based on individual local observations. However, it is still challenging for them to match the performance of the best search-based algorithms, especially in large-scale settings. This work proposes an imitation-learning-based LMAPF solver that introduces a novel communication module as well as systematic single-step collision resolution and global guidance techniques. Our proposed solver, Scalable Imitation Learning for LMAPF (SILLM), inherits the fast reasoning speed of learning-based methods and the high solution quality of search-based methods with the help of modern GPUs. Across six large-scale maps with up to 10,000 agents and varying obstacle structures, SILLM surpasses the best learning- and search-based baselines, achieving average throughput improvements of 137.7% and 16.0%, respectively. Furthermore, SILLM also beats the winning solution of the 2023 League of Robot Runners, an international LMAPF competition. Finally, we validated SILLM with 10 real robots and 100 virtual robots in a mock warehouse environment.
Authors: Koen Ponse, Aske Plaat, Niki van Stein, Thomas M. Moerland
Abstract: Accurate economic simulations often require many experimental runs, particularly when combined with reinforcement learning. Unfortunately, training reinforcement learning agents in multi-agent economic environments can be slow. This paper introduces EconoJax, a fast simulated economy, based on the AI economist. EconoJax, and its training pipeline, are completely written in JAX. This allows EconoJax to scale to large population sizes and perform large experiments, while keeping training times within minutes. Through experiments with populations of 100 agents, we show how real-world economic behavior emerges through training within 15 minutes, in contrast to previous work that required several days. We additionally perform experiments in varying sized action spaces to test if some multi-agent methods produce more diverse behavior compared to others. Here, our findings indicate no notable differences in produced behavior with different methods as is sometimes suggested in earlier works. To aid further research, we open-source EconoJax on Github.
Authors: William F. Bradley
Abstract: As large language models (LLMs) become increasingly powerful, traditional evaluation metrics tend to saturate, making it challenging to distinguish between models. We propose a general method to transform existing LLM evaluations into a series of progressively more difficult tasks. These enhanced evaluations emphasize reasoning capabilities and can reveal relative performance differences that are not apparent in the original assessments. To demonstrate the effectiveness of our approach, we create a new multiple-choice test corpus, extend it into a family of evaluations, and assess a collection of LLMs. Our results offer insights into the comparative abilities of these models, particularly highlighting the differences between base LLMs and more recent "reasoning" models.
Authors: Wentang Song, Zhiyuan Yan, Yuzhen Lin, Taiping Yao, Changsheng Chen, Shen Chen, Yandan Zhao, Shouhong Ding, Bin Li
Abstract: Detecting AI-generated images, particularly deepfakes, has become increasingly crucial, with the primary challenge being the generalization to previously unseen manipulation methods. This paper tackles this issue by leveraging the forgery quality of training data to improve the generalization performance of existing deepfake detectors. Generally, the forgery quality of different deepfakes varies: some have easily recognizable forgery clues, while others are highly realistic. Existing works often train detectors on a mix of deepfakes with varying forgery qualities, potentially leading detectors to short-cut the easy-to-spot artifacts from low-quality forgery samples, thereby hurting generalization performance. To tackle this issue, we propose a novel quality-centric framework for generic deepfake detection, which is composed of a Quality Evaluator, a low-quality data enhancement module, and a learning pacing strategy that explicitly incorporates forgery quality into the training process. Our framework is inspired by curriculum learning, which is designed to gradually enable the detector to learn more challenging deepfake samples, starting with easier samples and progressing to more realistic ones. We employ both static and dynamic assessments to assess the forgery quality, combining their scores to produce a final rating for each training sample. The rating score guides the selection of deepfake samples for training, with higher-rated samples having a higher probability of being chosen. Furthermore, we propose a novel frequency data augmentation method specifically designed for low-quality forgery samples, which helps to reduce obvious forgery traces and improve their overall realism. Extensive experiments demonstrate that our proposed framework can be applied plug-and-play to existing detection models and significantly enhance their generalization performance in detection.
Authors: Andrew Shannon, Conor Houghton, David Barton, Martin Homer
Abstract: Surrogate modeling of non-linear oscillator networks remains challenging due to discrepancies between simplified analytical models and real-world complexity. To bridge this gap, we investigate hybrid reservoir computing, combining reservoir computing with "expert" analytical models. Simulating the absence of an exact model, we first test the surrogate models with parameter errors in their expert model. Second, in a residual physics task, we assess their performance when their expert model lacks key non-linear coupling terms present in an extended ground-truth model. We focus on short-term forecasting across diverse dynamical regimes, evaluating the use of these surrogates for control applications. We show that hybrid reservoir computers generally outperform standard reservoir computers and exhibit greater robustness to parameter tuning. This advantage is less pronounced in the residual physics task. Notably, unlike standard reservoir computers, the performance of the hybrid does not degrade when crossing an observed spectral radius threshold. Furthermore, there is good performance for dynamical regimes not accessible to the expert model, demonstrating the contribution of the reservoir.
Authors: Yanxi Chen, Xuchen Pan, Yaliang Li, Bolin Ding, Jingren Zhou
Abstract: We propose two simple, principled and practical algorithms that enjoy provable scaling laws for the test-time compute of large language models (LLMs). The first one is a two-stage knockout-style algorithm: given an input problem, it first generates multiple candidate solutions, and then aggregate them via a knockout tournament for the final output. Assuming that the LLM can generate a correct solution with non-zero probability and do better than a random guess in comparing a pair of correct and incorrect solutions, we prove theoretically that the failure probability of this algorithm decays to zero exponentially or by a power law (depending on the specific way of scaling) as its test-time compute grows. The second one is a two-stage league-style algorithm, where each candidate is evaluated by its average win rate against multiple opponents, rather than eliminated upon loss to a single opponent. Under analogous but more robust assumptions, we prove that its failure probability also decays to zero exponentially with more test-time compute. Both algorithms require a black-box LLM and nothing else (e.g., no verifier or reward model) for a minimalistic implementation, which makes them appealing for practical applications and easy to adapt for different tasks. Through extensive experiments with diverse models and datasets, we validate the proposed theories and demonstrate the outstanding scaling properties of both algorithms.
Authors: Wenbo Zhang, Lu Zhang, Ping Hu, Liqian Ma, Yunzhi Zhuge, Huchuan Lu
Abstract: Injecting semantics into 3D Gaussian Splatting (3DGS) has recently garnered significant attention. While current approaches typically distill 3D semantic features from 2D foundational models (e.g., CLIP and SAM) to facilitate novel view segmentation and semantic understanding, their heavy reliance on 2D supervision can undermine cross-view semantic consistency and necessitate complex data preparation processes, therefore hindering view-consistent scene understanding. In this work, we present FreeGS, an unsupervised semantic-embedded 3DGS framework that achieves view-consistent 3D scene understanding without the need for 2D labels. Instead of directly learning semantic features, we introduce the IDentity-coupled Semantic Field (IDSF) into 3DGS, which captures both semantic representations and view-consistent instance indices for each Gaussian. We optimize IDSF with a two-step alternating strategy: semantics help to extract coherent instances in 3D space, while the resulting instances regularize the injection of stable semantics from 2D space. Additionally, we adopt a 2D-3D joint contrastive loss to enhance the complementarity between view-consistent 3D geometry and rich semantics during the bootstrapping process, enabling FreeGS to uniformly perform tasks such as novel-view semantic segmentation, object selection, and 3D object detection. Extensive experiments on LERF-Mask, 3D-OVS, and ScanNet datasets demonstrate that FreeGS performs comparably to state-of-the-art methods while avoiding the complex data preprocessing workload. Our code is publicly available at https://github.com/wb014/FreeGS.
Authors: Prajwal Singh, Ashish Tiwari, Gautam Vashishtha, Shanmuganathan Raman
Abstract: Neural radiance fields (NeRF) have revolutionized photorealistic rendering of novel views for 3D scenes. Despite their growing popularity and efficiency as 3D resources, NeRFs face scalability challenges due to the need for separate models per scene and the cumulative increase in training time for multiple scenes. The potential for incrementally encoding multiple 3D scenes into a single NeRF model remains largely unexplored. To address this, we introduce Continual-Neural Graphics Primitives (C-NGP), a novel continual learning framework that integrates multiple scenes incrementally into a single neural radiance field. Using a generative replay approach, C-NGP adapts to new scenes without requiring access to old data. We demonstrate that C-NGP can accommodate multiple scenes without increasing the parameter count, producing high-quality novel-view renderings on synthetic and real datasets. Notably, C-NGP models all $8$ scenes from the Real-LLFF dataset together, with only a $2.2\%$ drop in PSNR compared to vanilla NeRF, which models each scene independently. Further, C-NGP allows multiple style edits in the same network.
Authors: Takayuki Osogami, Hirota Kinoshita, Segev Wasserkrug
Abstract: We analytically derive a class of optimal solutions to a linear program (LP) for automated mechanism design that satisfies efficiency, incentive compatibility, strong budget balance (SBB), and individual rationality (IR), where SBB and IR are enforced in expectation. These solutions can be expressed using a set of essential variables whose cardinality is exponentially smaller than the total number of variables in the original formulation. However, evaluating a key term in the solutions requires exponentially many optimization steps as the number of players $N$ increases. We address this by translating the evaluation of this term into a multi-armed bandit (MAB) problem and develop a probably approximately correct (PAC) estimator with asymptotically optimal sample complexity. This MAB-based approach reduces the optimization complexity from exponential to $O(N\log N)$. Numerical experiments confirm that our method efficiently computes mechanisms with the target properties, scaling to problems with up to $N=128$ players -- substantially improving over prior work.
Authors: Kevin Murphy
Abstract: This manuscript gives a big-picture, up-to-date overview of the field of (deep) reinforcement learning and sequential decision making, covering value-based methods, policy-based methods, model-based methods, multi-agent RL, LLMs and RL, and various other topics (e.g., offline RL, hierarchical RL, intrinsic reward).
Authors: Haizhou Shi, Yibin Wang, Ligong Han, Huan Zhang, Hao Wang
Abstract: Estimating the uncertainty of responses from Large Language Models (LLMs) remains a critical challenge. While recent Bayesian methods have demonstrated effectiveness in quantifying uncertainty through low-rank weight updates, they typically require complex fine-tuning or post-training procedures. In this paper, we propose Training-Free Bayesianization (TFB), a simple yet theoretically grounded framework that efficiently transforms trained low-rank adapters into Bayesian ones without additional training. TFB systematically searches for the maximally acceptable level of variance in the weight posterior, constrained within a family of low-rank isotropic Gaussian distributions. Our theoretical analysis shows that under mild conditions, this search process is equivalent to KL-regularized variational optimization, a generalized form of variational inference. Through comprehensive experiments, we show that TFB achieves superior uncertainty estimation and generalization compared to existing methods while eliminating the need for complex Bayesianization training procedures. Code will be available at https://github.com/Wang-ML-Lab/bayesian-peft.
Authors: Kangyu Zhu, Peng Xia, Yun Li, Hongtu Zhu, Sheng Wang, Huaxiu Yao
Abstract: The advancement of Large Vision-Language Models (LVLMs) has propelled their application in the medical field. However, Medical LVLMs (Med-LVLMs) encounter factuality challenges due to modality misalignment, where the models prioritize textual knowledge over visual input, leading to hallucinations that contradict information in medical images. Previous attempts to enhance modality alignment in Med-LVLMs through preference optimization have inadequately mitigated clinical relevance in preference data, making these samples easily distinguishable and reducing alignment effectiveness. To address this challenge, we propose MMedPO, a novel multimodal medical preference optimization approach that considers the clinical relevance of preference samples to enhance Med-LVLM alignment. MMedPO curates multimodal preference data by introducing two types of dispreference: (1) plausible hallucinations injected through target Med-LVLMs or GPT-4o to produce medically inaccurate responses, and (2) lesion region neglect achieved through local lesion-noising, disrupting visual understanding of critical areas. We then calculate clinical relevance for each sample based on scores from multiple Med-LLMs and visual tools, and integrate these scores into the preference optimization process as weights, enabling effective alignment. Our experiments demonstrate that MMedPO significantly enhances factual accuracy in Med-LVLMs, achieving substantial improvements over existing preference optimization methods by averaging 14.2% and 51.7% across the Med-VQA and report generation tasks. Our code are available in https://github.com/aiming-lab/MMedPO.
Authors: Minjae Kwon, Ingy ElSayed-Aly, Lu Feng
Abstract: There is a surge of interest in using formal languages such as Linear Temporal Logic (LTL) to precisely and succinctly specify complex tasks and derive reward functions for Reinforcement Learning (RL). However, existing methods often assign sparse rewards (e.g., giving a reward of 1 only if a task is completed and 0 otherwise). By providing feedback solely upon task completion, these methods fail to encourage successful subtask completion. This is particularly problematic in environments with inherent uncertainty, where task completion may be unreliable despite progress on intermediate goals. To address this limitation, we propose a suite of reward functions that incentivize an RL agent to complete a task specified by an LTL formula as much as possible, and develop an adaptive reward shaping approach that dynamically updates reward functions during the learning process. Experimental results on a range of benchmark RL environments demonstrate that the proposed approach generally outperforms baselines, achieving earlier convergence to a better policy with higher expected return and task completion rate.
Authors: Giorgio Giannone, Ruoteng Li, Qianli Feng, Evgeny Perevodchikov, Rui Chen, Aleix Martinez
Abstract: Vision-language models (VLMs) have demonstrated remarkable potential in integrating visual and linguistic information, but their performance is often constrained by the need for extensive, high-quality image-text training data. Curation of these image-text pairs is both time-consuming and computationally expensive. To address this challenge, we introduce SVP (Sampling-based Visual Projection), a novel framework that enhances vision-language alignment without relying on manually curated text-image pairs or preference annotation. SVP leverages a small set of manually selected images, self-captioning and a pre-trained grounding model as a feedback mechanism to elicit latent information in VLMs. We evaluate our approach across six key areas: captioning, referring, visual question answering, multitasking, hallucination control, and object recall. Results demonstrate significant improvements, including a 14 % average improvement in captioning tasks, up to 12 % increase in object recall, and significantly reduced hallucinations, while maintaining question-answering capabilities. Using SVP, a small VLM achieves hallucination reductions similar to a model five times larger, while a VLM with initially poor referring capabilities more than doubles its performance, approaching parity with a model twice its size.
Authors: Maria Nareklishvili, Nick Polson, Vadim Sokolov
Abstract: Generative methods (Gen-AI) are reviewed with a particular goal of solving tasks in machine learning and Bayesian inference. Generative models require one to simulate a large training dataset and to use deep neural networks to solve a supervised learning problem. To do this, we require high-dimensional regression methods and tools for dimensionality reduction (a.k.a. feature selection). The main advantage of Gen-AI methods is their ability to be model-free and to use deep neural networks to estimate conditional densities or posterior quintiles of interest. To illustrate generative methods , we analyze the well-known Ebola data set. Finally, we conclude with directions for future research.
Authors: Sydney Vernon, Eviatar Bach, Oliver R. A. Dunbar
Abstract: Ensemble Kalman inversion (EKI) is a derivative-free, particle-based optimization method for solving inverse problems. It can be shown that EKI approximates a gradient flow, which allows the application of methods for accelerating gradient descent. Here, we show that Nesterov acceleration is effective in speeding up the reduction of the EKI cost function on a variety of inverse problems. We also implement Nesterov acceleration for two EKI variants, unscented Kalman inversion and ensemble transform Kalman inversion. Our specific implementation takes the form of a particle-level nudge that is demonstrably simple to couple in a black-box fashion with any existing EKI variant algorithms, comes with no additional computational expense, and with no additional tuning hyperparameters. This work shows a pathway for future research to translate advances in gradient-based optimization into advances in gradient-free Kalman optimization.
Authors: Vincent Koc
Abstract: The global crisis of language endangerment meets a technological turning point as Generative AI (GenAI) and Large Language Models (LLMs) unlock new frontiers in automating corpus creation, transcription, translation, and tutoring. However, this promise is imperiled by fragmented practices and the critical lack of a methodology to navigate the fraught balance between LLM capabilities and the profound risks of data scarcity, cultural misappropriation, and ethical missteps. This paper introduces a novel analytical framework that systematically evaluates GenAI applications against language-specific needs, embedding community governance and ethical safeguards as foundational pillars. We demonstrate its efficacy through the Te Reo M\=aori revitalization, where it illuminates successes, such as community-led Automatic Speech Recognition achieving 92% accuracy, while critically surfacing persistent challenges in data sovereignty and model bias for digital archives and educational tools. Our findings underscore that GenAI can indeed revolutionize language preservation, but only when interventions are rigorously anchored in community-centric data stewardship, continuous evaluation, and transparent risk management. Ultimately, this framework provides an indispensable toolkit for researchers, language communities, and policymakers, aiming to catalyze the ethical and high-impact deployment of LLMs to safeguard the world's linguistic heritage.
Authors: Zikun Li, Zhuofu Chen, Remi Delacourt, Gabriele Oliaro, Zeyu Wang, Qinghan Chen, Shuhuai Lin, April Yang, Zhihao Zhang, Zhuoming Chen, Sean Lai, Xinhao Cheng, Xupeng Miao, Zhihao Jia
Abstract: Modern large language model (LLM) applications exhibit diverse service-level objectives (SLOs), from low-latency requirements in interactive coding assistants to more relaxed constraints in data wrangling tasks. Existing LLM serving systems, which rely on uniform batching and scheduling strategies, often fail to meet these heterogeneous SLOs concurrently. We present AdaServe, the first LLM serving system designed to support efficient multi-SLO serving through SLO-customized speculative decoding. AdaServe formulates multi-SLO serving as a constrained optimization problem and introduces a hardware-aware algorithm that constructs a speculation tree tailored to each request's latency target. It features a speculate-select-verify pipeline that enables fine-grained control over decoding speed while maximizing system throughput. AdaServe further adapts to workload variation by dynamically adjusting speculation parameters. Evaluations across diverse workloads show that AdaServe reduces SLO violations by up to 4.3$\times$ and improves goodput by up to 1.9$\times$ compared to the best performing baselines, highlighting its effectiveness in multi-SLO serving.
Authors: Yingchao Yu, Yaochu Jin, Kuangrong Hao, Yuchen Xiao, Yuping Yan, Hengjie Yu, Zeqi Zheng
Abstract: The emergence of abstract rules from exemplars is central to the brain's capability of flexible generalization and rapid adaptation. However, the internal organizing mechanisms underlying rule abstraction remain elusive, largely due to the limitations of conventional models that lack intrinsic neuronal heterogeneity, making it hard to examine neuronal and structural differentiations. Inspired by astrocyte-mediated neuromodulation, this work introduces a hierarchically modulated recurrent spiking neural network (HM-RSNN) that can tune intrinsic neuronal properties, where a global stage simulates calcium wave-driven task-specific configuration and a local one mimics gliotransmitter-mediated fine-tuning. We conduct modeling using HM-RSNN across four cognitive tasks and rule abstraction contingent differentiation is observed at both network and neuron levels, leading to better performance compared to artificial neural networks. These findings highlight the critical role of dynamic internal organization in supporting the accomplishment of various cognitive tasks.
Authors: Zhenhao Zhu, Bulou Liu, Qingyao Ai, Yiqun Liu
Abstract: Multiple choice questions (MCQs) are a popular and important task for evaluating large language models (LLMs). Based on common strategies people use when answering MCQs, the process of elimination (PoE) has been proposed as an effective problem-solving method. Existing PoE methods typically either have LLMs directly identify incorrect options or score options and replace lower-scoring ones with [MASK]. However, both methods suffer from inapplicability or suboptimal performance. To address these issues, this paper proposes a novel option-ID based PoE ($\text{PoE}_{\text{ID}}$). $\text{PoE}_{\text{ID}}$ critically incorporates a debiasing technique to counteract LLMs token bias, enhancing robustness over naive ID-based elimination. It features two strategies: $\text{PoE}_{\text{ID}}^{\text{log}}$, which eliminates options whose IDs have log probabilities below the average threshold, and $\text{PoE}_{\text{ID}}^{\text{seq}}$, which iteratively removes the option with the lowest ID probability. We conduct extensive experiments with 6 different LLMs on 4 diverse datasets. The results demonstrate that $\text{PoE}_{\text{ID}}$, especially $\text{PoE}_{\text{ID}}^{\text{log}}$, significantly improves zero-shot and few-shot MCQs performance, particularly in datasets with more options. Our analyses demonstrate that $\text{PoE}_{\text{ID}}^{\text{log}}$ enhances the LLMs' confidence in selecting the correct option, and the option elimination strategy outperforms methods relying on [MASK] replacement. We further investigate the limitations of LLMs in directly identifying incorrect options, which stem from their inherent deficiencies.
Authors: Mohamad Qadri, Gokul Swamy, Jonathan Francis, Michael Kaess, Andrea Bajcsy
Abstract: Inverse Constraint Learning (ICL) is the problem of inferring constraints from safe (i.e., constraint-satisfying) demonstrations. The hope is that these inferred constraints can then be used downstream to search for safe policies for new tasks and, potentially, under different dynamics. Our paper explores the question of what mathematical entity ICL recovers. Somewhat surprisingly, we show that both in theory and in practice, ICL recovers the set of states where failure is inevitable, rather than the set of states where failure has already happened. In the language of safe control, this means we recover a backwards reachable tube (BRT) rather than a failure set. In contrast to the failure set, the BRT depends on the dynamics of the data collection system. We discuss the implications of the dynamics-conditionedness of the recovered constraint on both the sample-efficiency of policy search and the transferability of learned constraints.
Authors: Kaito Takanami, Takashi Takahashi, Ayaka Sakata
Abstract: Self-distillation (SD), a technique where a model improves itself using its own predictions, has attracted attention as a simple yet powerful approach in machine learning. Despite its widespread use, the mechanisms underlying its effectiveness remain unclear. In this study, we investigate the efficacy of hyperparameter-tuned multi-stage SD with a linear classifier for binary classification on noisy Gaussian mixture data. For the analysis, we employ the replica method from statistical physics. Our findings reveal that the primary driver of SD's performance improvement is denoising through hard pseudo-labels, with the most notable gains observed in moderately sized datasets. We also identify two practical heuristics to enhance SD: early stopping that limits the number of stages, which is broadly effective, and bias parameter fixing, which helps under label imbalance. To empirically validate our theoretical findings derived from our toy model, we conduct additional experiments on CIFAR-10 classification using pretrained ResNet backbone. These results provide both theoretical and practical insights, advancing our understanding and application of SD in noisy settings.
Authors: Siyuan Jiang, Yihan Hu, Wenjie Li, Pengcheng Zeng
Abstract: Functional data - observations in the form of curves or trajectories - arise in diverse domains such as biomedical sensing, motion capture, and handwriting recognition. A core challenge in functional data analysis (FDA) is accounting for phase variability, where misaligned temporal patterns hinder accurate inference. We introduce DeepFRC, an end-to-end deep learning framework for joint functional registration and classification. Unlike conventional approaches that decouple alignment and prediction, DeepFRC integrates class-aware elastic warping and a learnable basis representation into a unified architecture. This design enables temporal alignment and dimensionality reduction to be jointly optimized with classification, improving both interpretability and accuracy. We establish the first theoretical connection between alignment quality and generalization error, and validate our model on synthetic and real-world benchmarks. DeepFRC consistently outperforms state-of-the-art methods, especially in scenarios with complex temporal misalignment. Code is available at: https://github.com/Drivergo-93589/DeepFRC.
Authors: Haoyu Liang, Youran Sun, Yunfeng Cai, Jun Zhu, Bo Zhang
Abstract: The security issue of large language models (LLMs) has gained wide attention recently, with various defense mechanisms developed to prevent harmful output, among which safeguards based on text embedding models serve as a fundamental defense. Through testing, we discover that the output distribution of text embedding models is severely biased with a large mean. Inspired by this observation, we propose novel, efficient methods to search for **universal magic words** that attack text embedding models. Universal magic words as suffixes can shift the embedding of any text towards the bias direction, thus manipulating the similarity of any text pair and misleading safeguards. Attackers can jailbreak the safeguards by appending magic words to user prompts and requiring LLMs to end answers with magic words. Experiments show that magic word attacks significantly degrade safeguard performance on JailbreakBench, cause real-world chatbots to produce harmful outputs in full-pipeline attacks, and generalize across input/output texts, models, and languages. To eradicate this security risk, we also propose defense methods against such attacks, which can correct the bias of text embeddings and improve downstream performance in a train-free manner.
Authors: Saul Santos, Ant\'onio Farinhas, Daniel C. McNamee, Andr\'e F. T. Martins
Abstract: Current video-language models struggle with long-video understanding due to limited context lengths and reliance on sparse frame subsampling, often leading to information loss. This paper introduces $\infty$-Video, which can process arbitrarily long videos through a continuous-time long-term memory (LTM) consolidation mechanism. Our framework augments video Q-formers by allowing them to process unbounded video contexts efficiently and without requiring additional training. Through continuous attention, our approach dynamically allocates higher granularity to the most relevant video segments, forming "sticky" memories that evolve over time. Experiments with Video-LLaMA and VideoChat2 demonstrate improved performance in video question-answering tasks, showcasing the potential of continuous-time LTM mechanisms to enable scalable and training-free comprehension of long videos.
Authors: Alexander Nikulin, Ilya Zisman, Denis Tarasov, Nikita Lyubaykin, Andrei Polubarov, Igor Kiselev, Vladislav Kurenkov
Abstract: Recently, latent action learning, pioneered by Latent Action Policies (LAPO), have shown remarkable pre-training efficiency on observation-only data, offering potential for leveraging vast amounts of video available on the web for embodied AI. However, prior work has focused on distractor-free data, where changes between observations are primarily explained by ground-truth actions. Unfortunately, real-world videos contain action-correlated distractors that may hinder latent action learning. Using Distracting Control Suite (DCS) we empirically investigate the effect of distractors on latent action learning and demonstrate that LAPO struggle in such scenario. We propose LAOM, a simple LAPO modification that improves the quality of latent actions by 8x, as measured by linear probing. Importantly, we show that providing supervision with ground-truth actions, as few as 2.5% of the full dataset, during latent action learning improves downstream performance by 4.2x on average. Our findings suggest that integrating supervision during Latent Action Models (LAM) training is critical in the presence of distractors, challenging the conventional pipeline of first learning LAM and only then decoding from latent to ground-truth actions.
Authors: Maxime Meyer, Soumik Adhikary, Naixu Guo, Patrick Rebentrost
Abstract: Quantum state tomography, the task of learning an unknown quantum state, is a fundamental problem in quantum information. In standard settings, the complexity of this problem depends significantly on the type of quantum state that one is trying to learn, with pure states being substantially easier to learn than general mixed states. A natural question is whether this separation holds for any quantum state learning setting. In this work, we consider the online learning framework and prove the surprising result that learning pure states in this setting is as hard as learning mixed states. More specifically, we show that both classes share almost the same sequential fat-shattering dimension, leading to identical regret scaling. We also generalize previous results on full quantum state tomography in the online setting to (i) the $\epsilon$-realizable setting and (ii) learning the density matrix only partially, using smoothed analysis.
Authors: Wen Lai, Alexander Fraser, Ivan Titov
Abstract: Parameter-efficient fine-tuning (PEFT) methods, such as LoRA, are commonly used to adapt LLMs. However, the effectiveness of standard PEFT methods is limited in low-resource scenarios with only a few hundred examples. Recent advances in interpretability research have inspired the emergence of activation editing (or steering) techniques, which modify the activations of specific model components. These methods, due to their extremely small parameter counts, show promise for small datasets. However, their performance is highly dependent on identifying the correct modules to edit and often lacks stability across different datasets. In this paper, we propose Joint Localization and Activation Editing (JoLA), a method that jointly learns (1) which heads in the Transformer to edit (2) whether the intervention should be additive, multiplicative, or both and (3) the intervention parameters themselves - the vectors applied as additive offsets or multiplicative scalings to the head output. Through evaluations on three benchmarks spanning commonsense reasoning, natural language understanding, and natural language generation, we demonstrate that JoLA consistently outperforms existing methods. The code for the method is released at https://github.com/wenlai-lavine/jola.
Authors: Yichen Wang, Yudong Chen, Lorenzo Rosasco, Fanghui Liu
Abstract: Understanding how the test risk scales with model complexity is a central question in machine learning. Classical theory is challenged by the learning curves observed for large over-parametrized deep networks. Capacity measures based on parameter count typically fail to account for these empirical observations. To tackle this challenge, we consider norm-based capacity measures and develop our study for random features based estimators, widely used as simplified theoretical models for more complex networks. In this context, we provide a precise characterization of how the estimator's norm concentrates and how it governs the associated test error. Our results show that the predicted learning curve admits a phase transition from under- to over-parameterization, but no double descent behavior. This confirms that more classical U-shaped behavior is recovered considering appropriate capacity measures based on models norms rather than size. From a technical point of view, we leverage deterministic equivalence as the key tool and further develop new deterministic quantities which are of independent interest.
Authors: Da Yu, Edith Cohen, Badih Ghazi, Yangsibo Huang, Pritish Kamath, Ravi Kumar, Daogao Liu, Chiyuan Zhang
Abstract: We propose SCONE ($S$calable, $C$ontextualized, $O$ffloaded, $N$-gram $E$mbedding), a new method for extending input embedding layers to enhance language model performance. To avoid increased decoding costs, SCONE retains the original vocabulary while introducing embeddings for a set of frequent $n$-grams. These embeddings provide contextualized representation for each input token and are learned with a separate model during training. After training, embeddings are precomputed and stored in off-accelerator memory; during inference, querying them has minimal impact on latency due to the low complexity of embedding lookups. SCONE enables two new scaling strategies: increasing the number of $n$-gram embeddings and scaling the model used to learn them, both while maintaining fixed accelerator usage during inference (in terms of FLOPS and memory). We show that scaling both aspects enables a model with 1B accelerator-resident parameters to outperform a 1.9B-parameter baseline across diverse corpora, while using only about half the FLOPS and accelerator memory during inference.
Authors: Peter Chen, Bryan Chang, Olivia Annette Creasey, Julie Beth Sneddon, Zev Gartner, Yining Liu
Abstract: 3D cell segmentation methods are often hindered by \emph{oversegmentation}, where a single cell is incorrectly split into multiple fragments. This degrades the final segmentation quality and is notoriously difficult to resolve, as oversegmentation errors often resemble \emph{natural gaps} between adjacent cells. Our work makes two key contributions. First, for 3D cell segmentation, we are the first work to formulate oversegmentation as a concrete problem and propose a geometric framework to identify and correct these errors. Our approach builds a pre-trained classifier using both 2D geometric and 3D topological features extracted from flawed 3D segmentation results. Second, we introduce a novel metric, \emph{Geo-Wasserstein} divergence, to quantify changes in 2D geometries. This captures the evolving trends in cell mask shape changes in a geometry-aware manner. We validate our method through extensive experiments on in-domain plant datasets, including both synthesized and real cases, as well as on out-of-domain animal datasets to demonstrate transfer learning performance. An ablation study further highlights the contribution of the \emph{Geo-Wasserstein} divergence. A clear pipeline is provided for end-users to build pre-trained models to any labeled dataset.
Authors: Zelai Xu, Ruize Zhang, Chao Yu, Huining Yuan, Xiangmin Yi, Shilong Ji, Chuqi Wang, Wenhao Tang, Feng Gao, Wenbo Ding, Xinlei Chen, Yu Wang
Abstract: Robot sports, characterized by well-defined objectives, explicit rules, and dynamic interactions, present ideal scenarios for demonstrating embodied intelligence. In this paper, we present VolleyBots, a novel robot sports testbed where multiple drones cooperate and compete in the sport of volleyball under physical dynamics. VolleyBots integrates three features within a unified platform: competitive and cooperative gameplay, turn-based interaction structure, and agile 3D maneuvering. Competitive and cooperative gameplay challenges each drone to coordinate with its teammates while anticipating and countering opposing teams' tactics. Turn-based interaction demands precise timing, accurate state prediction, and management of long-horizon temporal dependencies. Agile 3D maneuvering requires rapid accelerations, sharp turns, and precise 3D positioning despite the quadrotor's underactuated dynamics. These intertwined features yield a complex problem combining motion control and strategic play, with no available expert demonstrations. We provide a comprehensive suite of tasks ranging from single-drone drills to multi-drone cooperative and competitive tasks, accompanied by baseline evaluations of representative multi-agent reinforcement learning (MARL) and game-theoretic algorithms. Simulation results show that on-policy reinforcement learning (RL) methods outperform off-policy methods in single-agent tasks, but both approaches struggle in complex tasks that combine motion control and strategic play. We additionally design a hierarchical policy which achieves a 69.5% percent win rate against the strongest baseline in the 3 vs 3 task, underscoring its potential as an effective solution for tackling the complex interplay between low-level control and high-level strategy. The project page is at https://sites.google.com/view/thu-volleybots.
Authors: Junguang Jiang, Yanwen Huang, Bin Liu, Xiaoyu Kong, Xinhang Li, Ziru Xu, Han Zhu, Jian Xu, Bo Zheng
Abstract: In real-world recommender systems, different retrieval objectives are typically addressed using task-specific datasets with carefully designed model architectures. We demonstrate that Large Language Models (LLMs) can function as universal retrievers, capable of handling multiple objectives within a generative retrieval framework. To model complex user-item relationships within generative retrieval, we propose multi-query representation. To address the challenge of extremely large candidate sets in industrial recommender systems, we introduce matrix decomposition to boost model learnability, discriminability, and transferability, and we incorporate probabilistic sampling to reduce computation costs. Finally, our Universal Retrieval Model (URM) can adaptively generate a set from tens of millions of candidates based on arbitrary given objective while keeping the latency within tens of milliseconds. Applied to industrial-scale data, URM outperforms expert models elaborately designed for different retrieval objectives on offline experiments and significantly improves the core metric of online advertising platform by $3\%$.
Authors: Xiaoyang Liu, Kangjie Bao, Jiashuo Zhang, Yunqi Liu, Yuntian Liu, Yu Chen, Yang Jiao, Tao Luo
Abstract: Autoformalization, the automatic translation of mathematical content from natural language into machine-verifiable formal languages, has seen significant progress driven by advances in large language models (LLMs). Nonetheless, a primary barrier to further improvements is the limited availability of parallel corpora that map informal mathematical text to its formal counterpart. To address this limitation, we propose ATLAS (Autoformalizing Theorems through Lifting, Augmentation, and Synthesis of Data), a novel data generation framework designed to produce large-scale, high-quality parallel corpora of theorem statements. Distinct from prior approaches, ATLAS begins with a concept repository, accelerates the improvement of student model through expert iteration combined with knowledge distillation, and introduces two novel augmentation strategies that exploit the structural characteristics of formal languages. With the proposed ATLAS running for 10 iterations, we construct an undergraduate-level dataset comprising 117k theorem statements and develop ATLAS Translator, which demonstrates statistically significant improvements over both the HERALD Translator and the Kimina-Autoformalizer across all benchmarks ($p<0.05$, two-sided t-test), achieving a new state of the art. The datasets, model, and code will be released to the public soon.
Authors: Yongcheng Zeng, Xinyu Cui, Xuanfa Jin, Guoqing Liu, Zexu Sun, Dong Li, Ning Yang, Jianye Hao, Haifeng Zhang, Jun Wang
Abstract: While large language models (LLMs) have demonstrated remarkable general performance, enabling smaller models to achieve capabilities comparable to their larger counterparts remains a critical challenge. For humans, iterative refinement of problem analysis and responses is a common strategy to enhance answer quality. However, we observe that existing LLMs exhibit limited ability to refine their outputs for quality improvement. In this paper, we first investigate mechanisms to unlock and progressively enhance self-refinement ability in smaller models within an iterative preference optimization framework, aiming to bridge the performance gap with larger models. To this end, we propose EVOLVE, a novel post-training and inference framework that iteratively integrates preference training with self-refinement-driven data collection. During training, EVOLVE strengthens the model's direct question-answering ability while simultaneously unlocking its self-refinement potential. At inference, the framework leverages this capability to generate progressively refined responses, which are filtered to construct datasets for subsequent rounds of preference training. Experiments demonstrate EVOLVE's exceptional performance: when applied to Llama-3.1-8B base model and under the self-refinement setting, it surpasses state-of-the-art models including Llama-3.1-405B-Instruct and GPT-4o, achieving a 62.3% length-controlled win rate and 63.3% raw win rate on AlpacaEval 2, along with a 50.3% win rate on Arena-Hard. Furthermore, EVOLVE consistently enhances performance on mathematical reasoning tasks like GSM8K and MATH.
Authors: Wenhao Wang, Adam Dziedzic, Grace C. Kim, Michael Backes, Franziska Boenisch
Abstract: Multi-modal models, such as CLIP, have demonstrated strong performance in aligning visual and textual representations, excelling in tasks like image retrieval and zero-shot classification. Despite this success, the mechanisms by which these models utilize training data, particularly the role of memorization, remain unclear. In uni-modal models, both supervised and self-supervised, memorization has been shown to be essential for generalization. However, it is not well understood how these findings would apply to CLIP, which incorporates elements from both supervised learning via captions that provide a supervisory signal similar to labels, and from self-supervised learning via the contrastive objective. To bridge this gap in understanding, we propose a formal definition of memorization in CLIP (CLIPMem) and use it to quantify memorization in CLIP models. Our results indicate that CLIP's memorization behavior falls between the supervised and self-supervised paradigms, with "mis-captioned" samples exhibiting highest levels of memorization. Additionally, we find that the text encoder contributes more to memorization than the image encoder, suggesting that mitigation strategies should focus on the text domain. Building on these insights, we propose multiple strategies to reduce memorization while at the same time improving utility--something that had not been shown before for traditional learning paradigms where reducing memorization typically results in utility decrease.
Authors: Erfan Zinvandi, Morteza Alikhani, Mehran Sarmadi, Zahra Pourbahman, Sepehr Arvin, Reza Kazemi, Arash Amini
Abstract: In this paper, we introduce a comprehensive benchmark for Persian (Farsi) text embeddings, built upon the Massive Text Embedding Benchmark (MTEB). Our benchmark includes 63 datasets spanning seven different tasks: classification, clustering, pair classification, reranking, retrieval, summary retrieval, and semantic textual similarity. The datasets are formed as a combination of existing, translated, and newly generated data, offering a diverse evaluation framework for Persian language models. Given the increasing use of text embedding models in chatbots, evaluation datasets are becoming inseparable ingredients in chatbot challenges and Retrieval-Augmented Generation systems. As a contribution, we include chatbot evaluation datasets in the MTEB benchmark for the first time. In addition, in this paper, we introduce the new task of summary retrieval which is not part of the tasks included in standard MTEB. Another contribution of this paper is the introduction of a substantial number of new Persian language NLP datasets suitable for training and evaluation, some of which have no previous counterparts in Persian. We evaluate the performance of several Persian and multilingual embedding models in a range of tasks. This work introduces an open-source benchmark with datasets, code and a public leaderboard.
Authors: Zihao Zhu, Hongbao Zhang, Ruotong Wang, Ke Xu, Siwei Lyu, Baoyuan Wu
Abstract: Large Reasoning Models (LRMs) are designed to solve complex tasks by generating explicit reasoning traces before producing final answers. However, we reveal a critical vulnerability in LRMs -- termed Unthinking Vulnerability -- wherein the thinking process can be bypassed by manipulating special delimiter tokens. It is empirically demonstrated to be widespread across mainstream LRMs, posing both a significant risk and potential utility, depending on how it is exploited. In this paper, we systematically investigate this vulnerability from both malicious and beneficial perspectives. On the malicious side, we introduce Breaking of Thought (BoT), a novel attack that enables adversaries to bypass the thinking process of LRMs, thereby compromising their reliability and availability. We present two variants of BoT: a training-based version that injects backdoor during the fine-tuning stage, and a training-free version based on adversarial attack during the inference stage. As a potential defense, we propose thinking recovery alignment to partially mitigate the vulnerability. On the beneficial side, we introduce Monitoring of Thought (MoT), a plug-and-play framework that allows model owners to enhance efficiency and safety. It is implemented by leveraging the same vulnerability to dynamically terminate redundant or risky reasoning through external monitoring. Extensive experiments show that BoT poses a significant threat to reasoning reliability, while MoT provides a practical solution for preventing overthinking and jailbreaking. Our findings expose an inherent flaw in current LRM architectures and underscore the need for more robust reasoning systems in the future.
Authors: Liviu Aolaritei, Zheyu Oliver Wang, Julie Zhu, Michael I. Jordan, Youssef Marzouk
Abstract: Conformal prediction provides a powerful framework for constructing prediction intervals with finite-sample guarantees, yet its robustness under distribution shifts remains a significant challenge. This paper addresses this limitation by modeling distribution shifts using Levy-Prokhorov (LP) ambiguity sets, which capture both local and global perturbations. We provide a self-contained overview of LP ambiguity sets and their connections to popular metrics such as Wasserstein and Total Variation. We show that the link between conformal prediction and LP ambiguity sets is a natural one: by propagating the LP ambiguity set through the scoring function, we reduce complex high-dimensional distribution shifts to manageable one-dimensional distribution shifts, enabling exact quantification of worst-case quantiles and coverage. Building on this analysis, we construct robust conformal prediction intervals that remain valid under distribution shifts, explicitly linking LP parameters to interval width and confidence levels. Experimental results on real-world datasets demonstrate the effectiveness of the proposed approach.
Authors: Natalia Koliou, George Vouros
Abstract: Game-theoretic solution concepts, such as the Nash equilibrium, have been key to finding stable joint actions in multi-player games. However, it has been shown that the dynamics of agents' interactions, even in simple two-player games with few strategies, are incapable of reaching Nash equilibria, exhibiting complex and unpredictable behavior. Instead, evolutionary approaches can describe the long-term persistence of strategies and filter out transient ones, accounting for the long-term dynamics of agents' interactions. Our goal is to identify agents' joint strategies that result in stable behavior, being resistant to changes, while also accounting for agents' payoffs, in dynamic games. Towards this goal, and building on previous results, this paper proposes transforming dynamic games into their empirical forms by considering agents' strategies instead of agents' actions, and applying the evolutionary methodology $\alpha$-Rank to evaluate and rank strategy profiles according to their long-term dynamics. This methodology not only allows us to identify joint strategies that are strong through agents' long-term interactions, but also provides a descriptive, transparent framework regarding the high ranking of these strategies. Experiments report on agents that aim to collaboratively solve a stochastic version of the graph coloring problem. We consider different styles of play as strategies to define the empirical game, and train policies realizing these strategies, using the DQN algorithm. Then we run simulations to generate the payoff matrix required by $\alpha$-Rank to rank joint strategies.
Authors: George Drayson, Emine Yilmaz, Vasileios Lampos
Abstract: As Large Language Models (LLMs) become increasingly prevalent, their generated outputs are proliferating across the web, risking a future where machine-generated content dilutes human-authored text. Since online data is the primary resource for LLM pre-training, subsequent models could be trained on an unknown portion of synthetic samples. This will lead to model collapse, a degenerative process whereby LLMs reinforce their own errors, converge to a low variance output distribution, and ultimately yield a declining performance. In this study, we investigate the impact of decoding strategy on model collapse, analysing the text characteristics at each model generation, the similarity to human references, and the resulting model performance. Using the decoding strategies that lead to the most significant degradation, we evaluate model collapse in more realistic scenarios where the origin of the data (human or synthetic) is unknown. We train a machine-generated text detector and propose an importance sampling approach to alleviate model collapse. Our method is validated on two LLM variants (GPT-2 and SmolLM2), across a range of model sizes (124M to 1.7B), on the open-ended text generation task. We demonstrate that it can not only prevent model collapse but also improve performance when sufficient human-authored samples are present. Source code: github.com/GeorgeDrayson/model_collapse.
Authors: Christian Marius Lillelund, Shi-ang Qi, Russell Greiner
Abstract: Conventional survival metrics, such as Harrell's concordance index (CI) and the Brier Score, rely on the independent censoring assumption for valid inference with right-censored data. However, in the presence of so-called dependent censoring, where the probability of censoring is related to the event of interest, these metrics can give biased estimates of the underlying model error. In this paper, we introduce three new evaluation metrics for survival analysis based on Archimedean copulas that can account for dependent censoring. We also develop a framework to generate realistic, semi-synthetic datasets with dependent censoring to facilitate the evaluation of the metrics. Our experiments in synthetic and semi-synthetic data demonstrate that the proposed metrics can provide more accurate estimates of the model error than conventional metrics under dependent censoring.
Authors: Ju-Hyung Lee, Andreas F. Molisch
Abstract: This paper introduces AutoBS, a reinforcement learning (RL)-based framework for optimal base station (BS) deployment in 6G radio access networks (RAN). AutoBS leverages the Proximal Policy Optimization (PPO) algorithm and fast, site-specific pathloss predictions from PMNet-a generative model for digital network twins (DNT). By efficiently learning deployment strategies that balance coverage and capacity, AutoBS achieves about 95% of the capacity of exhaustive search in single BS scenarios (and in 90% for multiple BSs), while cutting inference time from hours to milliseconds, making it highly suitable for real-time applications (e.g., ad-hoc deployments). AutoBS therefore provides a scalable, automated solution for large-scale 6G networks, meeting the demands of dynamic environments with minimal computational overhead.
Authors: Yihong Dong, Ge Li, Xue Jiang, Yongding Tao, Kechi Zhang, Hao Zhu, Huanyu Liu, Jiazheng Ding, Jia Li, Jinliang Deng, Hong Mei
Abstract: Periodicity, as one of the most important basic characteristics, lays the foundation for facilitating structured knowledge acquisition and systematic cognitive processes within human learning paradigms. However, the potential flaws of periodicity modeling in Transformer affect the learning efficiency and establishment of underlying principles from data for large language models (LLMs) built upon it. In this paper, we demonstrate that integrating effective periodicity modeling can improve the learning efficiency and performance of LLMs. We introduce FANformer, which adapts Fourier Analysis Network (FAN) into attention mechanism to achieve efficient periodicity modeling, by modifying the feature projection process of attention mechanism. Extensive experimental results on language modeling show that FANformer consistently outperforms Transformer when scaling up model size and training tokens, underscoring its superior learning efficiency. Our pretrained FANformer-1B exhibits marked improvements on downstream tasks compared to open-source LLMs with similar model parameters or training tokens. Moreover, we reveal that FANformer exhibits superior ability to learn and apply rules for reasoning compared to Transformer. The results position FANformer as an effective and promising architecture for advancing LLMs.
Authors: Samet Demir, Zafer Dogan
Abstract: In this work, we study the training and generalization performance of two-layer neural networks (NNs) after one gradient descent step under structured data modeled by Gaussian mixtures. While previous research has extensively analyzed this model under isotropic data assumption, such simplifications overlook the complexities inherent in real-world datasets. Our work addresses this limitation by analyzing two-layer NNs under Gaussian mixture data assumption in the asymptotically proportional limit, where the input dimension, number of hidden neurons, and sample size grow with finite ratios. We characterize the training and generalization errors by leveraging recent advancements in Gaussian universality. Specifically, we prove that a high-order polynomial model performs equivalent to the nonlinear neural networks under certain conditions. The degree of the equivalent model is intricately linked to both the "data spread" and the learning rate employed during one gradient step. Through extensive simulations, we demonstrate the equivalence between the original model and its polynomial counterpart across various regression and classification tasks. Additionally, we explore how different properties of Gaussian mixtures affect learning outcomes. Finally, we illustrate experimental results on Fashion-MNIST classification, indicating that our findings can translate to realistic data.
Authors: Yicong Zheng, Nora Wolf, Charan Ranganath, Randall C. O'Reilly, Kevin L. McKee
Abstract: Many tasks require flexibly modifying perception and behavior based on current goals. Humans can retrieve episodic memories from days to years ago, using them to contextualize and generalize behaviors across novel but structurally related situations. The brain's ability to control episodic memories based on task demands is often attributed to interactions between the prefrontal cortex (PFC) and hippocampus (HPC). We propose a reinforcement learning model that incorporates a PFC-HPC interaction mechanism for goal-directed generalization. In our model, the PFC learns to generate query-key representations to encode and retrieve goal-relevant episodic memories, modulating HPC memories top-down based on current task demands. Moreover, the PFC adapts its encoding and retrieval strategies dynamically when faced with multiple goals presented in a blocked, rather than interleaved, manner. Our results show that: (1) combining working memory with selectively retrieved episodic memory allows transfer of decisions among similar environments or situations, (2) top-down control from PFC over HPC improves learning of arbitrary structural associations between events for generalization to novel environments compared to a bottom-up sensory-driven approach, and (3) the PFC encodes generalizable representations during both encoding and retrieval of goal-relevant memories, whereas the HPC exhibits event-specific representations. Together, these findings highlight the importance of goal-directed prefrontal control over hippocampal episodic memory for decision-making in novel situations and suggest a computational mechanism by which PFC-HPC interactions enable flexible behavior.
Authors: Changliang Zhou, Xi Lin, Zhenkun Wang, Qingfu Zhang
Abstract: Constructive neural combinatorial optimization (NCO) has attracted growing research attention due to its ability to solve complex routing problems without relying on handcrafted rules. However, existing NCO methods face significant challenges in generalizing to large-scale problems due to high computational complexity and inefficient capture of structural patterns. To address this issue, we propose a novel learning-based search space reduction method that adaptively selects a small set of promising candidate nodes at each step of the constructive NCO process. Unlike traditional methods that rely on fixed heuristics, our selection model dynamically prioritizes nodes based on learned patterns, significantly reducing the search space while maintaining solution quality. Experimental results demonstrate that our method, trained solely on 100-node instances from uniform distribution, generalizes remarkably well to large-scale Traveling Salesman Problem (TSP) and Capacitated Vehicle Routing Problem (CVRP) instances with up to 1 million nodes from the uniform distribution and over 80K nodes from other distributions.
Authors: Run He, Di Fang, Yicheng Xu, Yawen Cui, Ming Li, Cen Chen, Ziqian Zeng, Huiping Zhuang
Abstract: Exemplar-Free Class-Incremental Learning (EFCIL) aims to sequentially learn from distinct categories without retaining exemplars but easily suffers from catastrophic forgetting of learned knowledge. While existing EFCIL methods leverage knowledge distillation to alleviate forgetting, they still face two critical challenges: semantic shift and decision bias. Specifically, the embeddings of old tasks shift in the embedding space after learning new tasks, and the classifier becomes biased towards new tasks due to training solely with new data, hindering the balance between old and new knowledge. To address these issues, we propose the Dual-Projection Shift Estimation and Classifier Reconstruction (DPCR) approach for EFCIL. DPCR effectively estimates semantic shift through a dual-projection, which combines a learnable transformation with a row-space projection to capture both task-wise and category-wise shifts. Furthermore, to mitigate decision bias, DPCR employs ridge regression to reformulate a classifier reconstruction process. This reconstruction exploits previous in covariance and prototype of each class after calibration with estimated shift, thereby reducing decision bias. Extensive experiments demonstrate that, on various datasets, DPCR effectively balances old and new tasks, outperforming state-of-the-art EFCIL methods. Our codes are available at https://github.com/RHe502/ICML25-DPCR.
Authors: Oskar van der Wal, Pietro Lesci, Max Muller-Eberstein, Naomi Saphra, Hailey Schoelkopf, Willem Zuidema, Stella Biderman
Abstract: The stability of language model pre-training and its effects on downstream performance are still understudied. Prior work shows that the training process can yield significantly different results in response to slight variations in initial conditions, e.g., the random seed. Crucially, the research community still lacks sufficient resources and tools to systematically investigate pre-training stability, particularly for decoder-only language models. We introduce the PolyPythias, a set of 45 new training runs for the Pythia model suite: 9 new seeds across 5 model sizes, from 14M to 410M parameters, resulting in about 7k new checkpoints that we release. Using these new 45 training runs, in addition to the 5 already available, we study the effects of different initial conditions determined by the seed -- i.e., parameters' initialisation and data order -- on (i) downstream performance, (ii) learned linguistic representations, and (iii) emergence of training phases. In addition to common scaling behaviours, our analyses generally reveal highly consistent training dynamics across both model sizes and initial conditions. Further, the new seeds for each model allow us to identify outlier training runs and delineate their characteristics. Our findings show the potential of using these methods to predict training stability.
Authors: Jonathan Zheng, Sauvik Das, Alan Ritter, Wei Xu
Abstract: Probabilistic reasoning is a key aspect of both human and artificial intelligence that allows for handling uncertainty and ambiguity in decision-making. In this paper, we introduce a new numerical reasoning task under uncertainty for large language models, focusing on estimating the privacy risk of user-generated documents containing privacy-sensitive information. We propose BRANCH, a new LLM methodology that estimates the k-privacy value of a text-the size of the population matching the given information. BRANCH factorizes a joint probability distribution of personal information as random variables. The probability of each factor in a population is estimated separately using a Bayesian network and combined to compute the final k-value. Our experiments show that this method successfully estimates the k-value 73% of the time, a 13% increase compared to o3-mini with chain-of-thought reasoning. We also find that LLM uncertainty is a good indicator for accuracy, as high-variance predictions are 37.47% less accurate on average.
Authors: Adam \v{S}torek, Mukur Gupta, Noopur Bhatt, Aditya Gupta, Janie Kim, Prashast Srivastava, Suman Jana
Abstract: AI coding assistants are widely used for tasks like code generation. These tools now require large and complex contexts, automatically sourced from various origins$\unicode{x2014}$across files, projects, and contributors$\unicode{x2014}$forming part of the prompt fed to underlying LLMs. This automatic context-gathering introduces new vulnerabilities, allowing attackers to subtly poison input to compromise the assistant's outputs, potentially generating vulnerable code or introducing critical errors. We propose a novel attack, Cross-Origin Context Poisoning (XOXO), that is challenging to detect as it relies on adversarial code modifications that are semantically equivalent. Traditional program analysis techniques struggle to identify these perturbations since the semantics of the code remains correct, making it appear legitimate. This allows attackers to manipulate coding assistants into producing incorrect outputs, while shifting the blame to the victim developer. We introduce a novel, task-agnostic, black-box attack algorithm GCGS that systematically searches the transformation space using a Cayley Graph, achieving a 75.72% attack success rate on average across five tasks and eleven models, including GPT 4.1 and Claude 3.5 Sonnet v2 used by popular AI coding assistants. Furthermore, defenses like adversarial fine-tuning are ineffective against our attack, underscoring the need for new security measures in LLM-powered coding tools.
Authors: Kasra Borazjani, Payam Abdisarabshali, Naji Khosravan, Seyyedali Hosseinalipour
Abstract: Federated Learning (FL) represents a paradigm shift in distributed machine learning (ML), enabling clients to train models collaboratively while keeping their raw data private. This paradigm shift from traditional centralized ML introduces challenges due to the non-iid (non-independent and identically distributed) nature of data across clients, significantly impacting FL's performance. Existing literature, predominantly model data heterogeneity by imposing label distribution skew across clients. In this paper, we show that label distribution skew fails to fully capture the real-world data heterogeneity among clients in computer vision tasks beyond classification. Subsequently, we demonstrate that current approaches overestimate FL's performance by relying on label/class distribution skew, exposing an overlooked gap in the literature. By utilizing pre-trained deep neural networks to extract task-specific data embeddings, we define task-specific data heterogeneity through the lens of each vision task and introduce a new level of data heterogeneity called embedding-based data heterogeneity. Our methodology involves clustering data points based on embeddings and distributing them among clients using the Dirichlet distribution. Through extensive experiments, we evaluate the performance of different FL methods under our revamped notion of data heterogeneity, introducing new benchmark performance measures to the literature. We further unveil a series of open research directions that can be pursued.
Authors: Minh-Nhat Phung, Minh-Binh Tran
Abstract: We study the fundamental computational problem of approximating optimal transport (OT) equations using neural differential equations (Neural ODEs). More specifically, we develop a novel framework for approximating unbalanced optimal transport (UOT) in the continuum using Neural ODEs. By generalizing a discrete UOT problem with Pearson divergence, we constructively design vector fields for Neural ODEs that converge to the true UOT dynamics, thereby advancing the mathematical foundations of computational transport and machine learning. To this end, we design a numerical scheme inspired by the Sinkhorn algorithm to solve the corresponding minimization problem and rigorously prove its convergence, providing explicit error estimates. From the obtained numerical solutions, we derive vector fields defining the transport dynamics and construct the corresponding transport equation. Finally, from the numerically obtained transport equation, we construct a neural differential equation whose flow converges to the true transport dynamics in an appropriate limiting regime.
Authors: NVIDIA, :, Alisson Azzolini, Junjie Bai, Hannah Brandon, Jiaxin Cao, Prithvijit Chattopadhyay, Huayu Chen, Jinju Chu, Yin Cui, Jenna Diamond, Yifan Ding, Liang Feng, Francesco Ferroni, Rama Govindaraju, Jinwei Gu, Siddharth Gururani, Imad El Hanafi, Zekun Hao, Jacob Huffman, Jingyi Jin, Brendan Johnson, Rizwan Khan, George Kurian, Elena Lantz, Nayeon Lee, Zhaoshuo Li, Xuan Li, Maosheng Liao, Tsung-Yi Lin, Yen-Chen Lin, Ming-Yu Liu, Xiangyu Lu, Alice Luo, Andrew Mathau, Yun Ni, Lindsey Pavao, Wei Ping, David W. Romero, Misha Smelyanskiy, Shuran Song, Lyne Tchapmi, Andrew Z. Wang, Boxin Wang, Haoxiang Wang, Fangyin Wei, Jiashu Xu, Yao Xu, Dinghao Yang, Xiaodong Yang, Zhuolin Yang, Jingxu Zhang, Xiaohui Zeng, Zhe Zhang
Abstract: Physical AI systems need to perceive, understand, and perform complex actions in the physical world. In this paper, we present the Cosmos-Reason1 models that can understand the physical world and generate appropriate embodied decisions (e.g., next step action) in natural language through long chain-of-thought reasoning processes. We begin by defining key capabilities for Physical AI reasoning, with a focus on physical common sense and embodied reasoning. To represent physical common sense, we use a hierarchical ontology that captures fundamental knowledge about space, time, and physics. For embodied reasoning, we rely on a two-dimensional ontology that generalizes across different physical embodiments. Building on these capabilities, we develop two multimodal large language models, Cosmos-Reason1-7B and Cosmos-Reason1-56B. We curate data and train our models in two stages: Physical AI supervised fine-tuning (SFT) and Physical AI reinforcement learning (RL). To evaluate our models, we build comprehensive benchmarks for physical common sense and embodied reasoning according to our ontologies. Evaluation results show that Physical AI SFT and RL bring significant improvements. To facilitate the development of Physical AI, we make our code and pre-trained models available under the NVIDIA Open Model License at https://github.com/nvidia-cosmos/cosmos-reason1.
Authors: Daniele Bracale, Moulinath Banerjee, Cong Shi, Yuekai Sun
Abstract: We consider price competition among multiple sellers over a selling horizon of $T$ periods. In each period, sellers simultaneously offer their prices and subsequently observe their respective demand that is unobservable to competitors. The demand function for each seller depends on all sellers' prices through a private, unknown, and nonlinear relationship. To address this challenge, we propose a semi-parametric least-squares estimation of the nonlinear mean function, which does not require sellers to communicate demand information. We show that when all sellers employ our policy, their prices converge at a rate of $O(T^{-1/7})$ to the Nash equilibrium prices that sellers would reach if they were fully informed. Each seller incurs a regret of $O(T^{5/7})$ relative to a dynamic benchmark policy. A theoretical contribution of our work is proving the existence of equilibrium under shape-constrained demand functions via the concept of $s$-concavity and establishing regret bounds of our proposed policy. Technically, we also establish new concentration results for the least squares estimator under shape constraints. Our findings offer significant insights into dynamic competition-aware pricing and contribute to the broader study of non-parametric learning in strategic decision-making.
Authors: Sven Klaassen, Jan Rabenseifner, Jannis Kueck, Philipp Bach
Abstract: The partitioning of data for estimation and calibration critically impacts the performance of propensity score based estimators like inverse probability weighting (IPW) and double/debiased machine learning (DML) frameworks. We extend recent advances in calibration techniques for propensity score estimation, improving the robustness of propensity scores in challenging settings such as limited overlap, small sample sizes, or unbalanced data. Our contributions are twofold: First, we provide a theoretical analysis of the properties of calibrated estimators in the context of DML. To this end, we refine existing calibration frameworks for propensity score models, with a particular emphasis on the role of sample-splitting schemes in ensuring valid causal inference. Second, through extensive simulations, we show that calibration reduces variance of inverse-based propensity score estimators while also mitigating bias in IPW, even in small-sample regimes. Notably, calibration improves stability for flexible learners (e.g., gradient boosting) while preserving the doubly robust properties of DML. A key insight is that, even when methods perform well without calibration, incorporating a calibration step does not degrade performance, provided that an appropriate sample-splitting approach is chosen.
Authors: Yuheng Ding, Bo Qiang, Yiran Zhou, Jie Yu, Qi Li, Liangren Zhang, Yusong Wang, Zhenmin Liu
Abstract: Natural products, as metabolites from microorganisms, animals, or plants, exhibit diverse biological activities, making them crucial for drug discovery. Nowadays, existing deep learning methods for natural products research primarily rely on supervised learning approaches designed for specific downstream tasks. However, such one-model-for-a-task paradigm often lacks generalizability and leaves significant room for performance improvement. Additionally, existing molecular characterization methods are not well-suited for the unique tasks associated with natural products. To address these limitations, we have pre-trained a foundation model for natural products based on their unique properties. Our approach employs a novel pretraining strategy that is especially tailored to natural products. By incorporating contrastive learning and masked graph learning objectives, we emphasize evolutional information from molecular scaffolds while capturing side-chain information. Our framework achieves state-of-the-art (SOTA) results in various downstream tasks related to natural product mining and drug discovery. We first compare taxonomy classification with synthesized molecule-focused baselines to demonstrate that current models are inadequate for understanding natural synthesis. Furthermore, by diving into a fine-grained analysis at both the gene and microbial levels, NaFM demonstrates the ability to capture evolutionary information. Eventually, our method is experimented with virtual screening, illustrating informative natural product representations that can lead to more effective identification of potential drug candidates.
Authors: Dor Elimelech, Wasim Huleihel
Abstract: The problems of detecting and recovering planted structures/subgraphs in Erd\H{o}s-R\'{e}nyi random graphs, have received significant attention over the past three decades, leading to many exciting results and mathematical techniques. However, prior work has largely focused on specific ad hoc planted structures and inferential settings, while a general theory has remained elusive. In this paper, we bridge this gap by investigating the detection of an \emph{arbitrary} planted subgraph $\Gamma = \Gamma_n$ in an Erd\H{o}s-R\'{e}nyi random graph $\mathcal{G}(n, q_n)$, where the edge probability within $\Gamma$ is $p_n$. We examine both the statistical and computational aspects of this problem and establish the following results. In the dense regime, where the edge probabilities $p_n$ and $q_n$ are fixed, we tightly characterize the information-theoretic and computational thresholds for detecting $\Gamma$, and provide conditions under which a computational-statistical gap arises. Most notably, these thresholds depend on $\Gamma$ only through its number of edges, maximum degree, and maximum subgraph density. Our lower and upper bounds are general and apply to any value of $p_n$ and $q_n$ as functions of $n$. Accordingly, we also analyze the sparse regime where $q_n = \Theta(n^{-\alpha})$ and $p_n-q_n =\Theta(q_n)$, with $\alpha\in[0,2]$, as well as the critical regime where $p_n=1-o(1)$ and $q_n = \Theta(n^{-\alpha})$, both of which have been widely studied, for specific choices of $\Gamma$. For these regimes, we show that our bounds are tight for all planted subgraphs investigated in the literature thus far\textemdash{}and many more. Finally, we identify conditions under which detection undergoes sharp phase transition, where the boundaries at which algorithms succeed or fail shift abruptly as a function of $q_n$.
Authors: Vladimir Vargas-Calder\'on
Abstract: This paper introduces the quantum deep sets model, expanding the quantum machine learning tool-box by enabling the possibility of learning variadic functions using quantum systems. A couple of variants are presented for this model. The first one focuses on mapping sets to quantum systems through state vector averaging: each element of the set is mapped to a quantum state, and the quantum state of the set is the average of the corresponding quantum states of its elements. This approach allows the definition of a permutation-invariant variadic model. The second variant is useful for ordered sets, i.e., sequences, and relies on optimal coherification of tristochastic tensors that implement products of mixed states: each element of the set is mapped to a density matrix, and the quantum state of the set is the product of the corresponding density matrices of its elements. Such variant can be relevant in tasks such as natural language processing. The resulting quantum state in any of the variants is then processed to realise a function that solves a machine learning task such as classification, regression or density estimation. Through synthetic problem examples, the efficacy and versatility of quantum deep sets and sequences (QDSs) is demonstrated.
Authors: Amit Rand, Hadi Ibrahim
Abstract: Medical imaging, particularly X-ray analysis, often involves detecting multiple conditions simultaneously within a single scan, making multi-label classification crucial for real-world clinical applications. We present the Medical X-ray Attention (MXA) block, a novel attention mechanism tailored specifically to address the unique challenges of X-ray abnormality detection. The MXA block enhances traditional Multi-Head Self Attention (MHSA) by integrating a specialized module that efficiently captures both detailed local information and broader global context. To the best of our knowledge, this is the first work to propose a task-specific attention mechanism for diagnosing chest X-rays, as well as to attempt multi-label classification using an Efficient Vision Transformer (EfficientViT). By embedding the MXA block within the EfficientViT architecture and employing knowledge distillation, our proposed model significantly improves performance on the CheXpert dataset, a widely used benchmark for multi-label chest X-ray abnormality detection. Our approach achieves an area under the curve (AUC) of 0.85, an absolute improvement of 0.19 compared to our baseline model's AUC of 0.66, corresponding to a substantial approximate 233% relative improvement over random guessing (AUC = 0.5).
Authors: Weizheng Zhang, Hao Pan, Lin Lu, Xiaowei Duan, Xin Yan, Ruonan Wang, Qiang Du
Abstract: Heat exchangers are critical components in a wide range of engineering applications, from energy systems to chemical processing, where efficient thermal management is essential. The design objectives for heat exchangers include maximizing the heat exchange rate while minimizing the pressure drop, requiring both a large interface area and a smooth internal structure. State-of-the-art designs, such as triply periodic minimal surfaces (TPMS), have proven effective in optimizing heat exchange efficiency. However, TPMS designs are constrained by predefined mathematical equations, limiting their adaptability to freeform boundary shapes. Additionally, TPMS structures do not inherently control flow directions, which can lead to flow stagnation and undesirable pressure drops. This paper presents DualMS, a novel computational framework for optimizing dual-channel minimal surfaces specifically for heat exchanger designs in freeform shapes. To the best of our knowledge, this is the first attempt to directly optimize minimal surfaces for two-fluid heat exchangers, rather than relying on TPMS. Our approach formulates the heat exchange maximization problem as a constrained connected maximum cut problem on a graph, with flow constraints guiding the optimization process. To address undesirable pressure drops, we model the minimal surface as a classification boundary separating the two fluids, incorporating an additional regularization term for area minimization. We employ a neural network that maps spatial points to binary flow types, enabling it to classify flow skeletons and automatically determine the surface boundary. DualMS demonstrates greater flexibility in surface topology compared to TPMS and achieves superior thermal performance, with lower pressure drops while maintaining a similar heat exchange rate under the same material cost.
Authors: Shan An, Ziyu Meng, Chao Tang, Yuning Zhou, Tengyu Liu, Fangqiang Ding, Shufang Zhang, Yao Mu, Ran Song, Wei Zhang, Zeng-Guang Hou, Hong Zhang
Abstract: Dexterous manipulation, which refers to the ability of a robotic hand or multi-fingered end-effector to skillfully control, reorient, and manipulate objects through precise, coordinated finger movements and adaptive force modulation, enables complex interactions similar to human hand dexterity. With recent advances in robotics and machine learning, there is a growing demand for these systems to operate in complex and unstructured environments. Traditional model-based approaches struggle to generalize across tasks and object variations due to the high dimensionality and complex contact dynamics of dexterous manipulation. Although model-free methods such as reinforcement learning (RL) show promise, they require extensive training, large-scale interaction data, and carefully designed rewards for stability and effectiveness. Imitation learning (IL) offers an alternative by allowing robots to acquire dexterous manipulation skills directly from expert demonstrations, capturing fine-grained coordination and contact dynamics while bypassing the need for explicit modeling and large-scale trial-and-error. This survey provides an overview of dexterous manipulation methods based on imitation learning, details recent advances, and addresses key challenges in the field. Additionally, it explores potential research directions to enhance IL-driven dexterous manipulation. Our goal is to offer researchers and practitioners a comprehensive introduction to this rapidly evolving domain.
Authors: Mohammad Amaan Sayeed, Engin Tekin, Maryam Nadeem, Nancy A. ElNaker, Aahan Singh, Natalia Vassilieva, Boulbaba Ben Amor
Abstract: Unlocking the next generation of biotechnology and therapeutic innovation demands overcoming the inherent complexity and resource-intensity of conventional protein engineering methods. Recent GenAI-powered computational techniques often rely on the availability of the target protein's 3D structures and specific binding sites to generate high-affinity binders, constraints exhibited by models such as AlphaProteo and RFdiffusion. In this work, we explore the use of Protein Language Models (pLMs) for high-affinity binder generation. We introduce Prot42, a novel family of Protein Language Models (pLMs) pretrained on vast amounts of unlabeled protein sequences. By capturing deep evolutionary, structural, and functional insights through an advanced auto-regressive, decoder-only architecture inspired by breakthroughs in natural language processing, Prot42 dramatically expands the capabilities of computational protein design based on language only. Remarkably, our models handle sequences up to 8,192 amino acids, significantly surpassing standard limitations and enabling precise modeling of large proteins and complex multi-domain sequences. Demonstrating powerful practical applications, Prot42 excels in generating high-affinity protein binders and sequence-specific DNA-binding proteins. Our innovative models are publicly available, offering the scientific community an efficient and precise computational toolkit for rapid protein engineering.
Authors: Donghao Ren, Fred Hohman, Dominik Moritz
Abstract: Interactive visualization of embedding projections is a useful technique for understanding data and evaluating machine learning models. Labeling data within these visualizations is critical for interpretation, as labels provide an overview of the projection and guide user navigation. However, most methods for producing labels require clustering the points, which can be computationally expensive as the number of points grows. In this paper, we describe an efficient clustering approach using kernel density estimation in the projected 2D space instead of points. This algorithm can produce high-quality cluster regions from a 2D density map in a few hundred milliseconds, orders of magnitude faster than current approaches. We contribute the design of the algorithm, benchmarks, and applications that demonstrate the utility of the algorithm, including labeling and summarization.
Authors: Zhixuan Pan, Shaowen Wang, Jian Li
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities across numerous tasks, yet principled explanations for their underlying mechanisms and several phenomena, such as scaling laws, hallucinations, and related behaviors, remain elusive. In this work, we revisit the classical relationship between compression and prediction, grounded in Kolmogorov complexity and Shannon information theory, to provide deeper insights into LLM behaviors. By leveraging the Kolmogorov Structure Function and interpreting LLM compression as a two-part coding process, we offer a detailed view of how LLMs acquire and store information across increasing model and data scales -- from pervasive syntactic patterns to progressively rarer knowledge elements. Motivated by this theoretical perspective and natural assumptions inspired by Heap's and Zipf's laws, we introduce a simplified yet representative hierarchical data-generation framework called the Syntax-Knowledge model. Under the Bayesian setting, we show that prediction and compression within this model naturally lead to diverse learning and scaling behaviors of LLMs. In particular, our theoretical analysis offers intuitive and principled explanations for both data and model scaling laws, the dynamics of knowledge acquisition during training and fine-tuning, factual knowledge hallucinations in LLMs. The experimental results validate our theoretical predictions.
Authors: Guanyuan Pan, Tiansheng Zhou, Bingtao Ma, Yaqi Wang, Jianxiang Zhao, Zhi Li, Yugui Lin, Pietro Lio, Shuai Wang
Abstract: Circuit link prediction identifying missing component connections from incomplete netlists is crucial in automating analog circuit design. However, existing methods face three main challenges: 1) Insufficient use of topological patterns in circuit graphs reduces prediction accuracy; 2) Data scarcity due to the complexity of annotations hinders model generalization; 3) Limited adaptability to various netlist formats. We propose GNN-ACLP, a Graph Neural Networks (GNNs) based framework featuring three innovations to tackle these challenges. First, we introduce the SEAL (Subgraphs, Embeddings, and Attributes for Link Prediction) framework and achieve port-level accuracy in circuit link prediction. Second, we propose Netlist Babel Fish, a netlist format conversion tool leveraging retrieval-augmented generation (RAG) with a large language model (LLM) to enhance the compatibility of netlist formats. Finally, we construct SpiceNetlist, a comprehensive dataset that contains 775 annotated circuits across 10 different component classes. Experimental results achieve accuracy improvements of 16.08% on SpiceNetlist, 11.38% on Image2Net, and 16.01% on Masala-CHAI in intra-dataset evaluation, while maintaining accuracy from 92.05% to 99.07% in cross-dataset evaluation, exhibiting robust feature transfer capabilities.
Authors: Kun Wang, Guibin Zhang, Zhenhong Zhou, Jiahao Wu, Miao Yu, Shiqian Zhao, Chenlong Yin, Jinhu Fu, Yibo Yan, Hanjun Luo, Liang Lin, Zhihao Xu, Haolang Lu, Xinye Cao, Xinyun Zhou, Weifei Jin, Fanci Meng, Junyuan Mao, Yu Wang, Hao Wu, Minghe Wang, Fan Zhang, Junfeng Fang, Wenjie Qu, Yue Liu, Chengwei Liu, Yifan Zhang, Qiankun Li, Chongye Guo, Yalan Qin, Zhaoxin Fan, Yi Ding, Donghai Hong, Jiaming Ji, Yingxin Lai, Zitong Yu, Xinfeng Li, Yifan Jiang, Yanhui Li, Xinyu Deng, Junlin Wu, Dongxia Wang, Yihao Huang, Yufei Guo, Jen-tse Huang, Qiufeng Wang, Wenxuan Wang, Dongrui Liu, Yanwei Yue, Wenke Huang, Guancheng Wan, Heng Chang, Tianlin Li, Yi Yu, Chenghao Li, Jiawei Li, Lei Bai, Jie Zhang, Qing Guo, Jingyi Wang, Tianlong Chen, Joey Tianyi Zhou, Xiaojun Jia, Weisong Sun, Cong Wu, Jing Chen, Xuming Hu, Yiming Li, Xiao Wang, Ningyu Zhang, Luu Anh Tuan, Guowen Xu, Jiaheng Zhang, Tianwei Zhang, Xingjun Ma, Jindong Gu, Xiang Wang, Bo An, Jun Sun, Mohit Bansal, Shirui Pan, Lingjuan Lyu, Yuval Elovici, Bhavya Kailkhura, Yaodong Yang, Hongwei Li, Wenyuan Xu, Yizhou Sun, Wei Wang, Qing Li, Ke Tang, Yu-Gang Jiang, Felix Juefei-Xu, Hui Xiong, Xiaofeng Wang, Dacheng Tao, Philip S. Yu, Qingsong Wen, Yang Liu
Abstract: The remarkable success of Large Language Models (LLMs) has illuminated a promising pathway toward achieving Artificial General Intelligence for both academic and industrial communities, owing to their unprecedented performance across various applications. As LLMs continue to gain prominence in both research and commercial domains, their security and safety implications have become a growing concern, not only for researchers and corporations but also for every nation. Currently, existing surveys on LLM safety primarily focus on specific stages of the LLM lifecycle, e.g., deployment phase or fine-tuning phase, lacking a comprehensive understanding of the entire "lifechain" of LLMs. To address this gap, this paper introduces, for the first time, the concept of "full-stack" safety to systematically consider safety issues throughout the entire process of LLM training, deployment, and eventual commercialization. Compared to the off-the-shelf LLM safety surveys, our work demonstrates several distinctive advantages: (I) Comprehensive Perspective. We define the complete LLM lifecycle as encompassing data preparation, pre-training, post-training, deployment and final commercialization. To our knowledge, this represents the first safety survey to encompass the entire lifecycle of LLMs. (II) Extensive Literature Support. Our research is grounded in an exhaustive review of over 800+ papers, ensuring comprehensive coverage and systematic organization of security issues within a more holistic understanding. (III) Unique Insights. Through systematic literature analysis, we have developed reliable roadmaps and perspectives for each chapter. Our work identifies promising research directions, including safety in data generation, alignment techniques, model editing, and LLM-based agent systems. These insights provide valuable guidance for researchers pursuing future work in this field.
Authors: Junwei Liao, Muning Wen, Jun Wang, Weinan Zhang
Abstract: LLM-based Multi-Agent Systems have demonstrated remarkable capabilities in addressing complex, agentic tasks, from generating high-quality presentation slides to even conducting sophisticated scientific research. Meanwhile, RL has been widely recognized for its effectiveness in enhancing agent intelligence, but limited research has investigated the fine-tuning of LaMAS using foundational RL techniques. Moreover, the direct application of MARL methods to LaMAS introduces significant challenges, stemming from the unique characteristics and mechanisms inherent to LaMAS. To address these challenges, this article presents a comprehensive study of LLM-based MARL and proposes a novel paradigm termed Multi-Agent Reinforcement Fine-Tuning (MARFT). We introduce a brand-new POMDP called Flex-POMDP, which aligns with the LaMAS optimization in real-world applications and a universal algorithmic framework tailored specifically for LaMAS, outlining the conceptual foundations, key distinctions, and practical implementation strategies. We review the evolution from RL to RFT, setting the stage for a parallel analysis in the multi-agent domain. In the context of LaMAS, we elucidate critical differences between MARL and MARFT. These differences motivate a transition toward a LaMAS-oriented formulation of RFT. Central to this work is a robust and scalable MARFT framework. We detail the core algorithm and provide a complete, open-source implementation to facilitate adoption and further research. The latter sections of the paper explore real-world application perspectives and opening challenges in MARFT. By bridging theoretical underpinnings with practical methodologies, this work serves as a roadmap for researchers seeking to advance MARFT toward resilient and adaptive solutions in agentic systems. Our implementation of the proposed framework is publicly available at: https://github.com/jwliao-ai/MARFT.
Authors: Jiaqi Chen, Bang Zhang, Ruotian Ma, Peisong Wang, Xiaodan Liang, Zhaopeng Tu, Xiaolong Li, Kwan-Yee K. Wong
Abstract: Evaluating the step-by-step reliability of large language model (LLM) reasoning, such as Chain-of-Thought, remains challenging due to the difficulty and cost of obtaining high-quality step-level supervision. In this paper, we introduce Self-Play Critic (SPC), a novel approach where a critic model evolves its ability to assess reasoning steps through adversarial self-play games, eliminating the need for manual step-level annotation. SPC involves fine-tuning two copies of a base model to play two roles, namely a "sneaky generator" that deliberately produces erroneous steps designed to be difficult to detect, and a "critic" that analyzes the correctness of reasoning steps. These two models engage in an adversarial game in which the generator aims to fool the critic, while the critic model seeks to identify the generator's errors. Using reinforcement learning based on the game outcomes, the models iteratively improve; the winner of each confrontation receives a positive reward and the loser receives a negative reward, driving continuous self-evolution. Experiments on three reasoning process benchmarks (ProcessBench, PRM800K, DeltaBench) demonstrate that our SPC progressively enhances its error detection capabilities (e.g., accuracy increases from 70.8% to 77.7% on ProcessBench) and surpasses strong baselines, including distilled R1 model. Furthermore, SPC can guide the test-time search of diverse LLMs and significantly improve their mathematical reasoning performance on MATH500 and AIME2024, surpassing those guided by state-of-the-art process reward models.
Authors: Kaichen Zhang, Yuzhong Hong, Junwei Bao, Hongfei Jiang, Yang Song, Dingqian Hong, Hui Xiong
Abstract: Post-training plays a crucial role in refining and aligning large language models to meet specific tasks and human preferences. While recent advancements in post-training techniques, such as Group Relative Policy Optimization (GRPO), leverage increased sampling with relative reward scoring to achieve superior performance, these methods often suffer from training instability that limits their practical adoption. To address this challenge, we present Group Variance Policy Optimization (GVPO). GVPO incorporates the analytical solution to KL-constrained reward maximization directly into its gradient weights, ensuring alignment with the optimal policy. The method provides intuitive physical interpretations: its gradient mirrors the mean squared error between the central distance of implicit rewards and that of actual rewards. GVPO offers two key advantages: (1) it guarantees a unique optimal solution, exactly the KL-constrained reward maximization objective, (2) it supports flexible sampling distributions that avoids on-policy and importance sampling limitations. By unifying theoretical guarantees with practical adaptability, GVPO establishes a new paradigm for reliable and versatile LLM post-training.
Authors: Run Luo, Renke Shan, Longze Chen, Ziqiang Liu, Lu Wang, Min Yang, Xiaobo Xia
Abstract: Large Vision-Language Models (LVLMs) are pivotal for real-world AI tasks like embodied intelligence due to their strong vision-language reasoning abilities. However, current LVLMs process entire images at the token level, which is inefficient compared to humans who analyze information and generate content at the conceptual level, extracting relevant visual concepts with minimal effort. This inefficiency, stemming from the lack of a visual concept model, limits LVLMs' usability in real-world applications. To address this, we propose VCM, an end-to-end self-supervised visual concept modeling framework. VCM leverages implicit contrastive learning across multiple sampled instances and vision-language fine-tuning to construct a visual concept model without requiring costly concept-level annotations. Our results show that VCM significantly reduces computational costs (e.g., 85\% fewer FLOPs for LLaVA-1.5-7B) while maintaining strong performance across diverse image understanding tasks. Moreover, VCM enhances visual encoders' capabilities in classic visual concept perception tasks. Extensive quantitative and qualitative experiments validate the effectiveness and efficiency of VCM.
Authors: Woongyeong Yeo, Kangsan Kim, Soyeong Jeong, Jinheon Baek, Sung Ju Hwang
Abstract: Retrieval-Augmented Generation (RAG) has shown substantial promise in improving factual accuracy by grounding model responses with external knowledge relevant to queries. However, most existing RAG approaches are limited to a text-only corpus, and while recent efforts have extended RAG to other modalities such as images and videos, they typically operate over a single modality-specific corpus. In contrast, real-world queries vary widely in the type of knowledge they require, which a single type of knowledge source cannot address. To address this, we introduce UniversalRAG, a novel RAG framework designed to retrieve and integrate knowledge from heterogeneous sources with diverse modalities and granularities. Specifically, motivated by the observation that forcing all modalities into a unified representation space derived from a single aggregated corpus causes a modality gap, where the retrieval tends to favor items from the same modality as the query, we propose a modality-aware routing mechanism that dynamically identifies the most appropriate modality-specific corpus and performs targeted retrieval within it. Also, beyond modality, we organize each modality into multiple granularity levels, enabling fine-tuned retrieval tailored to the complexity and scope of the query. We validate UniversalRAG on 8 benchmarks spanning multiple modalities, showing its superiority over various modality-specific and unified baselines.
Authors: Jiamin Chang, Haoyang Li, Hammond Pearce, Ruoxi Sun, Bo Li, Minhui Xue
Abstract: The growing adoption of artificial intelligence (AI) has amplified concerns about trustworthiness, including integrity, privacy, robustness, and bias. To assess and attribute these threats, we propose ConceptLens, a generic framework that leverages pre-trained multimodal models to identify the root causes of integrity threats by analyzing Concept Shift in probing samples. ConceptLens demonstrates strong detection performance for vanilla data poisoning attacks and uncovers vulnerabilities to bias injection, such as the generation of covert advertisements through malicious concept shifts. It identifies privacy risks in unaltered but high-risk samples, filters them before training, and provides insights into model weaknesses arising from incomplete or imbalanced training data. Additionally, at the model level, it attributes concepts that the target model is overly dependent on, identifies misleading concepts, and explains how disrupting key concepts negatively impacts the model. Furthermore, it uncovers sociological biases in generative content, revealing disparities across sociological contexts. Strikingly, ConceptLens reveals how safe training and inference data can be unintentionally and easily exploited, potentially undermining safety alignment. Our study informs actionable insights to breed trust in AI systems, thereby speeding adoption and driving greater innovation.
Authors: Yanhao 'Max' Wei, Zhenling Jiang
Abstract: We explore pretraining estimators for structural econometric models. The estimator is "pretrained" in the sense that the bulk of the computational cost and researcher effort occur during the construction of the estimator. Subsequent applications of the estimator to different datasets require little computational cost or researcher effort. The estimation leverages a neural net to recognize the structural model's parameter from data patterns. As an initial trial, this paper builds a pretrained estimator for a sequential search model that is known to be difficult to estimate. We evaluate the pretrained estimator on 12 real datasets. The estimation takes seconds to run and shows high accuracy. We provide the estimator at pnnehome.github.io. More generally, pretrained, off-the-shelf estimators can make structural models more accessible to researchers and practitioners.
Authors: Jing Liu, Yao Du, Kun Yang, Yan Wang, Xiping Hu, Zehua Wang, Yang Liu, Peng Sun, Azzedine Boukerche, Victor C. M. Leung
Abstract: Edge-cloud collaborative computing (ECCC) has emerged as a pivotal paradigm for addressing the computational demands of modern intelligent applications, integrating cloud resources with edge devices to enable efficient, low-latency processing. Recent advancements in AI, particularly deep learning and large language models (LLMs), have dramatically enhanced the capabilities of these distributed systems, yet introduce significant challenges in model deployment and resource management. In this survey, we comprehensive examine the intersection of distributed intelligence and model optimization within edge-cloud environments, providing a structured tutorial on fundamental architectures, enabling technologies, and emerging applications. Additionally, we systematically analyze model optimization approaches, including compression, adaptation, and neural architecture search, alongside AI-driven resource management strategies that balance performance, energy efficiency, and latency requirements. We further explore critical aspects of privacy protection and security enhancement within ECCC systems and examines practical deployments through diverse applications, spanning autonomous driving, healthcare, and industrial automation. Performance analysis and benchmarking techniques are also thoroughly explored to establish evaluation standards for these complex systems. Furthermore, the review identifies critical research directions including LLMs deployment, 6G integration, neuromorphic computing, and quantum computing, offering a roadmap for addressing persistent challenges in heterogeneity management, real-time processing, and scalability. By bridging theoretical advancements and practical deployments, this survey offers researchers and practitioners a holistic perspective on leveraging AI to optimize distributed computing environments, fostering innovation in next-generation intelligent systems.
Authors: Javier E. Santos, Agnese Marcato, Roman Colman, Nicholas Lubbers, Yen Ting Lin
Abstract: Generative diffusion models have achieved remarkable success in producing high-quality images. However, these models typically operate in continuous intensity spaces, diffusing independently across pixels and color channels. As a result, they are fundamentally ill-suited for applications involving inherently discrete quantities-such as particle counts or material units-that are constrained by strict conservation laws like mass conservation, limiting their applicability in scientific workflows. To address this limitation, we propose Discrete Spatial Diffusion (DSD), a framework based on a continuous-time, discrete-state jump stochastic process that operates directly in discrete spatial domains while strictly preserving particle counts in both forward and reverse diffusion processes. By using spatial diffusion to achieve particle conservation, we introduce stochasticity naturally through a discrete formulation. We demonstrate the expressive flexibility of DSD by performing image synthesis, class conditioning, and image inpainting across standard image benchmarks, while exactly conditioning total image intensity. We validate DSD on two challenging scientific applications: porous rock microstructures and lithium-ion battery electrodes, demonstrating its ability to generate structurally realistic samples under strict mass conservation constraints, with quantitative evaluation using state-of-the-art metrics for transport and electrochemical performance.
Authors: Xiusi Chen, Gaotang Li, Ziqi Wang, Bowen Jin, Cheng Qian, Yu Wang, Hongru Wang, Yu Zhang, Denghui Zhang, Tong Zhang, Hanghang Tong, Heng Ji
Abstract: Reward modeling is essential for aligning large language models with human preferences through reinforcement learning from human feedback. To provide accurate reward signals, a reward model (RM) should stimulate deep thinking and conduct interpretable reasoning before assigning a score or a judgment. Inspired by recent advances of long chain-of-thought on reasoning-intensive tasks, we hypothesize and validate that integrating reasoning capabilities into reward modeling significantly enhances RMs interpretability and performance. To this end, we introduce a new class of generative reward models - Reasoning Reward Models (ReasRMs) - which formulate reward modeling as a reasoning task. We propose a reasoning-oriented training pipeline and train a family of ReasRMs, RM-R1. RM-R1 features a chain-of-rubrics (CoR) mechanism - self-generating sample-level chat rubrics or math/code solutions, and evaluating candidate responses against them. The training of RM-R1 consists of two key stages: (1) distillation of high-quality reasoning chains and (2) reinforcement learning with verifiable rewards. Empirically, our models achieve state-of-the-art performance across three reward model benchmarks on average, outperforming much larger open-weight models (e.g., INF-ORM-Llama3.1-70B) and proprietary ones (e.g., GPT-4o) by up to 4.9%. Beyond final performance, we perform thorough empirical analyses to understand the key ingredients of successful ReasRM training. To facilitate future research, we release six REASRM models along with code and data at https://github.com/RM-R1-UIUC/RM-R1.
Authors: Srecharan Selvam
Abstract: Automating leaf manipulation in agricultural settings faces significant challenges, including the variability of plant morphologies and deformable leaves. We propose a novel hybrid geometric-neural approach for autonomous leaf grasping that combines traditional computer vision with neural networks through self-supervised learning. Our method integrates YOLOv8 for instance segmentation and RAFT-Stereo for 3D depth estimation to build rich leaf representations, which feed into both a geometric feature scoring pipeline and a neural refinement module (GraspPointCNN). The key innovation is our confidence-weighted fusion mechanism that dynamically balances the contribution of each approach based on prediction certainty. Our self-supervised framework uses the geometric pipeline as an expert teacher to automatically generate training data. Experiments demonstrate that our approach achieves an 88.0% success rate in controlled environments and 84.7% in real greenhouse conditions, significantly outperforming both purely geometric (75.3%) and neural (60.2%) methods. This work establishes a new paradigm for agricultural robotics where domain expertise is seamlessly integrated with machine learning capabilities, providing a foundation for fully automated crop monitoring systems.
Authors: Yuxin Zhang, Meihao Fan, Ju Fan, Mingyang Yi, Yuyu Luo, Jian Tan, Guoliang Li
Abstract: Recent advances in large language models (LLMs) have significantly improved performance on the Text-to-SQL task by leveraging their powerful reasoning capabilities. To enhance accuracy during the reasoning process, external Process Reward Models (PRMs) can be introduced during training and inference to provide fine-grained supervision. However, if misused, PRMs may distort the reasoning trajectory and lead to suboptimal or incorrect SQL generation. To address this challenge, we propose Reward-SQL, a framework that systematically explores how to incorporate PRMs into the Text-to-SQL reasoning process effectively. Our approach follows a "cold start, then PRM supervision" paradigm. Specifically, we first train the model to decompose SQL queries into structured stepwise reasoning chains using common table expressions (Chain-of-CTEs), establishing a strong and interpretable reasoning baseline. Then, we investigate four strategies for integrating PRMs, and find that combining PRM as an online training signal (e.g.,GRPO) with PRM-guided inference (e.g., best-of-N sampling) yields the best results. Empirically, on the BIRD benchmark, Reward-SQL enables models supervised by PRM (7B) to achieve a 13.1% performance gain across various guidance strategies. Notably, our GRPO-aligned policy model based on Qwen2.5-Coder-7B-Instruct achieves 68.9% accuracy on the BIRD development set, outperforming all baseline methods under the same model size. These results demonstrate the effectiveness of Reward-SQL in leveraging reward-based supervision for Text-to-SQL reasoning.
Authors: Zhongweiyang Xu, Xulin Fan, Zhong-Qiu Wang, Xilin Jiang, Romit Roy Choudhury
Abstract: Blind Speech Separation (BSS) aims to separate multiple speech sources from audio mixtures recorded by a microphone array. The problem is challenging because it is a blind inverse problem, i.e., the microphone array geometry, the room impulse response (RIR), and the speech sources, are all unknown. We propose ArrayDPS to solve the BSS problem in an unsupervised, array-agnostic, and generative manner. The core idea builds on diffusion posterior sampling (DPS), but unlike DPS where the likelihood is tractable, ArrayDPS must approximate the likelihood by formulating a separate optimization problem. The solution to the optimization approximates room acoustics and the relative transfer functions between microphones. These approximations, along with the diffusion priors, iterate through the ArrayDPS sampling process and ultimately yield separated voice sources. We only need a simple single-speaker speech diffusion model as a prior along with the mixtures recorded at the microphones; no microphone array information is necessary. Evaluation results show that ArrayDPS outperforms all baseline unsupervised methods while being comparable to supervised methods in terms of SDR. Audio demos are provided at: https://arraydps.github.io/ArrayDPSDemo/.
Authors: Botian Xu, Haoyang Weng, Qingzhou Lu, Yang Gao, Huazhe Xu
Abstract: Reinforcement learning (RL) has made significant strides in legged robot control, enabling locomotion across diverse terrains and complex loco-manipulation capabilities. However, the commonly used position or velocity tracking-based objectives are agnostic to forces experienced by the robot, leading to stiff and potentially dangerous behaviors and poor control during forceful interactions. To address this limitation, we present \emph{Force-Adaptive Control via Impedance Reference Tracking} (FACET). Inspired by impedance control, we use RL to train a control policy to imitate a virtual mass-spring-damper system, allowing fine-grained control under external forces by manipulating the virtual spring. In simulation, we demonstrate that our quadruped robot achieves improved robustness to large impulses (up to 200 Ns) and exhibits controllable compliance, achieving an 80% reduction in collision impulse. The policy is deployed to a physical robot to showcase both compliance and the ability to engage with large forces by kinesthetic control and pulling payloads up to 2/3 of its weight. Further extension to a legged loco-manipulator and a humanoid shows the applicability of our method to more complex settings to enable whole-body compliance control. Project Website: https://facet.pages.dev/
URLs: https://facet.pages.dev/
Authors: Muzhi Dai, Chenxu Yang, Qingyi Si
Abstract: As Test-Time Scaling emerges as an active research focus in the large language model community, advanced post-training methods increasingly emphasize extending chain-of-thought (CoT) generation length, thereby enhancing reasoning capabilities to approach Deepseek R1-like reasoning models. However, recent studies reveal that reasoning models (even Qwen3) consistently exhibit excessive thought redundancy in CoT generation. This overthinking issue arises from the inherent limitations of conventional outcome-reward reinforcement learning, which systematically overlooks the regulation of intermediate reasoning processes. This paper introduces Serial-Group Decaying-Reward Policy Optimization (S-GRPO), a novel reinforcement learning paradigm that enables models to implicitly evaluate the sufficiency of intermediate reasoning steps, thereby facilitating early exit in CoT generation. Unlike GRPO, which samples multiple possible reasoning paths in parallel (parallel group), S-GRPO only samples one reasoning path and serially selects multiple temporal positions from the path to exit thinking and directly generate answers (serial group). For correct answers within a serial group, rewards gradually decrease based on the exit positions along the reasoning path from front to back. This design encourages the model to produce more accurate and concise thoughts, while also incentivizing early thinking termination when appropriate. Empirical evaluations demonstrate that S-GRPO is compatible with state-of-the-art reasoning models, including Qwen3 and Deepseek-distill. Across diverse benchmarks such as GSM8K, AIME 2024, AMC 2023, MATH-500, and GPQA Diamond, S-GRPO achieves a substantial reduction in sequence length (35.4% - 61.1%) while simultaneously improving accuracy (absolute 0.72% - 6.08%).
Authors: Tobias Schnabel, Kiran Tomlinson, Adith Swaminathan, Jennifer Neville
Abstract: Despite their many successes, transformer-based large language models (LLMs) continue to struggle with tasks that require complex reasoning over large parts of their input. We argue that these failures arise due to capacity limits on the accurate flow of information within LLMs. To formalize this issue, we introduce the bounded attention prefix oracle (BAPO) model, a new computational framework that models bandwidth constraints on attention heads, the mechanism for internal communication in LLMs. We show that several important reasoning problems like graph reachability require high communication bandwidth for BAPOs to solve; we call these problems BAPO-hard. Our experiments corroborate our theoretical predictions: GPT-4o, Claude, and Gemini succeed on BAPO-easy tasks and fail even on relatively small BAPO-hard tasks. BAPOs also reveal another benefit of chain of thought (CoT): we prove that breaking down a task using CoT can turn any BAPO-hard problem into a BAPO-easy one. Our results offer principled explanations for key LLM failures and suggest directions for architectures and inference methods that mitigate bandwidth limits.
Authors: Ninh Pham, Rasmus Pagh
Abstract: Approximation of non-linear kernels using random feature maps has become a powerful technique for scaling kernel methods to large datasets. We propose $\textit{Tensor Sketch}$, an efficient random feature map for approximating polynomial kernels. Given $n$ training samples in $\mathbb{R}^d$ Tensor Sketch computes low-dimensional embeddings in $\mathbb{R}^D$ in time $\mathcal{O}\left( n(d+D \log{D}) \right)$ making it well-suited for high-dimensional and large-scale settings. We provide theoretical guarantees on the approximation error, ensuring the fidelity of the resulting kernel function estimates. We also discuss extensions and highlight applications where Tensor Sketch serves as a central computational tool.
Authors: Nicholas Attolino, Alessio Capitanelli, Fulvio Mastrogiovanni
Abstract: PDDL-based symbolic task planning remains pivotal for robot autonomy yet struggles with dynamic human-robot collaboration due to scalability, re-planning demands, and delayed plan availability. Although a few neurosymbolic frameworks have previously leveraged LLMs such as GPT-3 to address these challenges, reliance on closed-source, remote models with limited context introduced critical constraints: third-party dependency, inconsistent response times, restricted plan length and complexity, and multi-domain scalability issues. We present Gideon, a novel framework that enables the transition to modern, smaller, local LLMs with extended context length. Gideon integrates a novel problem generator to systematically generate large-scale datasets of realistic domain-problem-plan tuples for any domain, and adapts neurosymbolic planning for local LLMs, enabling on-device execution and extended context for multi-domain support. Preliminary experiments in single-domain scenarios performed on Qwen-2.5 1.5B and trained on 8k-32k samples, demonstrate a valid plan percentage of 66.1% (32k model) and show that the figure can be further scaled through additional data. Multi-domain tests on 16k samples yield an even higher 70.6% planning validity rate, proving extensibility across domains and signaling that data variety can have a positive effect on learning efficiency. Although long-horizon planning and reduced model size make Gideon training much less efficient than baseline models based on larger LLMs, the results are still significant considering that the trained model is about 120x smaller than baseline and that significant advantages can be achieved in inference efficiency, scalability, and multi-domain adaptability, all critical factors in human-robot collaboration. Training inefficiency can be mitigated by Gideon's streamlined data generation pipeline.
Authors: Adnan Ahmad, Bahareh Nakisa, Mohammad Naim Rastgoo
Abstract: Emotion recognition through physiological signals such as electroencephalogram (EEG) has become an essential aspect of affective computing and provides an objective way to capture human emotions. However, physiological data characterized by cross-subject variability and noisy labels hinder the performance of emotion recognition models. Existing domain adaptation and continual learning methods struggle to address these issues, especially under realistic conditions where data is continuously streamed and unlabeled. To overcome these limitations, we propose a novel bi-level self-supervised continual learning framework, SSOCL, based on a dynamic memory buffer. This bi-level architecture iteratively refines the dynamic buffer and pseudo-label assignments to effectively retain representative samples, enabling generalization from continuous, unlabeled physiological data streams for emotion recognition. The assigned pseudo-labels are subsequently leveraged for accurate emotion prediction. Key components of the framework, including a fast adaptation module and a cluster-mapping module, enable robust learning and effective handling of evolving data streams. Experimental validation on two mainstream EEG tasks demonstrates the framework's ability to adapt to continuous data streams while maintaining strong generalization across subjects, outperforming existing approaches.
Authors: Yacine Izza, Alexey Ignatiev, Sasha Rubin, Joao Marques-Silva, Peter J. Stuckey
Abstract: Explainable Artificial Intelligence (XAI) is critical for attaining trust in the operation of AI systems. A key question of an AI system is ``why was this decision made this way''. Formal approaches to XAI use a formal model of the AI system to identify abductive explanations. While abductive explanations may be applicable to a large number of inputs sharing the same concrete values, more general explanations may be preferred for numeric inputs. So-called inflated abductive explanations give intervals for each feature ensuring that any input whose values fall withing these intervals is still guaranteed to make the same prediction. Inflated explanations cover a larger portion of the input space, and hence are deemed more general explanations. But there can be many (inflated) abductive explanations for an instance. Which is the best? In this paper, we show how to find a most general abductive explanation for an AI decision. This explanation covers as much of the input space as possible, while still being a correct formal explanation of the model's behaviour. Given that we only want to give a human one explanation for a decision, the most general explanation gives us the explanation with the broadest applicability, and hence the one most likely to seem sensible. (The paper has been accepted at IJCAI2025 conference.)