Authors: Zhibiao Wang, Yunlong Zhou, Ziwei Zhang, Mengmei Zhang, Shirui Pan, Chunming Hu, Xiao Wang
Abstract: Graph Transformers, leveraging the global attention to capture long-range dependencies in graph structures, have significantly advanced graph machine learning, but face prohibitive computational complexity. Tokenized Graph Learning Models (TGLMs) address this issue by converting graphs into ordered token lists for scalable processing. Besides, TGLMs also empower Large Language Models (LLMs) to handle text-attributed graphs more effectively and thus are also employed in Graph LLMs. However, existing TGLMs rely on hand-designed token lists and their adaptability to diverse graph learning scenarios remains unexplored. In this paper, we first conduct extensive empirical and theoretical preliminary studies for hand-designed token lists. Surprisingly, we identify an unexplored hop-overpriority problem: the common pre-defined token lists overemphasize nearby nodes and overwhelm the ability of TGLMs to balance local and global signals. This phenomenon is especially harmful for heterophilic graphs. To address this problem, we propose the Learnable Graph Token List (LGTL), a plug-and-play module to replace hand-designed token lists in TGLMs. Specifically, LGTL adaptively adjusts the weights across hops and prioritizes informative nodes within hops through a graph attention gate module and a selection module, respectively. In this way, contextually informative nodes can be adaptively emphasized for both homophilic and heterophilic graphs. Besides, we theoretically show that LGTL can address the hop-overpriority problem. Extensive experiments on benchmarks validate the efficacy of LGTL across both Graph Transformers and Graph LLM backbones.
Authors: Valentin Villecroze, Yixin Wang, Gabriel Loaiza-Ganem
Abstract: The task of quantifying the inherent uncertainty associated with neural network predictions is a key challenge in artificial intelligence. Bayesian neural networks (BNNs) and deep ensembles are among the most prominent approaches to tackle this task. Both approaches produce predictions by computing an expectation of neural network outputs over some distribution on the corresponding weights; this distribution is given by the posterior in the case of BNNs, and by a mixture of point masses for ensembles. Inspired by recent work showing that the distribution used by ensembles can be understood as a posterior corresponding to a learned data-dependent prior, we propose last layer empirical Bayes (LLEB). LLEB instantiates a learnable prior as a normalizing flow, which is then trained to maximize the evidence lower bound; to retain tractability we use the flow only on the last layer. We show why LLEB is well motivated, and how it interpolates between standard BNNs and ensembles in terms of the strength of the prior that they use. LLEB performs on par with existing approaches, highlighting that empirical Bayes is a promising direction for future research in uncertainty quantification.
Authors: Alex Kogan
Abstract: Quantization became a necessary tool for serving ever-increasing Large Language Models (LLMs). RTN (Round-to-Nearest) is perhaps the simplest quantization technique that has been around well before LLMs surged to the forefront of machine learning (ML) research. Yet, it has been largely dismissed by recent and more advanced quantization methods that claim superiority over RTN in nearly every aspect of performance. This work aims to dispel this established point of view, showing that RTN is not only much cheaper to apply, but also its token generation throughput can be better than and accuracy can be similar to more advanced alternatives. In particular, we discuss our implementation of RTN based on the recent Marlin kernels and demonstrate how the accuracy of RTN can be gradually improved by selectively increasing the data precision format of certain model layers and modules. Based on our results, we argue that RTN presents a viable and practical choice for quantizing LLMs.
Authors: Morteza Alizadeh, Mehrdad Oveisi, Sonya Falahati, Ghazal Mousavi, Mohsen Alambardar Meybodi, Somayeh Sadat Mehrnia, Ilker Hacihaliloglu, Arman Rahmim, Mohammad R. Salmanpour
Abstract: Machine learning (ML) models rely heavily on consistent and accurate performance metrics to evaluate and compare their effectiveness. However, existing libraries often suffer from fragmentation, inconsistent implementations, and insufficient data validation protocols, leading to unreliable results. Existing libraries have often been developed independently and without adherence to a unified standard, particularly concerning the specific tasks they aim to support. As a result, each library tends to adopt its conventions for metric computation, input/output formatting, error handling, and data validation protocols. This lack of standardization leads to both implementation differences (ID) and reporting differences (RD), making it difficult to compare results across frameworks or ensure reliable evaluations. To address these issues, we introduce AllMetrics, an open-source unified Python library designed to standardize metric evaluation across diverse ML tasks, including regression, classification, clustering, segmentation, and image-to-image translation. The library implements class-specific reporting for multi-class tasks through configurable parameters to cover all use cases, while incorporating task-specific parameters to resolve metric computation discrepancies across implementations. Various datasets from domains like healthcare, finance, and real estate were applied to our library and compared with Python, Matlab, and R components to identify which yield similar results. AllMetrics combines a modular Application Programming Interface (API) with robust input validation mechanisms to ensure reproducibility and reliability in model evaluation. This paper presents the design principles, architectural components, and empirical analyses demonstrating the ability to mitigate evaluation errors and to enhance the trustworthiness of ML workflows.
Authors: Yuxiang Wei, Yanteng Zhang, Xi Xiao, Tianyang Wang, Xiao Wang, Vince D. Calhoun
Abstract: Decoding visual experiences from fMRI offers a powerful avenue to understand human perception and develop advanced brain-computer interfaces. However, current progress often prioritizes maximizing reconstruction fidelity while overlooking interpretability, an essential aspect for deriving neuroscientific insight. To address this gap, we propose MoRE-Brain, a neuro-inspired framework designed for high-fidelity, adaptable, and interpretable visual reconstruction. MoRE-Brain uniquely employs a hierarchical Mixture-of-Experts architecture where distinct experts process fMRI signals from functionally related voxel groups, mimicking specialized brain networks. The experts are first trained to encode fMRI into the frozen CLIP space. A finetuned diffusion model then synthesizes images, guided by expert outputs through a novel dual-stage routing mechanism that dynamically weighs expert contributions across the diffusion process. MoRE-Brain offers three main advancements: First, it introduces a novel Mixture-of-Experts architecture grounded in brain network principles for neuro-decoding. Second, it achieves efficient cross-subject generalization by sharing core expert networks while adapting only subject-specific routers. Third, it provides enhanced mechanistic insight, as the explicit routing reveals precisely how different modeled brain regions shape the semantic and spatial attributes of the reconstructed image. Extensive experiments validate MoRE-Brain's high reconstruction fidelity, with bottleneck analyses further demonstrating its effective utilization of fMRI signals, distinguishing genuine neural decoding from over-reliance on generative priors. Consequently, MoRE-Brain marks a substantial advance towards more generalizable and interpretable fMRI-based visual decoding. Code will be publicly available soon: https://github.com/yuxiangwei0808/MoRE-Brain.
Authors: Aaron Zweig, Zaikang Lin, Elham Azizi, David Knowles
Abstract: We study identifiability of stochastic differential equation (SDE) models under multiple interventions. Our results give the first provable bounds for unique recovery of SDE parameters given samples from their stationary distributions. We give tight bounds on the number of necessary interventions for linear SDEs, and upper bounds for nonlinear SDEs in the small noise regime. We experimentally validate the recovery of true parameters in synthetic data, and motivated by our theoretical results, demonstrate the advantage of parameterizations with learnable activation functions.
Authors: Aaron J. Li, Suraj Srinivas, Usha Bhalla, Himabindu Lakkaraju
Abstract: Sparse autoencoders (SAEs) are commonly used to interpret the internal activations of large language models (LLMs) by mapping them to human-interpretable concept representations. While existing evaluations of SAEs focus on metrics such as the reconstruction-sparsity tradeoff, human (auto-)interpretability, and feature disentanglement, they overlook a critical aspect: the robustness of concept representations to input perturbations. We argue that robustness must be a fundamental consideration for concept representations, reflecting the fidelity of concept labeling. To this end, we formulate robustness quantification as input-space optimization problems and develop a comprehensive evaluation framework featuring realistic scenarios in which adversarial perturbations are crafted to manipulate SAE representations. Empirically, we find that tiny adversarial input perturbations can effectively manipulate concept-based interpretations in most scenarios without notably affecting the outputs of the base LLMs themselves. Overall, our results suggest that SAE concept representations are fragile and may be ill-suited for applications in model monitoring and oversight.
Authors: Mariia Seleznova, Hung-Hsu Chou, Claudio Mayrink Verdun, Gitta Kutyniok
Abstract: We introduce GradPCA, an Out-of-Distribution (OOD) detection method that exploits the low-rank structure of neural network gradients induced by Neural Tangent Kernel (NTK) alignment. GradPCA applies Principal Component Analysis (PCA) to gradient class-means, achieving more consistent performance than existing methods across standard image classification benchmarks. We provide a theoretical perspective on spectral OOD detection in neural networks to support GradPCA, highlighting feature-space properties that enable effective detection and naturally emerge from NTK alignment. Our analysis further reveals that feature quality -- particularly the use of pretrained versus non-pretrained representations -- plays a crucial role in determining which detectors will succeed. Extensive experiments validate the strong performance of GradPCA, and our theoretical framework offers guidance for designing more principled spectral OOD detectors.
Authors: Weiguo Gao, Ming Li
Abstract: Diffusion trajectory distillation methods aim to accelerate sampling in diffusion models, which produce high-quality outputs but suffer from slow sampling speeds. These methods train a student model to approximate the multi-step denoising process of a pretrained teacher model in a single step, enabling one-shot generation. However, theoretical insights into the trade-off between different distillation strategies and generative quality remain limited, complicating their optimization and selection. In this work, we take a first step toward addressing this gap. Specifically, we reinterpret trajectory distillation as an operator merging problem in the linear regime, where each step of the teacher model is represented as a linear operator acting on noisy data. These operators admit a clear geometric interpretation as projections and rescalings corresponding to the noise schedule. During merging, signal shrinkage occurs as a convex combination of operators, arising from both discretization and limited optimization time of the student model. We propose a dynamic programming algorithm to compute the optimal merging strategy that maximally preserves signal fidelity. Additionally, we demonstrate the existence of a sharp phase transition in the optimal strategy, governed by data covariance structures. Our findings enhance the theoretical understanding of diffusion trajectory distillation and offer practical insights for improving distillation strategies.
Authors: Alejandro Garc\'ia-Castellanos, David R. Wessels, Nicky J. van den Berg, Remco Duits, Dani\"el M. Pelt, Erik J. Bekkers
Abstract: We introduce Equivariant Neural Eikonal Solvers, a novel framework that integrates Equivariant Neural Fields (ENFs) with Neural Eikonal Solvers. Our approach employs a single neural field where a unified shared backbone is conditioned on signal-specific latent variables - represented as point clouds in a Lie group - to model diverse Eikonal solutions. The ENF integration ensures equivariant mapping from these latent representations to the solution field, delivering three key benefits: enhanced representation efficiency through weight-sharing, robust geometric grounding, and solution steerability. This steerability allows transformations applied to the latent point cloud to induce predictable, geometrically meaningful modifications in the resulting Eikonal solution. By coupling these steerable representations with Physics-Informed Neural Networks (PINNs), our framework accurately models Eikonal travel-time solutions while generalizing to arbitrary Riemannian manifolds with regular group actions. This includes homogeneous spaces such as Euclidean, position-orientation, spherical, and hyperbolic manifolds. We validate our approach through applications in seismic travel-time modeling of 2D and 3D benchmark datasets. Experimental results demonstrate superior performance, scalability, adaptability, and user controllability compared to existing Neural Operator-based Eikonal solver methods.
Authors: Jan T\"onshoff, Martin Grohe
Abstract: Boolean Satisfiability (SAT) solvers are foundational to computer science, yet their performance typically hinges on hand-crafted heuristics. This work introduces Reinforcement Learning from Algorithm Feedback (RLAF) as a paradigm for learning to guide SAT solver branching heuristics with Graph Neural Networks (GNNs). Central to our approach is a novel and generic mechanism for injecting inferred variable weights and polarities into the branching heuristics of existing SAT solvers. In a single forward pass, a GNN assigns these parameters to all variables. Casting this one-shot guidance as a reinforcement learning problem lets us train the GNN with off-the-shelf policy-gradient methods, such as GRPO, directly using the solver's computational cost as the sole reward signal. Extensive evaluations demonstrate that RLAF-trained policies significantly reduce the mean solve times of different base solvers across diverse SAT problem distributions, achieving more than a 2x speedup in some cases, while generalizing effectively to larger and harder problems after training. Notably, these policies consistently outperform expert-supervised approaches based on learning handcrafted weighting heuristics, offering a promising path towards data-driven heuristic design in combinatorial optimization.
Authors: Jingcong Liang, Siyuan Wang, Miren Tian, Yitong Li, Duyu Tang, Zhongyu Wei
Abstract: Mixture-of-Experts (MoE) enables efficient scaling of large language models (LLMs) with sparsely activated experts during inference. To effectively deploy large MoE models on memory-constrained devices, many systems introduce *expert offloading* that caches a subset of experts in fast memory, leaving others on slow memory to run on CPU or load on demand. While some research has exploited the locality of expert activations, where consecutive tokens activate similar experts, the degree of this **local routing consistency** varies across models and remains understudied. In this paper, we propose two metrics to measure local routing consistency of MoE models: (1) **Segment Routing Best Performance (SRP)**, which evaluates how well a fixed group of experts can cover the needs of a segment of tokens, and (2) **Segment Cache Best Hit Rate (SCH)**, which measures the optimal segment-level cache hit rate under a given cache size limit. We analyzed 20 MoE LLMs with diverse sizes and architectures and found that models that apply MoE on every layer and do not use shared experts exhibit the highest local routing consistency. We further showed that domain-specialized experts contribute more to routing consistency than vocabulary-specialized ones, and that most models can balance between cache effectiveness and efficiency with cache sizes approximately 2x the active experts. These findings pave the way for memory-efficient MoE design and deployment without compromising inference speed. We publish the code for replicating experiments at https://github.com/ljcleo/moe-lrc .
Authors: Mars Liyao Gao, J. Nathan Kutz, Bernat Font
Abstract: Identifying the governing equations of a dynamical system is one of the most important tasks for scientific modeling. However, this procedure often requires high-quality spatio-temporal data uniformly sampled on structured grids. In this paper, we propose mesh-free SINDy, a novel algorithm which leverages the power of neural network approximation as well as auto-differentiation to identify governing equations from arbitrary sensor placements and non-uniform temporal data sampling. We show that mesh-free SINDy is robust to high noise levels and limited data while remaining computationally efficient. In our implementation, the training procedure is straight-forward and nearly free of hyperparameter tuning, making mesh-free SINDy widely applicable to many scientific and engineering problems. In the experiments, we demonstrate its effectiveness on a series of PDEs including the Burgers' equation, the heat equation, the Korteweg-De Vries equation and the 2D advection-diffusion equation. We conduct detailed numerical experiments on all datasets, varying the noise levels and number of samples, and we also compare our approach to previous state-of-the-art methods. It is noteworthy that, even in high-noise and low-data scenarios, mesh-free SINDy demonstrates robust PDE discovery, achieving successful identification with up to 75% noise for the Burgers' equation using 5,000 samples and with as few as 100 samples and 1% noise. All of this is achieved within a training time of under one minute.
Authors: Shangding Gu, Donghao Ying, Ming Jin, Yu Joe Lu, Jun Wang, Javad Lavaei, Costas Spanos
Abstract: We introduce Model Feedback Learning (MFL), a novel test-time optimization framework for optimizing inputs to pre-trained AI models or deployed hardware systems without requiring any retraining of the models or modifications to the hardware. In contrast to existing methods that rely on adjusting model parameters, MFL leverages a lightweight reverse model to iteratively search for optimal inputs, enabling efficient adaptation to new objectives under deployment constraints. This framework is particularly advantageous in real-world settings, such as semiconductor manufacturing recipe generation, where modifying deployed systems is often infeasible or cost-prohibitive. We validate MFL on semiconductor plasma etching tasks, where it achieves target recipe generation in just five iterations, significantly outperforming both Bayesian optimization and human experts. Beyond semiconductor applications, MFL also demonstrates strong performance in chemical processes (e.g., chemical vapor deposition) and electronic systems (e.g., wire bonding), highlighting its broad applicability. Additionally, MFL incorporates stability-aware optimization, enhancing robustness to process variations and surpassing conventional supervised learning and random search methods in high-dimensional control settings. By enabling few-shot adaptation, MFL provides a scalable and efficient paradigm for deploying intelligent control in real-world environments.
Authors: Zhixu Silvia Tao, Kasper Vinken, Hao-Wei Yeh, Avi Cooper, Xavier Boix
Abstract: Mixing datasets for fine-tuning large models (LMs) has become critical for maximizing performance on downstream tasks. However, composing effective dataset mixtures typically relies on heuristics and trial-and-error, often requiring multiple fine-tuning runs to achieve the desired outcome. We propose a novel method, $\textit{Merge to Mix}$, that accelerates composing dataset mixtures through model merging. Model merging is a recent technique that combines the abilities of multiple individually fine-tuned LMs into a single LM by using a few simple arithmetic operations. Our key insight is that merging models individually fine-tuned on each dataset in a mixture can effectively serve as a surrogate for a model fine-tuned on the entire mixture. Merge to Mix leverages this insight to accelerate selecting dataset mixtures without requiring full fine-tuning on each candidate mixture. Our experiments demonstrate that Merge to Mix surpasses state-of-the-art methods in dataset selection for fine-tuning LMs.
Authors: Bart Kosko, Olaoluwa Adigun
Abstract: We present the new bidirectional variational autoencoder (BVAE) network architecture. The BVAE uses a single neural network both to encode and decode instead of an encoder-decoder network pair. The network encodes in the forward direction and decodes in the backward direction through the same synaptic web. Simulations compared BVAEs and ordinary VAEs on the four image tasks of image reconstruction, classification, interpolation, and generation. The image datasets included MNIST handwritten digits, Fashion-MNIST, CIFAR-10, and CelebA-64 face images. The bidirectional structure of BVAEs cut the parameter count by almost 50% and still slightly outperformed the unidirectional VAEs.
Authors: Soham Gadgil, Chris Lin, Su-In Lee
Abstract: Sparse autoencoders (SAEs) are used to decompose neural network activations into human-interpretable features. Typically, features learned by a single SAE are used for downstream applications. However, it has recently been shown that SAEs trained with different initial weights can learn different features, demonstrating that a single SAE captures only a limited subset of features that can be extracted from the activation space. Motivated by this limitation, we propose to ensemble multiple SAEs through naive bagging and boosting. Specifically, SAEs trained with different weight initializations are ensembled in naive bagging, whereas SAEs sequentially trained to minimize the residual error are ensembled in boosting. We evaluate our ensemble approaches with three settings of language models and SAE architectures. Our empirical results demonstrate that ensembling SAEs can improve the reconstruction of language model activations, diversity of features, and SAE stability. Furthermore, ensembling SAEs performs better than applying a single SAE on downstream tasks such as concept detection and spurious correlation removal, showing improved practical utility.
Authors: Jiahuan Long, Wenzhe Zhang, Ning Wang, Tingsong Jiang, Wen Yao
Abstract: Physical field reconstruction (PFR) aims to predict the state distribution of physical quantities (e.g., velocity, pressure, and temperature) based on limited sensor measurements. It plays a critical role in domains such as fluid dynamics and thermodynamics. However, existing deep learning methods often fail to capture long-range temporal dependencies, resulting in suboptimal performance on time-evolving physical systems. To address this, we propose FR-Mamba, a novel spatiotemporal flow field reconstruction framework based on state space modeling. Specifically, we design a hybrid neural network architecture that combines Fourier Neural Operator (FNO) and State Space Model (SSM) to capture both global spatial features and long-range temporal dependencies. We adopt Mamba, a recently proposed efficient SSM architecture, to model long-range temporal dependencies with linear time complexity. In parallel, the FNO is employed to capture non-local spatial features by leveraging frequency-domain transformations. The spatiotemporal representations extracted by these two components are then fused to reconstruct the full-field distribution of the physical system. Extensive experiments demonstrate that our approach significantly outperforms existing PFR methods in flow field reconstruction tasks, achieving high-accuracy performance on long sequences.
Authors: Ziqing Wang, Kexin Zhang, Zihan Zhao, Yibo Wen, Abhishek Pandey, Han Liu, Kaize Ding
Abstract: Large language models (LLMs) are introducing a paradigm shift in molecular discovery by enabling text-guided interaction with chemical spaces through natural language, symbolic notations, with emerging extensions to incorporate multi-modal inputs. To advance the new field of LLM for molecular discovery, this survey provides an up-to-date and forward-looking review of the emerging use of LLMs for two central tasks: molecule generation and molecule optimization. Based on our proposed taxonomy for both problems, we analyze representative techniques in each category, highlighting how LLM capabilities are leveraged across different learning settings. In addition, we include the commonly used datasets and evaluation protocols. We conclude by discussing key challenges and future directions, positioning this survey as a resource for researchers working at the intersection of LLMs and molecular science. A continuously updated reading list is available at https://github.com/REAL-Lab-NU/Awesome-LLM-Centric-Molecular-Discovery.
URLs: https://github.com/REAL-Lab-NU/Awesome-LLM-Centric-Molecular-Discovery.
Authors: Ziyi (Queena), Zhou, Nicholas Stern, Julien Laasri
Abstract: Much research has been done to analyze the stock market. After all, if one can determine a pattern in the chaotic frenzy of transactions, then they could make a hefty profit from capitalizing on these insights. As such, the goal of our project was to apply reinforcement learning (RL) to determine the best time to buy a stock within a given time frame. With only a few adjustments, our model can be extended to identify the best time to sell a stock as well. In order to use the format of free, real-world data to train the model, we define our own Markov Decision Process (MDP) problem. These two papers [5] [6] helped us in formulating the state space and the reward system of our MDP problem. We train a series of agents using Q-Learning, Q-Learning with linear function approximation, and deep Q-Learning. In addition, we try to predict the stock prices using machine learning regression and classification models. We then compare our agents to see if they converge on a policy, and if so, which one learned the best policy to maximize profit on the stock market.
Authors: Monirul Islam Mahmud
Abstract: Keylogger detection involves monitoring for unusual system behaviors such as delays between typing and character display, analyzing network traffic patterns for data exfiltration. In this study, we provide a comprehensive analysis for keylogger detection with traditional machine learning models - SVC, Random Forest, Decision Tree, XGBoost, AdaBoost, Logistic Regression and Naive Bayes and advanced ensemble methods including Stacking, Blending and Voting. Moreover, feature selection approaches such as Information gain, Lasso L1 and Fisher Score are thoroughly assessed to improve predictive performance and lower computational complexity. The Keylogger Detection dataset from publicly available Kaggle website is used in this project. In addition to accuracy-based classification, this study implements the approach for model interpretation using Explainable AI (XAI) techniques namely SHAP (Global) and LIME (Local) to deliver finer explanations for how much each feature contributes in assisting or hindering the detection process. To evaluate the models result, we have used AUC score, sensitivity, Specificity, Accuracy and F1 score. The best performance was achieved by AdaBoost with 99.76% accuracy, F1 score of 0.99, 100% precision, 98.6% recall, 1.0 specificity and 0.99 of AUC that is near-perfect classification with Fisher Score.
Authors: Panagiotis Lymperopoulos, Vasanth Sarathy
Abstract: Modern Large Language Models (LLMs) often require external tools, such as machine learning classifiers or knowledge retrieval systems, to provide accurate answers in domains where their pre-trained knowledge is insufficient. This integration of LLMs with external tools expands their utility but also introduces a critical challenge: determining the trustworthiness of responses generated by the combined system. In high-stakes applications, such as medical decision-making, it is essential to assess the uncertainty of both the LLM's generated text and the tool's output to ensure the reliability of the final response. However, existing uncertainty quantification methods do not account for the tool-calling scenario, where both the LLM and external tool contribute to the overall system's uncertainty. In this work, we present a novel framework for modeling tool-calling LLMs that quantifies uncertainty by jointly considering the predictive uncertainty of the LLM and the external tool. We extend previous methods for uncertainty quantification over token sequences to this setting and propose efficient approximations that make uncertainty computation practical for real-world applications. We evaluate our framework on two new synthetic QA datasets, derived from well-known machine learning datasets, which require tool-calling for accurate answers. Additionally, we apply our method to retrieval-augmented generation (RAG) systems and conduct a proof-of-concept experiment demonstrating the effectiveness of our uncertainty metrics in scenarios where external information retrieval is needed. Our results show that the framework is effective in enhancing trust in LLM-based systems, especially in cases where the LLM's internal knowledge is insufficient and external tools are required.
Authors: Aditya T. Vadlamani, Anutam Srinivasan, Pranav Maneriker, Ali Payani, Srinivasan Parthasarathy
Abstract: Conformal Prediction (CP) is a popular method for uncertainty quantification with machine learning models. While conformal prediction provides probabilistic guarantees regarding the coverage of the true label, these guarantees are agnostic to the presence of sensitive attributes within the dataset. In this work, we formalize \textit{Conformal Fairness}, a notion of fairness using conformal predictors, and provide a theoretically well-founded algorithm and associated framework to control for the gaps in coverage between different sensitive groups. Our framework leverages the exchangeability assumption (implicit to CP) rather than the typical IID assumption, allowing us to apply the notion of Conformal Fairness to data types and tasks that are not IID, such as graph data. Experiments were conducted on graph and tabular datasets to demonstrate that the algorithm can control fairness-related gaps in addition to coverage aligned with theoretical expectations.
Authors: Junhong Lin, Xinyue Zeng, Jie Zhu, Song Wang, Julian Shun, Jun Wu, Dawei Zhou
Abstract: Large Language Models (LLMs) have achieved remarkable success in complex reasoning tasks, but their inference remains computationally inefficient. We observe a common failure mode in many prevalent LLMs, overthinking, where models generate verbose and tangential reasoning traces even for simple queries. Recent works have tried to mitigate this by enforcing fixed token budgets, however, this can lead to underthinking, especially on harder problems. Through empirical analysis, we identify that this inefficiency often stems from unclear problem-solving strategies. To formalize this, we develop a theoretical model, BBAM (Bayesian Budget Allocation Model), which models reasoning as a sequence of sub-questions with varying uncertainty, and introduce the $E^3$ metric to capture the trade-off between correctness and computation efficiency. Building on theoretical results from BBAM, we propose Plan-and-Budget, a model-agnostic, test-time framework that decomposes complex queries into sub-questions and allocates token budgets based on estimated complexity using adaptive scheduling. Plan-and-Budget improves reasoning efficiency across a range of tasks and models, achieving up to +70% accuracy gains, -39% token reduction, and +187.5% improvement in $E^3$. Notably, it elevates a smaller model (DS-Qwen-32B) to match the efficiency of a larger model (DS-LLaMA-70B)-demonstrating Plan-and-Budget's ability to close performance gaps without retraining. Our code is available at anonymous.4open.science/r/P-and-B-6513/.
Authors: Kotaro Yoshida, Slavakis Konstantinos
Abstract: Invariant risk minimization (IRM) aims to enable out-of-distribution (OOD) generalization in deep learning by learning invariant representations. As IRM poses an inherently challenging bi-level optimization problem, most existing approaches -- including IRMv1 -- adopt penalty-based single-level approximations. However, empirical studies consistently show that these methods often fail to outperform well-tuned empirical risk minimization (ERM), highlighting the need for more robust IRM implementations. This work theoretically identifies a key limitation common to many IRM variants: their penalty terms are highly sensitive to limited environment diversity and over-parameterization, resulting in performance degradation. To address this issue, a novel extrapolation-based framework is proposed that enhances environmental diversity by augmenting the IRM penalty through synthetic distributional shifts. Extensive experiments -- ranging from synthetic setups to realistic, over-parameterized scenarios -- demonstrate that the proposed method consistently outperforms state-of-the-art IRM variants, validating its effectiveness and robustness.
Authors: Zehong Wang, Zheyuan Zhang, Tianyi Ma, Chuxu Zhang, Yanfang Ye
Abstract: Graph neural networks (GNNs) has been predominantly driven by message-passing, where node representations are iteratively updated via local neighborhood aggregation. Despite their success, message-passing suffers from fundamental limitations -- including constrained expressiveness, over-smoothing, over-squashing, and limited capacity to model long-range dependencies. These issues hinder scalability: increasing data size or model size often fails to yield improved performance, limiting the viability of GNNs as backbones for graph foundation models. In this work, we explore pathways beyond message-passing and introduce Generative Graph Pattern Machine (G$^2$PM), a generative Transformer pre-training framework for graphs. G$^2$PM represents graph instances (nodes, edges, or entire graphs) as sequences of substructures, and employs generative pre-training over the sequences to learn generalizable, transferable representations. Empirically, G$^2$PM demonstrates strong scalability: on the ogbn-arxiv benchmark, it continues to improve with model sizes up to 60M parameters, outperforming prior generative approaches that plateau at significantly smaller scales (e.g., 3M). In addition, we systematically analyze the model design space, highlighting key architectural choices that contribute to its scalability and generalization. Across diverse tasks -- including node classification, graph classification, and transfer learning -- G$^2$PM consistently outperforms strong baselines, establishing a compelling foundation for scalable graph learning. The code and dataset are available at https://github.com/Zehong-Wang/G2PM.
Authors: Heqiang Wang, Xiang Liu, Xiaoxiong Zhong, Lixing Chen, Fangming Liu, Weizhe Zhang
Abstract: The Internet of Things (IoT) ecosystem generates vast amounts of multimodal data from heterogeneous sources such as sensors, cameras, and microphones. As edge intelligence continues to evolve, IoT devices have progressed from simple data collection units to nodes capable of executing complex computational tasks. This evolution necessitates the adoption of distributed learning strategies to effectively handle multimodal data in an IoT environment. Furthermore, the real-time nature of data collection and limited local storage on edge devices in IoT call for an online learning paradigm. To address these challenges, we introduce the concept of Multimodal Online Federated Learning (MMO-FL), a novel framework designed for dynamic and decentralized multimodal learning in IoT environments. Building on this framework, we further account for the inherent instability of edge devices, which frequently results in missing modalities during the learning process. We conduct a comprehensive theoretical analysis under both complete and missing modality scenarios, providing insights into the performance degradation caused by missing modalities. To mitigate the impact of modality missing, we propose the Prototypical Modality Mitigation (PMM) algorithm, which leverages prototype learning to effectively compensate for missing modalities. Experimental results on two multimodal datasets further demonstrate the superior performance of PMM compared to benchmarks.
Authors: Chongjie Si, Kangtao Lv, Jingjing Jiang, Yadao Wang, Yongwei Wang, Xiaokang Yang, Wenbo Su, Bo Zheng, Wei Shen
Abstract: Model merging offers a training-free alternative to multi-task learning by combining independently fine-tuned models into a unified one without access to raw data. However, existing approaches often rely on heuristics to determine the merging coefficients, limiting their scalability and generality. In this work, we revisit model merging through the lens of least-squares optimization and show that the optimal merging weights should scale with the amount of task-specific information encoded in each model. Based on this insight, we propose NAN, a simple yet effective method that estimates model merging coefficients via the inverse of parameter norm. NAN is training-free, plug-and-play, and applicable to a wide range of merging strategies. Extensive experiments on show that NAN consistently improves performance of baseline methods.
Authors: Chongjie Si, Yidan Cui, Fuchao Yang, Xiaokang Yang, Wei Shen
Abstract: Learning from inaccurate annotations has gained significant attention due to the high cost of precise labeling. However, despite the presence of erroneous labels, models trained on noisy data often retain the ability to make accurate predictions. This intriguing phenomenon raises a fundamental yet largely unexplored question: why models can still extract correct label information from inaccurate annotations remains unexplored. In this paper, we conduct a comprehensive investigation into this issue. By analyzing weight matrices from both empirical and theoretical perspectives, we find that label inaccuracy primarily accumulates noise in lower singular components and subtly perturbs the principal subspace. Within a certain range, the principal subspaces of weights trained on inaccurate labels remain largely aligned with those learned from clean labels, preserving essential task-relevant information. We formally prove that the angles of principal subspaces exhibit minimal deviation under moderate label inaccuracy, explaining why models can still generalize effectively. Building on these insights, we propose LIP, a lightweight plug-in designed to help classifiers retain principal subspace information while mitigating noise induced by label inaccuracy. Extensive experiments on tasks with various inaccuracy conditions demonstrate that LIP consistently enhances the performance of existing algorithms. We hope our findings can offer valuable theoretical and practical insights to understand of model robustness under inaccurate supervision.
Authors: Navid Seidi, Satyaki Roy, Sajal Das
Abstract: Federated Learning (FL) holds great promise for digital health by enabling collaborative model training without compromising patient data privacy. However, heterogeneity across institutions, lack of sustained reputation, and unreliable contributions remain major challenges. In this paper, we propose a robust, peer-driven reputation mechanism for federated healthcare that employs a hybrid communication model to integrate decentralized peer feedback with clustering-based noise handling to enhance model aggregation. Crucially, our approach decouples the federated aggregation and reputation mechanisms by applying differential privacy to client-side model updates before sharing them for peer evaluation. This ensures sensitive information remains protected during reputation computation, while unaltered updates are sent to the server for global model training. Using the Cox Proportional Hazards model for survival analysis across multiple federated nodes, our framework addresses both data heterogeneity and reputation deficit by dynamically adjusting trust scores based on local performance improvements measured via the concordance index. Experimental evaluations on both synthetic datasets and the SEER dataset demonstrate that our method consistently achieves high and stable C-index values, effectively down-weighing noisy client updates and outperforming FL methods that lack a reputation system.
Authors: Ichiro Hashimoto
Abstract: In this paper, we prove directional convergence of network parameters of fixed width leaky ReLU two-layer neural networks optimized by gradient descent with exponential loss, which was previously only known for gradient flow. By a careful analysis of the convergent direction, we establish sufficient conditions of benign overfitting and discover a new phase transition in the test error bound. All of these results hold beyond the nearly orthogonal data setting which was studied in prior works. As an application, we demonstrate that benign overfitting occurs with high probability in sub-Gaussian mixture models.
Authors: Zhihang Cai, Xingjun Zhang, Zhendong Tan, Zheng Wei
Abstract: Large Language Models (LLMs) have demonstrated remarkable proficiency across a wide range of tasks. However, LLMs often require larger batch sizes to enhance throughput or longer context lengths to meet task demands, which significantly increases the memory resource consumption of the Key-Value (KV) cache during inference, becoming a major bottleneck in LLM deployment. To address this issue, quantization is a common and straightforward approach. Currently, quantization methods for activations are limited to 8-bit, and quantization to even lower bits can lead to substantial accuracy drops. To further save space by quantizing the KV cache to even lower bits, we analyzed the element distribution of the KV cache and designed the NQKV algorithm. Since the elements within each block of the KV cache follow a normal distribution, NQKV employs per-block quantile quantization to achieve information-theoretically optimal quantization error. Without significantly compromising model output quality, NQKV enables the OPT model to perform inference with an 2x larger batch size or a 4x longer context length, and it improves throughput by 9.3x compared to when the KV cache is not used.
Authors: Hon Tik Tse, Siddarth Chandrasekar, Marlos C. Machado
Abstract: In recent years, the successor representation (SR) has attracted increasing attention in reinforcement learning (RL), and it has been used to address some of its key challenges, such as exploration, credit assignment, and generalization. The SR can be seen as representing the underlying credit assignment structure of the environment by implicitly encoding its induced transition dynamics. However, the SR is reward-agnostic. In this paper, we discuss a similar representation that also takes into account the reward dynamics of the problem. We study the default representation (DR), a recently proposed representation with limited theoretical (and empirical) analysis. Here, we lay some of the theoretical foundation underlying the DR in the tabular case by (1) deriving dynamic programming and (2) temporal-difference methods to learn the DR, (3) characterizing the basis for the vector space of the DR, and (4) formally extending the DR to the function approximation case through default features. Empirically, we analyze the benefits of the DR in many of the settings in which the SR has been applied, including (1) reward shaping, (2) option discovery, (3) exploration, and (4) transfer learning. Our results show that, compared to the SR, the DR gives rise to qualitatively different, reward-aware behaviour and quantitatively better performance in several settings.
Authors: Zi-Jian Cheng, Zi-Yi Jia, Zhi Zhou, Yu-Feng Li, Lan-Zhe Guo
Abstract: Tabular data, owing to its ubiquitous presence in real-world domains, has garnered significant attention in machine learning research. While tree-based models have long dominated tabular machine learning tasks, the recently proposed deep learning model TabPFN v2 has emerged, demonstrating unparalleled performance and scalability potential. Although extensive research has been conducted on TabPFN v2 to further improve performance, the majority of this research remains confined to closed environments, neglecting the challenges that frequently arise in open environments. This raises the question: Can TabPFN v2 maintain good performance in open environments? To this end, we conduct the first comprehensive evaluation of TabPFN v2's adaptability in open environments. We construct a unified evaluation framework covering various real-world challenges and assess the robustness of TabPFN v2 under open environments scenarios using this framework. Empirical results demonstrate that TabPFN v2 shows significant limitations in open environments but is suitable for small-scale, covariate-shifted, and class-balanced tasks. Tree-based models remain the optimal choice for general tabular tasks in open environments. To facilitate future research on open environments challenges, we advocate for open environments tabular benchmarks, multi-metric evaluation, and universal modules to strengthen model robustness. We publicly release our evaluation framework at https://anonymous.4open.science/r/tabpfn-ood-4E65.
Authors: Runze Yan, Xun Shen, Akifumi Wachi, Sebastien Gros, Anni Zhao, Xiao Hu
Abstract: When applying offline reinforcement learning (RL) in healthcare scenarios, the out-of-distribution (OOD) issues pose significant risks, as inappropriate generalization beyond clinical expertise can result in potentially harmful recommendations. While existing methods like conservative Q-learning (CQL) attempt to address the OOD issue, their effectiveness is limited by only constraining action selection by suppressing uncertain actions. This action-only regularization imitates clinician actions that prioritize short-term rewards, but it fails to regulate downstream state trajectories, thereby limiting the discovery of improved long-term treatment strategies. To safely improve policy beyond clinician recommendations while ensuring that state-action trajectories remain in-distribution, we propose \textit{Offline Guarded Safe Reinforcement Learning} ($\mathsf{OGSRL}$), a theoretically grounded model-based offline RL framework. $\mathsf{OGSRL}$ introduces a novel dual constraint mechanism for improving policy with reliability and safety. First, the OOD guardian is established to specify clinically validated regions for safe policy exploration. By constraining optimization within these regions, it enables the reliable exploration of treatment strategies that outperform clinician behavior by leveraging the full patient state history, without drifting into unsupported state-action trajectories. Second, we introduce a safety cost constraint that encodes medical knowledge about physiological safety boundaries, providing domain-specific safeguards even in areas where training data might contain potentially unsafe interventions. Notably, we provide theoretical guarantees on safety and near-optimality: policies that satisfy these constraints remain in safe and reliable regions and achieve performance close to the best possible policy supported by the data.
Authors: Wenxuan Zhu, Qiyuan Wu, Tengda Tang, Renzi Meng, Sheng Chai, Xuehui Quan
Abstract: This paper addresses the limitations of multi-node perception and delayed scheduling response in distributed systems by proposing a GNN-based multi-node collaborative perception mechanism. The system is modeled as a graph structure. Message-passing and state-update modules are introduced. A multi-layer graph neural network is constructed to enable efficient information aggregation and dynamic state inference among nodes. In addition, a perception representation method is designed by fusing local states with global features. This improves each node's ability to perceive the overall system status. The proposed method is evaluated within a customized experimental framework. A dataset featuring heterogeneous task loads and dynamic communication topologies is used. Performance is measured in terms of task completion rate, average latency, load balancing, and transmission efficiency. Experimental results show that the proposed method outperforms mainstream algorithms under various conditions, including limited bandwidth and dynamic structural changes. It demonstrates superior perception capabilities and cooperative scheduling performance. The model achieves rapid convergence and efficient responses to complex system states.
Authors: Alaa Khaddaj, Logan Engstrom, Aleksander Madry
Abstract: Choice of training data distribution greatly influences model behavior. Yet, in large-scale settings, precisely characterizing how changes in training data affects predictions is often difficult due to model training costs. Current practice is to instead extrapolate from scaled down, inexpensive-to-train proxy models. However, changes in data do not influence smaller and larger models identically. Therefore, understanding how choice of data affects large-scale models raises the question: how does training data distribution influence model behavior across compute scale? We find that small- and large-scale language model predictions (generally) do highly correlate across choice of training data. Equipped with these findings, we characterize how proxy scale affects effectiveness in two downstream proxy model applications: data attribution and dataset selection.
Authors: Ilgee Hong, Changlong Yu, Liang Qiu, Weixiang Yan, Zhenghao Xu, Haoming Jiang, Qingru Zhang, Qin Lu, Xin Liu, Chao Zhang, Tuo Zhao
Abstract: Reinforcement learning from human feedback (RLHF) has become a powerful post-training paradigm for aligning large language models with human preferences. A core challenge in RLHF is constructing accurate reward signals, where the conventional Bradley-Terry reward models (BT RMs) often suffer from sensitivity to data size and coverage, as well as vulnerability to reward hacking. Generative reward models (GenRMs) offer a more robust alternative by generating chain-of-thought (CoT) rationales followed by a final reward. However, existing GenRMs rely on shallow, vertically scaled reasoning, limiting their capacity to handle nuanced or complex (e.g., reasoning-intensive) tasks. Moreover, their pairwise preference outputs are incompatible with standard RLHF algorithms that require pointwise reward signals. In this work, we introduce Think-RM, a training framework that enables long-horizon reasoning in GenRMs by modeling an internal thinking process. Rather than producing structured, externally provided rationales, Think-RM generates flexible, self-guided reasoning traces that support advanced capabilities such as self-reflection, hypothetical reasoning, and divergent reasoning. To elicit these reasoning abilities, we first warm-up the models by supervised fine-tuning (SFT) over long CoT data. We then further improve the model's long-horizon abilities by rule-based reinforcement learning (RL). In addition, we propose a novel pairwise RLHF pipeline that directly optimizes policies using pairwise preference rewards, eliminating the need for pointwise reward conversion and enabling more effective use of Think-RM outputs. Experiments show that Think-RM achieves state-of-the-art results on RM-Bench, outperforming both BT RM and vertically scaled GenRM by 8%. When combined with our pairwise RLHF pipeline, it demonstrates superior end-policy performance compared to traditional approaches.
Authors: Josh Alman, Zhao Song
Abstract: Attention mechanisms lie at the heart of modern large language models (LLMs). Straightforward algorithms for forward and backward (gradient) computation take quadratic time, and a line of work initiated by [Alman and Song NeurIPS 2023] and [Alman and Song NeurIPS 2024] has shown that quadratic time is necessary unless the model weights are small, in which case almost linear time algorithms are possible. In this paper, we show that large weights are necessary to avoid a strong preclusion to representational strength we call layer collapse, which means that the entire network can be approximated well by a network with only a single layer. Thus, the quadratic running time of attention is unavoidable for expressive transformers. The notion of layer collapse that we introduce is a variant on the notion of rank collapse from the work of [Dong, Cordonnier, and Loukas ICML 2021]. They showed that in Self Attention Networks with small weights and with skip connections, rank collapse must occur. This is typically interpreted as justifying the necessity of skip connections in expressive networks. However, our result shows that even with skip connections, if the weights are small, then layer collapse still occurs. Thus, only large weights, and not skip connections, can prevent these representational weaknesses.
Authors: Ronen Gradwohl, Eilam Shapira, Moshe Tennenholtz
Abstract: Algorithmic fairness has emerged as a central issue in ML, and it has become standard practice to adjust ML algorithms so that they will satisfy fairness requirements such as Equal Opportunity. In this paper we consider the effects of adopting such fair classifiers on the overall level of ecosystem fairness. Specifically, we introduce the study of fairness with competing firms, and demonstrate the failure of fair classifiers in yielding fair ecosystems. Our results quantify the loss of fairness in systems, under a variety of conditions, based on classifiers' correlation and the level of their data overlap. We show that even if competing classifiers are individually fair, the ecosystem's outcome may be unfair; and that adjusting biased algorithms to improve their individual fairness may lead to an overall decline in ecosystem fairness. In addition to these theoretical results, we also provide supporting experimental evidence. Together, our model and results provide a novel and essential call for action.
Authors: Bingyang Cheng, Zhongtao Chen, Yichen Jin, Hao Zhang, Chen Zhang, Edmud Y. Lam, Yik-Chung Wu
Abstract: Tensor CANDECOMP/PARAFAC decomposition (CPD) is a fundamental model for tensor reconstruction. Although the Bayesian framework allows for principled uncertainty quantification and automatic hyperparameter learning, existing methods do not scale well for large tensors because of high-dimensional matrix inversions. To this end, we introduce CP-GAMP, a scalable Bayesian CPD algorithm. This algorithm leverages generalized approximate message passing (GAMP) to avoid matrix inversions and incorporates an expectation-maximization routine to jointly infer the tensor rank and noise power. Through multiple experiments, for synthetic 100x100x100 rank 20 tensors with only 20% elements observed, the proposed algorithm reduces runtime by 82.7% compared to the state-of-the-art variational Bayesian CPD method, while maintaining comparable reconstruction accuracy.
Authors: Xingyu Zhang, Wenwen Qiang, Siyu Zhao, Huijie Guo, Jiangmeng Li, Chuxiong Sun, Changwen Zheng
Abstract: Most existing multivariate time series forecasting methods adopt an all-to-all paradigm that feeds all variable histories into a unified model to predict their future values without distinguishing their individual roles. However, this undifferentiated paradigm makes it difficult to identify variable-specific causal influences and often entangles causally relevant information with spurious correlations. To address this limitation, we propose an all-to-one forecasting paradigm that predicts each target variable separately. Specifically, we first construct a Structural Causal Model from observational data and then, for each target variable, we partition the historical sequence into four sub-segments according to the inferred causal structure: endogenous, direct causal, collider causal, and spurious correlation. The prediction relies solely on the first three causally relevant sub-segments, while the spurious correlation sub-segment is excluded. Furthermore, we propose Causal Informed Transformer (CAIFormer), a novel forecasting model comprising three components: Endogenous Sub-segment Prediction Block, Direct Causal Sub-segment Prediction Block, and Collider Causal Sub-segment Prediction Block, which process the endogenous, direct causal, and collider causal sub-segments, respectively. Their outputs are then combined to produce the final prediction. Extensive experiments on multiple benchmark datasets demonstrate the effectiveness of the CAIFormer.
Authors: Yangyang Wang, Jiawei Gu, Li Long, Xin Li, Li Shen, Zhouyu Fu, Xiangjun Zhou, Xu Jiang
Abstract: Accurate demand estimation is critical for the retail business in guiding the inventory and pricing policies of perishable products. However, it faces fundamental challenges from censored sales data during stockouts, where unobserved demand creates systemic policy biases. Existing datasets lack the temporal resolution and annotations needed to address this censoring effect. To fill this gap, we present FreshRetailNet-50K, the first large-scale benchmark for censored demand estimation. It comprises 50,000 store-product time series of detailed hourly sales data from 898 stores in 18 major cities, encompassing 863 perishable SKUs meticulously annotated for stockout events. The hourly stock status records unique to this dataset, combined with rich contextual covariates, including promotional discounts, precipitation, and temporal features, enable innovative research beyond existing solutions. We demonstrate one such use case of two-stage demand modeling: first, we reconstruct the latent demand during stockouts using precise hourly annotations. We then leverage the recovered demand to train robust demand forecasting models in the second stage. Experimental results show that this approach achieves a 2.73\% improvement in prediction accuracy while reducing the systematic demand underestimation from 7.37\% to near-zero bias. With unprecedented temporal granularity and comprehensive real-world information, FreshRetailNet-50K opens new research directions in demand imputation, perishable inventory optimization, and causal retail analytics. The unique annotation quality and scale of the dataset address long-standing limitations in retail AI, providing immediate solutions and a platform for future methodological innovation. The data (https://huggingface.co/datasets/Dingdong-Inc/FreshRetailNet-50K) and code (https://github.com/Dingdong-Inc/frn-50k-baseline}) are openly released.
URLs: https://huggingface.co/datasets/Dingdong-Inc/FreshRetailNet-50K), https://github.com/Dingdong-Inc/frn-50k-baseline
Authors: Woosung Koh, Wonbeen Oh, Jaein Jang, MinHyung Lee, Hyeongjin Kim, Ah Yeon Kim, Joonkee Kim, Junghyun Lee, Taehyeon Kim, Se-Young Yun
Abstract: Self-Taught Reasoners (STaR), synonymously known as Rejection sampling Fine-Tuning (RFT), is an integral part of the training pipeline of self-improving reasoning Language Models (LMs). The self-improving mechanism often employs random observation (data) sampling. However, this results in trained observation imbalance; inefficiently over-training on solved examples while under-training on challenging ones. In response, we introduce Adaptive STaR (AdaSTaR), a novel algorithm that rectifies this by integrating two adaptive sampling principles: (1) Adaptive Sampling for Diversity: promoting balanced training across observations, and (2) Adaptive Sampling for Curriculum: dynamically adjusting data difficulty to match the model's evolving strength. Across six benchmarks, AdaSTaR achieves best test accuracy in all instances (6/6) and reduces training FLOPs by an average of 58.6% against an extensive list of baselines. These improvements in performance and efficiency generalize to different pre-trained LMs and larger models, paving the way for more efficient and effective self-improving LMs.
Authors: Qian Tan, Dongzhan Zhou, Peng Xia, Wanhao Liu, Wanli Ouyang, Lei Bai, Yuqiang Li, Tianfan Fu
Abstract: Multimodal large language models (MLLMs) have made impressive progress in many applications in recent years. However, chemical MLLMs that can handle cross-modal understanding and generation remain underexplored. To fill this gap, in this paper, we propose ChemMLLM, a unified chemical multimodal large language model for molecule understanding and generation. Also, we design five multimodal tasks across text, molecular SMILES strings, and image, and curate the datasets. We benchmark ChemMLLM against a range of general leading MLLMs and Chemical LLMs on these tasks. Experimental results show that ChemMLLM achieves superior performance across all evaluated tasks. For example, in molecule image optimization task, ChemMLLM outperforms the best baseline (GPT-4o) by 118.9\% (4.27 vs 1.95 property improvement). The code is publicly available at https://github.com/bbsbz/ChemMLLM.git.
Authors: Chaerin Kong, Jiho Jang, Nojun Kwak
Abstract: Differential Transformer has recently gained significant attention for its impressive empirical performance, often attributed to its ability to perform noise canceled attention. However, precisely how differential attention achieves its empirical benefits remains poorly understood. Moreover, Differential Transformer architecture demands large-scale training from scratch, hindering utilization of open pretrained weights. In this work, we conduct an in-depth investigation of Differential Transformer, uncovering three key factors behind its success: (1) enhanced expressivity via negative attention, (2) reduced redundancy among attention heads, and (3) improved learning dynamics. Based on these findings, we propose DEX, a novel method to efficiently integrate the advantages of differential attention into pretrained language models. By reusing the softmax attention scores and adding a lightweight differential operation on the output value matrix, DEX effectively incorporates the key advantages of differential attention while remaining lightweight in both training and inference. Evaluations confirm that DEX substantially improves the pretrained LLMs across diverse benchmarks, achieving significant performance gains with minimal adaptation data (< 0.01\%).
Authors: Yunhui Jang, Jaehyung Kim, Sungsoo Ahn
Abstract: Large language models (LLMs) are increasingly recognized as powerful tools for scientific discovery, particularly in molecular science. A fundamental requirement for these models is the ability to accurately understand molecular structures, commonly encoded in the SMILES representation. However, current LLMs struggle to interpret SMILES, even failing to carry out basic tasks such as counting molecular rings. To address this limitation, we introduce CLEANMOL, a novel framework that formulates SMILES parsing into a suite of clean and deterministic tasks explicitly designed to promote graph-level molecular comprehension. These tasks span from subgraph matching to global graph matching, providing structured supervision aligned with molecular structural properties. We construct a molecular pretraining dataset with adaptive difficulty scoring and pre-train open-source LLMs on these tasks. Our results show that CLEANMOL not only enhances structural comprehension but also achieves the best or competes with the baseline on the Mol-Instructions benchmark.
Authors: Yaxin Hou, Yuheng Jia
Abstract: This paper studies the long-tailed semi-supervised learning (LTSSL) with distribution mismatch, where the class distribution of the labeled training data follows a long-tailed distribution and mismatches with that of the unlabeled training data. Most existing methods introduce auxiliary classifiers (experts) to model various unlabeled data distributions and produce pseudo-labels, but the expertises of various experts are not fully utilized. We observe that different experts are good at predicting different intervals of samples, e.g., long-tailed expert is skilled in samples located in the head interval and uniform expert excels in samples located in the medium interval. Therefore, we propose a dynamic expert assignment module that can estimate the class membership (i.e., head, medium, or tail class) of samples, and dynamically assigns suitable expert to each sample based on the estimated membership to produce high-quality pseudo-label in the training phase and produce prediction in the testing phase. We also theoretically reveal that integrating different experts' strengths will lead to a smaller generalization error bound. Moreover, we find that the deeper features are more biased toward the head class but with more discriminative ability, while the shallower features are less biased but also with less discriminative ability. We, therefore, propose a multi-depth feature fusion module to utilize different depth features to mitigate the model bias. Our method demonstrates its effectiveness through comprehensive experiments on the CIFAR-10-LT, STL-10-LT, and SVHN-LT datasets across various settings. The code is available at https://github.com/yaxinhou/Meta-Expert.
Authors: C\'eline Comte (CNRS, LAAS-SARA, LAAS-RISC), Pascal Moyal (IECL)
Abstract: In this paper, we introduce a versatile scheme for optimizing the arrival rates of quasi-reversible queueing systems. We first propose an alternative definition of quasi-reversibility that encompasses reversibility and highlights the importance of the definition of customer classes. In a second time, we introduce balanced arrival control policies, which generalize the notion of balanced arrival rates introduced in the context of Whittle networks, to the much broader class of quasi-reversible queueing systems. We prove that supplementing a quasi-reversible queueing system with a balanced arrival-control policy preserves the quasi-reversibility, and we specify the form of the stationary measures. We revisit two canonical examples of quasi-reversible queueing systems, Whittle networks and order-independent queues. Lastly, we focus on the problem of admission control and leverage our results in the frameworks of optimization and reinforcement learning.
Authors: Huishuai Zhang, Bohan Wang, Luoxin Chen
Abstract: We introduce AdamS, a simple yet effective alternative to Adam for large language model (LLM) pretraining and post-training. By leveraging a novel denominator, i.e., the root of weighted sum of squares of the momentum and the current gradient, AdamS eliminates the need for second-moment estimates. Hence, AdamS is efficient, matching the memory and compute footprint of SGD with momentum while delivering superior optimization performance. Moreover, AdamS is easy to adopt: it can directly inherit hyperparameters of AdamW, and is entirely model-agnostic, integrating seamlessly into existing pipelines without modifications to optimizer APIs or architectures. The motivation behind AdamS stems from the observed $(L_0, L_1)$ smoothness properties in transformer objectives, where local smoothness is governed by gradient magnitudes that can be further approximated by momentum magnitudes. We establish rigorous theoretical convergence guarantees and provide practical guidelines for hyperparameter selection. Empirically, AdamS demonstrates strong performance in various tasks, including pre-training runs on GPT-2 and Llama2 (up to 13B parameters) and reinforcement learning in post-training regimes. With its efficiency, simplicity, and theoretical grounding, AdamS stands as a compelling alternative to existing optimizers.
Authors: Manuel Ruiz-Botella, Marta Sales-Pardo, Roger Guimer\`a
Abstract: Developing new molecular compounds is crucial to address pressing challenges, from health to environmental sustainability. However, exploring the molecular space to discover new molecules is difficult due to the vastness of the space. Here we introduce CoCoGraph, a collaborative and constrained graph diffusion model capable of generating molecules that are guaranteed to be chemically valid. Thanks to the constraints built into the model and to the collaborative mechanism, CoCoGraph outperforms state-of-the-art approaches on standard benchmarks while requiring up to an order of magnitude fewer parameters. Analysis of 36 chemical properties also demonstrates that CoCoGraph generates molecules with distributions more closely matching real molecules than current models. Leveraging the model's efficiency, we created a database of 8.2M million synthetically generated molecules and conducted a Turing-like test with organic chemistry experts to further assess the plausibility of the generated molecules, and potential biases and limitations of CoCoGraph.
Authors: Huanyu Liu, Jia Li, Hao Zhu, Kechi Zhang, Yihong Dong, Ge Li
Abstract: How to design reinforcement learning (RL) tasks that effectively unleash the reasoning capability of large language models (LLMs) remains an open question. Existing RL tasks (e.g., math, programming, and constructing reasoning tasks) suffer from three key limitations: (1) Scalability. They rely heavily on human annotation or expensive LLM synthesis to generate sufficient training data. (2) Verifiability. LLMs' outputs are hard to verify automatically and reliably. (3) Controllable Difficulty. Most tasks lack fine-grained difficulty control, making it hard to train LLMs to develop reasoning ability from easy to hard. To address these limitations, we propose Saturn, a SAT-based RL framework that uses Boolean Satisfiability (SAT) problems to train and evaluate LLM reasoning. Saturn enables scalable task construction, rule-based verification, and precise difficulty control. Saturn designs a curriculum learning pipeline that continuously improves LLMs' reasoning capability by constructing SAT tasks of increasing difficulty and training LLMs from easy to hard. To ensure stable training, we design a principled mechanism to control difficulty transitions. We introduce Saturn-2.6k, a dataset of 2,660 SAT problems with varying difficulty. It supports the evaluation of how LLM reasoning changes with problem difficulty. We apply Saturn to DeepSeek-R1-Distill-Qwen and obtain Saturn-1.5B and Saturn-7B. We achieve several notable results: (1) On SAT problems, Saturn-1.5B and Saturn-7B achieve average pass@3 improvements of +14.0 and +28.1, respectively. (2) On math and programming tasks, Saturn-1.5B and Saturn-7B improve average scores by +4.9 and +1.8 on benchmarks (e.g., AIME, LiveCodeBench). (3) Compared to the state-of-the-art (SOTA) approach in constructing RL tasks, Saturn achieves further improvements of +8.8%. We release the source code, data, and models to support future research.
Authors: Ahmed K. Kadhim, Lei Jiao, Rishad Shafik, Ole-Christoffer Granmo
Abstract: The increasing complexity of large-scale language models has amplified concerns regarding their interpretability and reusability. While traditional embedding models like Word2Vec and GloVe offer scalability, they lack transparency and often behave as black boxes. Conversely, interpretable models such as the Tsetlin Machine (TM) have shown promise in constructing explainable learning systems, though they previously faced limitations in scalability and reusability. In this paper, we introduce Omni Tsetlin Machine AutoEncoder (Omni TM-AE), a novel embedding model that fully exploits the information contained in the TM's state matrix, including literals previously excluded from clause formation. This method enables the construction of reusable, interpretable embeddings through a single training phase. Extensive experiments across semantic similarity, sentiment classification, and document clustering tasks show that Omni TM-AE performs competitively with and often surpasses mainstream embedding models. These results demonstrate that it is possible to balance performance, scalability, and interpretability in modern Natural Language Processing (NLP) systems without resorting to opaque architectures.
Authors: Yang Chen, Zhuolin Yang, Zihan Liu, Chankyu Lee, Peng Xu, Mohammad Shoeybi, Bryan Catanzaro, Wei Ping
Abstract: Despite recent progress in large-scale reinforcement learning (RL) for reasoning, the training recipe for building high-performing reasoning models remains elusive. Key implementation details of frontier models, such as DeepSeek-R1, including data curation strategies and RL training recipe, are often omitted. Moreover, recent research indicates distillation remains more effective than RL for smaller models. In this work, we demonstrate that large-scale RL can significantly enhance the reasoning capabilities of strong, small- and mid-sized models, achieving results that surpass those of state-of-the-art distillation-based models. We systematically study the RL training process through extensive ablations and propose a simple yet effective approach: first training on math-only prompts, then on code-only prompts. Notably, we find that math-only RL not only significantly enhances the performance of strong distilled models on math benchmarks (e.g., +14.6% / +17.2% on AIME 2025 for the 7B / 14B models), but also code reasoning tasks (e.g., +6.8% / +5.8% on LiveCodeBench for the 7B / 14B models). In addition, extended code-only RL iterations further improve performance on code benchmarks with minimal or no degradation in math results. We develop a robust data curation pipeline to collect challenging prompts with high-quality, verifiable answers and test cases to enable verification-based RL across both domains. Finally, we identify key experimental insights, including curriculum learning with progressively increasing response lengths and the stabilizing effect of on-policy parameter updates. We find that RL not only elicits the foundational reasoning capabilities acquired during pretraining and supervised fine-tuning (e.g., distillation), but also pushes the limits of the model's reasoning ability, enabling it to solve problems that were previously unsolvable.
Authors: Xiaoqing Zhang, Huabin Zheng, Ang Lv, Yuhan Liu, Zirui Song, Flood Sung, Xiuying Chen, Rui Yan
Abstract: Large language models (LLMs) have been observed to suddenly exhibit advanced reasoning abilities during reinforcement learning (RL), resembling an ``aha moment'' triggered by simple outcome-based rewards. While RL has proven effective in eliciting such breakthroughs in tasks involving mathematics, coding, and vision, it faces significant challenges in multi-scenario games. The diversity of game rules, interaction modes, and environmental complexities often leads to policies that perform well in one scenario but fail to generalize to others. Simply combining multiple scenarios during training introduces additional challenges, such as training instability and poor performance. To overcome these challenges, we propose Divide-Fuse-Conquer, a framework designed to enhance generalization in multi-scenario RL. This approach starts by heuristically grouping games based on characteristics such as rules and difficulties. Specialized models are then trained for each group to excel at games in the group is what we refer to as the divide step. Next, we fuse model parameters from different groups as a new model, and continue training it for multiple groups, until the scenarios in all groups are conquered. Experiments across 18 TextArena games show that Qwen2.5-32B-Align trained with the Divide-Fuse-Conquer strategy reaches a performance level comparable to Claude3.5, achieving 7 wins and 4 draws. We hope our approach can inspire future research on using reinforcement learning to improve the generalization of LLMs.
Authors: Huazi Pan, Yanjun Zhang, Leo Yu Zhang, Scott Adams, Abbas Kouzani, Suiyang Khoo
Abstract: Manipulation of local training data and local updates, i.e., the poisoning attack, is the main threat arising from the collaborative nature of the federated learning (FL) paradigm. Most existing poisoning attacks aim to manipulate local data/models in a way that causes denial-of-service (DoS) issues. In this paper, we introduce a novel attack method, named Federated Learning Sliding Attack (FedSA) scheme, aiming at precisely introducing the extent of poisoning in a subtle controlled manner. It operates with a predefined objective, such as reducing global model's prediction accuracy by 10\%. FedSA integrates robust nonlinear control-Sliding Mode Control (SMC) theory with model poisoning attacks. It can manipulate the updates from malicious clients to drive the global model towards a compromised state, achieving this at a controlled and inconspicuous rate. Additionally, leveraging the robust control properties of FedSA allows precise control over the convergence bounds, enabling the attacker to set the global accuracy of the poisoned model to any desired level. Experimental results demonstrate that FedSA can accurately achieve a predefined global accuracy with fewer malicious clients while maintaining a high level of stealth and adjustable learning rates.
Authors: Zhaoxin Wang, Handing Wang, Cong Tian, Yaochu Jin
Abstract: Multimodal large language models (MLLMs) enable powerful cross-modal reasoning capabilities. However, the expanded input space introduces new attack surfaces. Previous jailbreak attacks often inject malicious instructions from text into less aligned modalities, such as vision. As MLLMs increasingly incorporate cross-modal consistency and alignment mechanisms, such explicit attacks become easier to detect and block. In this work, we propose a novel implicit jailbreak framework termed IJA that stealthily embeds malicious instructions into images via least significant bit steganography and couples them with seemingly benign, image-related textual prompts. To further enhance attack effectiveness across diverse MLLMs, we incorporate adversarial suffixes generated by a surrogate model and introduce a template optimization module that iteratively refines both the prompt and embedding based on model feedback. On commercial models like GPT-4o and Gemini-1.5 Pro, our method achieves attack success rates of over 90% using an average of only 3 queries.
Authors: Xinxing Shi, Xiaoyu Jiang, Mauricio A. \'Alvarez
Abstract: Gaussian Process (GP) Variational Autoencoders (VAEs) extend standard VAEs by replacing the fully factorised Gaussian prior with a GP prior, thereby capturing richer correlations among latent variables. However, performing exact GP inference in large-scale GPVAEs is computationally prohibitive, often forcing existing approaches to rely on restrictive kernel assumptions or large sets of inducing points. In this work, we propose a neighbour-driven approximation strategy that exploits local adjacencies in the latent space to achieve scalable GPVAE inference. By confining computations to the nearest neighbours of each data point, our method preserves essential latent dependencies, allowing more flexible kernel choices and mitigating the need for numerous inducing points. Through extensive experiments on tasks including representation learning, data imputation, and conditional generation, we demonstrate that our approach outperforms other GPVAE variants in both predictive performance and computational efficiency.
Authors: Seyedeh Fatemeh Ebrahimi, Jaakko Peltonen
Abstract: Topic models often fail to capture low-prevalence, domain-critical themes, so-called minority topics, such as mental health themes in online comments. While some existing methods can incorporate domain knowledge, such as expected topical content, methods allowing guidance may require overly detailed expected topics, hindering the discovery of topic divisions and variation. We propose a topic modeling solution via a specially constrained NMF. We incorporate a seed word list characterizing minority content of interest, but we do not require experts to pre-specify their division across minority topics. Through prevalence constraints on minority topics and seed word content across topics, we learn distinct data-driven minority topics as well as majority topics. The constrained NMF is fitted via Karush-Kuhn-Tucker (KKT) conditions with multiplicative updates. We outperform several baselines on synthetic data in terms of topic purity, normalized mutual information, and also evaluate topic quality using Jensen-Shannon divergence (JSD). We conduct a case study on YouTube vlog comments, analyzing viewer discussion of mental health content; our model successfully identifies and reveals this domain-relevant minority content.
Authors: Noga Amit, Omer Reingold, Guy N. Rothblum
Abstract: We revisit the foundations of fairness and its interplay with utility and efficiency in settings where the training data contain richer labels, such as individual types, rankings, or risk estimates, rather than just binary outcomes. In this context, we propose algorithms that achieve stronger notions of evidence-based fairness than are possible in standard supervised learning. Our methods support classification and ranking techniques that preserve accurate subpopulation classification rates, as suggested by the underlying data distributions, across a broad class of classification rules and downstream applications. Furthermore, our predictors enable loss minimization, whether aimed at maximizing utility or in the service of fair treatment. Complementing our algorithmic contributions, we present impossibility results demonstrating that simultaneously achieving accurate classification rates and optimal loss minimization is, in some cases, computationally infeasible. Unlike prior impossibility results, our notions are not inherently in conflict and are simultaneously satisfied by the Bayes-optimal predictor. Furthermore, we show that each notion can be satisfied individually via efficient learning. Our separation thus stems from the computational hardness of learning a sufficiently good approximation of the Bayes-optimal predictor. These computational impossibilities present a choice between two natural and attainable notions of accuracy that could both be motivated by fairness.
Authors: Majid Mohammadi, Siu Lun Chau, Krikamol Muandet
Abstract: Kernel methods are widely used in machine learning due to their flexibility and expressive power. However, their black-box nature poses significant challenges to interpretability, limiting their adoption in high-stakes applications. Shapley value-based feature attribution techniques, such as SHAP and kernel-specific variants like RKHS-SHAP, offer a promising path toward explainability. Yet, computing exact Shapley values remains computationally intractable in general, motivating the development of various approximation schemes. In this work, we introduce PKeX-Shapley, a novel algorithm that utilizes the multiplicative structure of product kernels to enable the exact computation of Shapley values in polynomial time. We show that product-kernel models admit a functional decomposition that allows for a recursive formulation of Shapley values. This decomposition not only yields computational efficiency but also enhances interpretability in kernel-based learning. We also demonstrate how our framework can be generalized to explain kernel-based statistical discrepancies such as the Maximum Mean Discrepancy (MMD) and the Hilbert-Schmidt Independence Criterion (HSIC), thus offering new tools for interpretable statistical inference.
Authors: Mohamed Amine Ketata, David L\"udke, Leo Schwinn, Stephan G\"unnemann
Abstract: Building generative models for relational databases (RDBs) is important for applications like privacy-preserving data release and augmenting real datasets. However, most prior work either focuses on single-table generation or relies on autoregressive factorizations that impose a fixed table order and generate tables sequentially. This approach limits parallelism, restricts flexibility in downstream applications like missing value imputation, and compounds errors due to commonly made conditional independence assumptions. We propose a fundamentally different approach: jointly modeling all tables in an RDB without imposing any order. By using a natural graph representation of RDBs, we propose the Graph-Conditional Relational Diffusion Model (GRDM). GRDM leverages a graph neural network to jointly denoise row attributes and capture complex inter-table dependencies. Extensive experiments on six real-world RDBs demonstrate that our approach substantially outperforms autoregressive baselines in modeling multi-hop inter-table correlations and achieves state-of-the-art performance on single-table fidelity metrics.
Authors: Alejandro Moreno Arcas, Albert Sanchis, Jorge Civera, Alfons Juan
Abstract: Adaptation of foundation models using low-rank methods is a widespread approach. Another way to adapt these models is to employ orthogonal fine-tuning methods, which are less time and memory efficient despite their good generalization properties. In this work, we propose Householder Orthogonal Fine-tuning (HOFT), a novel orthogonal fine-tuning method that aims to alleviate time and space complexity. Moreover, some theoretical properties of the orthogonal fine-tuning paradigm are explored. From this exploration, Scaled Householder Orthogonal Fine-tuning (SHOFT) is proposed. Both HOFT and SHOFT are evaluated in downstream tasks, namely commonsense reasoning, machine translation, subject-driven generation and mathematical reasoning. Compared with state-of-the-art adaptation methods, HOFT and SHOFT show comparable or better results.
Authors: Lucas Maystre, Gabriel Barello, Tudor Berariu, Aleix Cambray, Rares Dolga, Alvaro Ortega Gonzalez, Andrei Nica, David Barber
Abstract: We address the problem of incremental sequence classification, where predictions are updated as new elements in the sequence are revealed. Drawing on temporal-difference learning from reinforcement learning, we identify a temporal-consistency condition that successive predictions should satisfy. We leverage this condition to develop a novel loss function for training incremental sequence classifiers. Through a concrete example, we demonstrate that optimizing this loss can offer substantial gains in data efficiency. We apply our method to text classification tasks and show that it improves predictive accuracy over competing approaches on several benchmark datasets. We further evaluate our approach on the task of verifying large language model generations for correctness in grade-school math problems. Our results show that models trained with our method are better able to distinguish promising generations from unpromising ones after observing only a few tokens.
Authors: Trung V. Phan, George A. Kevrekidis, Soledad Villar, Yannis G. Kevrekidis, Juan M. Bello-Rivas
Abstract: The machine learning methods for data-driven identification of partial differential equations (PDEs) are typically defined for a given number of spatial dimensions and a choice of coordinates the data have been collected in. This dependence prevents the learned evolution equation from generalizing to other spaces. In this work, we reformulate the problem in terms of coordinate- and dimension-independent representations, paving the way toward what we call ``spatially liberated" PDE learning. To this end, we employ a machine learning approach to predict the evolution of scalar field systems expressed in the formalism of exterior calculus, which is coordinate-free and immediately generalizes to arbitrary dimensions by construction. We demonstrate the performance of this approach in the FitzHugh-Nagumo and Barkley reaction-diffusion models, as well as the Patlak-Keller-Segel model informed by in-situ chemotactic bacteria observations. We provide extensive numerical experiments that demonstrate that our approach allows for seamless transitions across various spatial contexts. We show that the field dynamics learned in one space can be used to make accurate predictions in other spaces with different dimensions, coordinate systems, boundary conditions, and curvatures.
Authors: Chen Gong, Rui Xing, Zhenzhe Zheng, Fan Wu
Abstract: The demand for machine learning (ML) model training on edge devices is escalating due to data privacy and personalized service needs. However, we observe that current on-device model training is hampered by the under-utilization of on-device data, due to low training throughput, limited storage and diverse data importance. To improve data resource utilization, we propose a two-stage data selection framework {\sf Titan} to select the most important data batch from streaming data for model training with guaranteed efficiency and effectiveness. Specifically, in the first stage, {\sf Titan} filters out a candidate dataset with potentially high importance in a coarse-grained manner.In the second stage of fine-grained selection, we propose a theoretically optimal data selection strategy to identify the data batch with the highest model performance improvement to current training round. To further enhance time-and-resource efficiency, {\sf Titan} leverages a pipeline to co-execute data selection and model training, and avoids resource conflicts by exploiting idle computing resources. We evaluate {\sf Titan} on real-world edge devices and three representative edge computing tasks with diverse models and data modalities. Empirical results demonstrate that {\sf Titan} achieves up to $43\%$ reduction in training time and $6.2\%$ increase in final accuracy with minor system overhead, such as data processing delay, memory footprint and energy consumption.
Authors: Thibaud Gloaguen, Mark Vero, Robin Staab, Martin Vechev
Abstract: Finetuning openly accessible Large Language Models (LLMs) has become standard practice for achieving task-specific performance improvements. Until now, finetuning has been regarded as a controlled and secure process in which training on benign datasets led to predictable behaviors. In this paper, we demonstrate for the first time that an adversary can create poisoned LLMs that initially appear benign but exhibit malicious behaviors once finetuned by downstream users. To this end, our proposed attack, FAB (Finetuning-Activated Backdoor), poisons an LLM via meta-learning techniques to simulate downstream finetuning, explicitly optimizing for the emergence of malicious behaviors in the finetuned models. At the same time, the poisoned LLM is regularized to retain general capabilities and to exhibit no malicious behaviors prior to finetuning. As a result, when users finetune the seemingly benign model on their own datasets, they unknowingly trigger its hidden backdoor behavior. We demonstrate the effectiveness of FAB across multiple LLMs and three target behaviors: unsolicited advertising, refusal, and jailbreakability. Additionally, we show that FAB-backdoors are robust to various finetuning choices made by the user (e.g., dataset, number of steps, scheduler). Our findings challenge prevailing assumptions about the security of finetuning, revealing yet another critical attack vector exploiting the complexities of LLMs.
Authors: Yu Zuo, Dalin Qin, Yi Wang
Abstract: The growing complexity of power systems has made accurate load forecasting more important than ever. An increasing number of advanced load forecasting methods have been developed. However, the static design of current methods offers no mechanism for human-model interaction. As the primary users of forecasting models, system operators often find it difficult to understand and apply these advanced models, which typically requires expertise in artificial intelligence (AI). This also prevents them from incorporating their experience and real-world contextual understanding into the forecasting process. Recent breakthroughs in large language models (LLMs) offer a new opportunity to address this issue. By leveraging their natural language understanding and reasoning capabilities, we propose an LLM-based multi-agent collaboration framework to bridge the gap between human operators and forecasting models. A set of specialized agents is designed to perform different tasks in the forecasting workflow and collaborate via a dedicated communication mechanism. This framework embeds interactive mechanisms throughout the load forecasting pipeline, reducing the technical threshold for non-expert users and enabling the integration of human experience. Our experiments demonstrate that the interactive load forecasting accuracy can be significantly improved when users provide proper insight in key stages. Our cost analysis shows that the framework remains affordable, making it practical for real-world deployment.
Authors: Max Weltevrede, Moritz A. Zanger, Matthijs T. J. Spaan, Wendelin B\"ohmer
Abstract: In the zero-shot policy transfer setting in reinforcement learning, the goal is to train an agent on a fixed set of training environments so that it can generalise to similar, but unseen, testing environments. Previous work has shown that policy distillation after training can sometimes produce a policy that outperforms the original in the testing environments. However, it is not yet entirely clear why that is, or what data should be used to distil the policy. In this paper, we prove, under certain assumptions, a generalisation bound for policy distillation after training. The theory provides two practical insights: for improved generalisation, you should 1) train an ensemble of distilled policies, and 2) distil it on as much data from the training environments as possible. We empirically verify that these insights hold in more general settings, when the assumptions required for the theory no longer hold. Finally, we demonstrate that an ensemble of policies distilled on a diverse dataset can generalise significantly better than the original agent.
Authors: Shpresim Sadiku, Kartikeya Chitranshi, Hiroshi Kera, Sebastian Pokutta
Abstract: Plausible counterfactual explanations (p-CFEs) are perturbations that minimally modify inputs to change classifier decisions while remaining plausible under the data distribution. In this study, we demonstrate that classifiers can be trained on p-CFEs labeled with induced \emph{incorrect} target classes to classify unperturbed inputs with the original labels. While previous studies have shown that such learning is possible with adversarial perturbations, we extend this paradigm to p-CFEs. Interestingly, our experiments reveal that learning from p-CFEs is even more effective: the resulting classifiers achieve not only high in-distribution accuracy but also exhibit significantly reduced bias with respect to spurious correlations.
Authors: Benjamin Herdeanu, Juan Nathaniel, Carla Roesch, Jatan Buch, Gregor Ramien, Johannes Haux, Pierre Gentine
Abstract: Causal discovery for dynamical systems poses a major challenge in fields where active interventions are infeasible. Most methods used to investigate these systems and their associated benchmarks are tailored to deterministic, low-dimensional and weakly nonlinear time-series data. To address these limitations, we present CausalDynamics, a large-scale benchmark and extensible data generation framework to advance the structural discovery of dynamical causal models. Our benchmark consists of true causal graphs derived from thousands of coupled ordinary and stochastic differential equations as well as two idealized climate models. We perform a comprehensive evaluation of state-of-the-art causal discovery algorithms for graph reconstruction on systems with noisy, confounded, and lagged dynamics. CausalDynamics consists of a plug-and-play, build-your-own coupling workflow that enables the construction of a hierarchy of physical systems. We anticipate that our framework will facilitate the development of robust causal discovery algorithms that are broadly applicable across domains while addressing their unique challenges. We provide a user-friendly implementation and documentation on https://kausable.github.io/CausalDynamics.
Authors: Victor Dheur, Souhaib Ben Taieb
Abstract: Reliably characterizing the full conditional distribution of a multivariate response variable given a set of covariates is crucial for trustworthy decision-making. However, misspecified or miscalibrated multivariate models may yield a poor approximation of the joint distribution of the response variables, leading to unreliable predictions and suboptimal decisions. Furthermore, standard recalibration methods are primarily limited to univariate settings, while conformal prediction techniques, despite generating multivariate prediction regions with coverage guarantees, do not provide a full probability density function. We address this gap by first introducing a novel notion of latent calibration, which assesses probabilistic calibration in the latent space of a conditional normalizing flow. Second, we propose latent recalibration (LR), a novel post-hoc model recalibration method that learns a transformation of the latent space with finite-sample bounds on latent calibration. Unlike existing methods, LR produces a recalibrated distribution with an explicit multivariate density function while remaining computationally efficient. Extensive experiments on both tabular and image datasets show that LR consistently improves latent calibration error and the negative log-likelihood of the recalibrated models.
Authors: Benedikt H\"oltgen, Nuria Oliver
Abstract: Fairness through Unawareness (FtU) describes the idea that discrimination against demographic groups can be avoided by not considering group membership in the decisions or predictions. This idea has long been criticized in the machine learning literature as not being sufficient to ensure fairness. In addition, the use of additional features is typically thought to increase the accuracy of the predictions for all groups, so that FtU is sometimes thought to be detrimental to all groups. In this paper, we show both theoretically and empirically that FtU can reduce algorithmic discrimination without necessarily reducing accuracy. We connect this insight with the literature on Model Multiplicity, to which we contribute with novel theoretical and empirical results. Furthermore, we illustrate how, in a real-life application, FtU can contribute to the deployment of more equitable policies without losing efficacy. Our findings suggest that FtU is worth considering in practical applications, particularly in high-risk scenarios, and that the use of protected attributes such as gender in predictive models should be accompanied by a clear and well-founded justification.
Authors: Zhichao Zhu, Yang Qi, Hengyuan Ma, Wenlian Lu, Jianfeng Feng
Abstract: The Forward-Forward (FF) algorithm provides a bottom-up alternative to backpropagation (BP) for training neural networks, relying on a layer-wise "goodness" function to guide learning. Existing goodness functions, inspired by energy-based learning (EBL), are typically defined as the sum of squared post-synaptic activations, neglecting the correlations between neurons. In this work, we propose a novel goodness function termed dimensionality compression that uses the effective dimensionality (ED) of fluctuating neural responses to incorporate second-order statistical structure. Our objective minimizes ED for clamped inputs when noise is considered while maximizing it across the sample distribution, promoting structured representations without the need to prepare negative samples. We demonstrate that this formulation achieves competitive performance compared to other non-BP methods. Moreover, we show that noise plays a constructive role that can enhance generalization and improve inference when predictions are derived from the mean of squared outputs, which is equivalent to making predictions based on the energy term. Our findings contribute to the development of more biologically plausible learning algorithms and suggest a natural fit for neuromorphic computing, where stochasticity is a computational resource rather than a nuisance. The code is available at https://github.com/ZhichaoZhu/StochasticForwardForward
URLs: https://github.com/ZhichaoZhu/StochasticForwardForward
Authors: Khoa Tran, Tri Le, Bao Huynh, Hung-Cuong Trinh, Vy-Rin Nguyen
Abstract: Accurate prediction of the Remaining Useful Life (RUL) is essential for enabling timely maintenance of lithium-ion batteries, impacting the operational efficiency of electric applications that rely on them. This paper proposes a RUL prediction approach that leverages data from recent charge-discharge cycles to estimate the number of remaining usable cycles. The approach introduces both a novel signal processing pipeline and a deep learning prediction model. In the signal preprocessing pipeline, a derived capacity feature is computed based on current and capacity signals. Alongside original capacity, voltage and current, these features are denoised and enhanced using statistical metrics and a delta-based method to capture differences between the current and previous cycles. In the prediction model, the processed features are then fed into a hybrid deep learning architecture composed of 1D Convolutional Neural Networks (CNN), Attentional Long Short-Term Memory (A-LSTM), and Ordinary Differential Equation-based LSTM (ODE-LSTM) modules. This architecture is designed to capture both local signal characteristics and long-range temporal dependencies while modeling the continuous-time dynamics of battery degradation. The model is further evaluated using transfer learning across different learning strategies and target data partitioning scenarios. Results indicate that the model maintains robust performance, even when fine-tuned on limited target data. Experimental results on two publicly available large-scale datasets demonstrate that the proposed method outperforms a baseline deep learning approach and machine learning techniques, achieving an RMSE of 101.59, highlighting its strong potential for real-world RUL prediction applications.
Authors: Yun-Cheng Tsai, Samuel Yen-Chi Chen
Abstract: Blockchain transaction data exhibits high dimensionality, noise, and intricate feature entanglement, presenting significant challenges for traditional clustering algorithms. In this study, we conduct a comparative analysis of three clustering approaches: (1) Classical K-Means Clustering, applied to pre-processed feature representations; (2) Hybrid Clustering, wherein classical features are enhanced with quantum random features extracted using randomly initialized quantum neural networks (QNNs); and (3) Fully Quantum Clustering, where a QNN is trained in a self-supervised manner leveraging a SwAV-based loss function to optimize the feature space for clustering directly. The proposed experimental framework systematically investigates the impact of quantum circuit depth and the number of learned prototypes, demonstrating that even shallow quantum circuits can effectively extract meaningful non-linear representations, significantly improving clustering performance.
Authors: Wenwen Qiang, Jingyao Wang, Zeen Song, Jiangmeng Li, Changwen Zheng
Abstract: In this paper, we focus on the out-of-distribution (OOD) generalization of self-supervised learning (SSL). By analyzing the mini-batch construction during the SSL training phase, we first give one plausible explanation for SSL having OOD generalization. Then, from the perspective of data generation and causal inference, we analyze and conclude that SSL learns spurious correlations during the training process, which leads to a reduction in OOD generalization. To address this issue, we propose a post-intervention distribution (PID) grounded in the Structural Causal Model. PID offers a scenario where the spurious variable and label variable is mutually independent. Besides, we demonstrate that if each mini-batch during SSL training satisfies PID, the resulting SSL model can achieve optimal worst-case OOD performance. This motivates us to develop a batch sampling strategy that enforces PID constraints through the learning of a latent variable model. Through theoretical analysis, we demonstrate the identifiability of the latent variable model and validate the effectiveness of the proposed sampling strategy. Experiments conducted on various downstream OOD tasks demonstrate the effectiveness of the proposed sampling strategy.
Authors: Filip Thor, Carl Nettelblad
Abstract: Sequencing a genome to determine an individual's DNA produces an enormous number of short nucleotide subsequences known as reads, which must be reassembled to reconstruct the full genome. We present a method for analyzing this type of data using contrastive learning, in which an encoder model is trained to produce embeddings that cluster together sequences from the same genomic region. The sequential nature of genomic regions is preserved in the form of trajectories through this embedding space. Trained solely to reflect the structure of the genome, the resulting model provides a general representation of $k$-mer sequences, suitable for a range of downstream tasks involving read data. We apply our framework to learn the structure of the $E.\ coli$ genome, and demonstrate its use in simulated ancient DNA (aDNA) read mapping and identification of structural variations. Furthermore, we illustrate the potential of using this type of model for metagenomic species identification. We show how incorporating a domain-specific noise model can enhance embedding robustness, and how a supervised contrastive learning setting can be adopted when a linear reference genome is available, by introducing a distance thresholding parameter $\Gamma$. The model can also be trained fully self-supervised on read data, enabling analysis without the need to construct a full genome assembly using specialized algorithms. Small prediction heads based on a pre-trained embedding are shown to perform on par with BWA-aln, the current gold standard approach for aDNA mapping, in terms of accuracy and runtime for short genomes. Given the method's favorable scaling properties with respect to total genome size, inference using our approach is highly promising for metagenomic applications and for mapping to genomes comparable in size to the human genome.
Authors: Beier Luo, Shuoyuan Wang, Yixuan Li, Hongxin Wei
Abstract: Post-training of large language models is essential for adapting pre-trained language models (PLMs) to align with human preferences and downstream tasks. While PLMs typically exhibit well-calibrated confidence, post-trained language models (PoLMs) often suffer from over-confidence, assigning high confidence to both correct and incorrect outputs, which can undermine reliability in critical applications. A major obstacle in calibrating PoLMs is the scarcity of labeled data for individual downstream tasks. To address this, we propose Disagreement-Aware Confidence Alignment (DACA), a novel unsupervised method to optimize the parameters (e.g., temperature $\tau$) in post-hoc confidence calibration. Our method is motivated by the under-confidence issue caused by prediction disagreement between the PLM and PoLM while aligning their confidence via temperature scaling. Theoretically, the PLM's confidence underestimates PoLM's prediction accuracy on disagreement examples, causing a larger $\tau$ and producing under-confident predictions. DACA mitigates this by selectively using only agreement examples for calibration, effectively decoupling the influence of disagreement. In this manner, our method avoids an overly large $\tau$ in temperature scaling caused by disagreement examples, improving calibration performance. Extensive experiments demonstrate the effectiveness of our method, improving the average ECE of open-sourced and API-based LLMs (e.g. GPT-4o) by up to 15.08$\%$ on common benchmarks.
Authors: Seonghwan Park, Jueun Mun, Donghyun Oh, Namhoon Lee
Abstract: Concept bottleneck models (CBMs) ensure interpretability by decomposing predictions into human interpretable concepts. Yet the annotations used for training CBMs that enable this transparency are often noisy, and the impact of such corruption is not well understood. In this study, we present the first systematic study of noise in CBMs and show that even moderate corruption simultaneously impairs prediction performance, interpretability, and the intervention effectiveness. Our analysis identifies a susceptible subset of concepts whose accuracy declines far more than the average gap between noisy and clean supervision and whose corruption accounts for most performance loss. To mitigate this vulnerability we propose a two-stage framework. During training, sharpness-aware minimization stabilizes the learning of noise-sensitive concepts. During inference, where clean labels are unavailable, we rank concepts by predictive entropy and correct only the most uncertain ones, using uncertainty as a proxy for susceptibility. Theoretical analysis and extensive ablations elucidate why sharpness-aware training confers robustness and why uncertainty reliably identifies susceptible concepts, providing a principled basis that preserves both interpretability and resilience in the presence of noise.
Authors: Wenhao Li, Yuxin Zhang, Gen Luo, Daohai Yu, Rongrong Ji
Abstract: While long-context large language models (LLMs) exhibit remarkable document processing capabilities, their prohibitively high training costs often hinder customized applications. To mitigate this issue, we propose \textit{Sequential Chunk-wise Optimization} (SeCO), a memory-efficient training paradigm that partitions lengthy inputs into manageable chunks. Each chunk independently constructs its computational graph and performs localized backpropagation, ensuring that only one chunk's forward activations are stored in memory. Building on SeCO, we further introduce \textit{Sparse Chunk-wise Optimization} (SpaCO), which reduces computational overhead by selectively propagating gradients to specific chunks and incorporates a carefully designed compensation factor to ensure unbiased gradient estimation. SpaCO decouples the computational cost of backpropagation from the context length, enabling training time to gradually converge to inference time as sequences become longer. Implemented as lightweight training wrappers, both SeCO and SpaCO offer substantial practical benefits. For example, when fine-tuning an 8B model with LoRA on a single RTX 3090 GPU, SeCO expands maximum sequence length from 1K to 16K tokens, while SpaCO demonstrates accelerated training speed -- achieving up to 3x faster than SeCO under the same experimental setup. These innovations provide new insights into optimizing long-context models, making them more accessible for practical applications. We have open-sourced the code at \href{https://github.com/wenhaoli-xmu/seco}{here}.
Authors: Konstantinos Barmpas, Na Lee, Yannis Panagakis, Dimitrios A. Adamos, Nikolaos Laskaris, Stefanos Zafeiriou
Abstract: Recent advances in large-scale pre-trained Electroencephalogram (EEG) models have shown great promise, driving progress in Brain-Computer Interfaces (BCIs) and healthcare applications. However, despite their success, many existing pre-trained models have struggled to fully capture the rich information content of neural oscillations, a limitation that fundamentally constrains their performance and generalizability across diverse BCI tasks. This limitation is frequently rooted in suboptimal architectural design choices which constrain their representational capacity. In this work, we introduce LaBraM++, an enhanced Large Brainwave Foundation Model (LBM) that incorporates principled improvements grounded in robust signal processing foundations. LaBraM++ demonstrates substantial gains across a variety of tasks, consistently outperforming its originally-based architecture and achieving competitive results when compared to other open-source LBMs. Its superior performance and training efficiency highlight its potential as a strong foundation for future advancements in LBMs.
Authors: Phillip Mueller, Jannik Wiese, Sebastian Mueller, Lars Mikelsons
Abstract: Datasets in engineering domains are often small, sparsely labeled, and contain numerical as well as categorical conditions. Additionally. computational resources are typically limited in practical applications which hinders the adoption of generative models for engineering tasks. We introduce a novel masked-conditioning approach, that enables generative models to work with sparse, mixed-type data. We mask conditions during training to simulate sparse conditions at inference time. For this purpose, we explore the use of various sparsity schedules that show different strengths and weaknesses. In addition, we introduce a flexible embedding that deals with categorical as well as numerical conditions. We integrate our method into an efficient variational autoencoder as well as a latent diffusion model and demonstrate the applicability of our approach on two engineering-related datasets of 2D point clouds and images. Finally, we show that small models trained on limited data can be coupled with large pretrained foundation models to improve generation quality while retaining the controllability induced by our conditioning scheme.
Authors: Hany Abdulsamad, Sahel Iqbal, Simo S\"arkk\"a
Abstract: Optimal decision-making under partial observability requires agents to balance reducing uncertainty (exploration) against pursuing immediate objectives (exploitation). In this paper, we introduce a novel policy optimization framework for continuous partially observable Markov decision processes (POMDPs) that explicitly addresses this challenge. Our method casts policy learning as probabilistic inference in a non-Markovian Feynman--Kac model that inherently captures the value of information gathering by anticipating future observations, without requiring extrinsic exploration bonuses or handcrafted heuristics. To optimize policies under this model, we develop a nested sequential Monte Carlo~(SMC) algorithm that efficiently estimates a history-dependent policy gradient under samples from the optimal trajectory distribution induced by the POMDP. We demonstrate the effectiveness of our algorithm across standard continuous POMDP benchmarks, where existing methods struggle to act under uncertainty.
Authors: Ziwei Luo, Fredrik K. Gustafsson, Jens Sj\"olund, Thomas B. Sch\"on
Abstract: This work presents a forward-only diffusion (FoD) approach for generative modelling. In contrast to traditional diffusion models that rely on a coupled forward-backward diffusion scheme, FoD directly learns data generation through a single forward diffusion process, yielding a simple yet efficient generative framework. The core of FoD is a state-dependent linear stochastic differential equation that involves a mean-reverting term in both the drift and diffusion functions. This mean-reversion property guarantees the convergence to clean data, naturally simulating a stochastic interpolation between source and target distributions. More importantly, FoD is analytically tractable and is trained using a simple stochastic flow matching objective, enabling a few-step non-Markov chain sampling during inference. The proposed FoD model, despite its simplicity, achieves competitive performance on various image-conditioned (e.g., image restoration) and unconditional generation tasks, demonstrating its effectiveness in generative modelling. Our code is available at https://github.com/Algolzw/FoD.
Authors: Bang You, Puze Liu, Huaping Liu, Jan Peters, Oleg Arenz
Abstract: Simplicity is a powerful inductive bias. In reinforcement learning, regularization is used for simpler policies, data augmentation for simpler representations, and sparse reward functions for simpler objectives, all that, with the underlying motivation to increase generalizability and robustness by focusing on the essentials. Supplementary to these techniques, we investigate how to promote simple behavior throughout the episode. To that end, we introduce a modification of the reinforcement learning problem that additionally maximizes the total correlation within the induced trajectories. We propose a practical algorithm that optimizes all models, including policy and state representation, based on a lower-bound approximation. In simulated robot environments, our method naturally generates policies that induce periodic and compressible trajectories, and that exhibit superior robustness to noise and changes in dynamics compared to baseline methods, while also improving performance in the original tasks.
Authors: Nicolas Keriven
Abstract: Oversmoothing has long been identified as a major limitation of Graph Neural Networks (GNNs): input node features are smoothed at each layer and converge to a non-informative representation, if the weights of the GNN are sufficiently bounded. This assumption is crucial: if, on the contrary, the weights are sufficiently large, then oversmoothing may not happen. Theoretically, GNN could thus learn to not oversmooth. However it does not really happen in practice, which prompts us to examine oversmoothing from an optimization point of view. In this paper, we analyze backward oversmoothing, that is, the notion that backpropagated errors used to compute gradients are also subject to oversmoothing from output to input. With non-linear activation functions, we outline the key role of the interaction between forward and backward smoothing. Moreover, we show that, due to backward oversmoothing, GNNs provably exhibit many spurious stationary points: as soon as the last layer is trained, the whole GNN is at a stationary point. As a result, we can exhibit regions where gradients are near-zero while the loss stays high. The proof relies on the fact that, unlike forward oversmoothing, backward errors are subjected to a linear oversmoothing even in the presence of non-linear activation function, such that the average of the output error plays a key role. Additionally, we show that this phenomenon is specific to deep GNNs, and exhibit counter-example Multi-Layer Perceptron. This paper is a step toward a more complete comprehension of the optimization landscape specific to GNNs.
Authors: Chengcan Wu, Zhixin Zhang, Zeming Wei, Yihao Zhang, Meng Sun
Abstract: The significant progress of large language models (LLMs) has led to remarkable achievements across numerous applications. However, their ability to generate harmful content has sparked substantial safety concerns. Despite the implementation of safety alignment techniques during the pre-training phase, recent research indicates that fine-tuning LLMs on adversarial or even benign data can inadvertently compromise their safety. In this paper, we re-examine the fundamental issue of why fine-tuning on non-harmful data still results in safety degradation. We introduce a safety-aware probing (SAP) optimization framework designed to mitigate the safety risks of fine-tuning LLMs. Specifically, SAP incorporates a safety-aware probe into the gradient propagation process, mitigating the model's risk of safety degradation by identifying potential pitfalls in gradient directions, thereby enhancing task-specific performance while successfully preserving model safety. Our extensive experimental results demonstrate that SAP effectively reduces harmfulness below the original fine-tuned model and achieves comparable test loss to standard fine-tuning methods. Our code is available at https://github.com/ChengcanWu/SAP.
Authors: Pilhwa Lee, Shashank Gupta
Abstract: Minimum attention applies the least action principle in the changes of control concerning state and time, first proposed by Brockett. The involved regularization is highly relevant in emulating biological control, such as motor learning. We apply minimum attention in reinforcement learning (RL) as part of the rewards and investigate its connection to meta-learning and stabilization. Specifically, model-based meta-learning with minimum attention is explored in high-dimensional nonlinear dynamics. Ensemble-based model learning and gradient-based meta-policy learning are alternately performed. Empirically, we show that the minimum attention does show outperforming competence in comparison to the state-of-the-art algorithms in model-free and model-based RL, i.e., fast adaptation in few shots and variance reduction from the perturbations of the model and environment. Furthermore, the minimum attention demonstrates the improvement in energy efficiency.
Authors: Julien Laasri, Marc Revol
Abstract: As Kalyan T. Talluri and Garrett J. Van Ryzin describe in their work [3], Revenue Management aims to maximize an organization's revenue by considering three types of decision categories: structural, pricing, and quantity. In this document, our primary focus will be on decisions related to pricing and quantity for the sale of airline tickets on a direct flight over a certain number of time periods. More specifically, we will only focus on the optimization aspect of this problem. We will assume the demand data to be given, since Air France estimates it beforehand using real data. Similarly, we assume all price options to be predetermined by Air France's algorithms and verified by their analysts. Our objective will be to maximize the revenue of a direct flight by choosing the prices for each product from the predefined set of options. -- Comme d\'ecrit par Kalyan T. Talluri et Garrett J. Van Ryzin dans leur ouvrage [3], le Revenue Management consiste en la maximisation du revenu d'un organisme \`a partir de trois types de cat\'egories de d\'ecision : structurelles, prix et quantit\'e. Dans ce document, nous nous int\'eresserons principalement aux d\'ecisions de type prix et quantit\'e pour la vente de billets d'avion sur un vol direct au cours d'un certain nombre de pas de temps. Plus pr\'ecis\'ement, nous nous situerons dans la partie optimisation du probl\`eme. Nous prendrons ainsi les donn\'ees de demande comme acquises, car elles sont estim\'ees au pr\'ealable par Air France \`a partir des donn\'ees r\'eelles. De m\^eme, pour chaque produit que l'on cherchera \`a vendre, on nous impose en amont les prix possibles que l'on a droit d'utiliser et qui se basent sur des algorithmes d'Air France dont les r\'esultats sont v\'erifi\'es par des analystes. Notre but sera alors de maximiser le revenu d'un vol direct en choisissant les prix de chaque produit parmi ceux impos\'es.
Authors: Hannah Markgraf, Michael Eichelbeck, Daria Cappey, Selin Demirt\"urk, Yara Schattschneider, Matthias Althoff
Abstract: Offline reinforcement learning (RL) has gained traction as a powerful paradigm for learning control policies from pre-collected data, eliminating the need for costly or risky online interactions. While many open-source libraries offer robust implementations of offline RL algorithms, they all rely on datasets composed of experience tuples consisting of state, action, next state, and reward. Managing, curating, and distributing such datasets requires suitable infrastructure. Although static datasets exist for established benchmark problems, no standardized or scalable solution supports developing and sharing datasets for novel or user-defined benchmarks. To address this gap, we introduce PyTupli, a Python-based tool to streamline the creation, storage, and dissemination of benchmark environments and their corresponding tuple datasets. PyTupli includes a lightweight client library with defined interfaces for uploading and retrieving benchmarks and data. It supports fine-grained filtering at both the episode and tuple level, allowing researchers to curate high-quality, task-specific datasets. A containerized server component enables production-ready deployment with authentication, access control, and automated certificate provisioning for secure use. By addressing key barriers in dataset infrastructure, PyTupli facilitates more collaborative, reproducible, and scalable offline RL research.
Authors: Ayano Nakai-Kasai, Tadashi Wadayama
Abstract: Graph-structured data is a type of data to be obtained associated with a graph structure where vertices and edges describe some kind of data correlation. This paper proposes a regression method on graph-structured data, which is based on multi-output Gaussian processes (MOGP), to capture both the correlation between vertices and the correlation between associated data. The proposed formulation is built on the definition of MOGP. This allows it to be applied to a wide range of data configurations and scenarios. Moreover, it has high expressive capability due to its flexibility in kernel design. It includes existing methods of Gaussian processes for graph-structured data as special cases and is possible to remove restrictions on data configurations, model selection, and inference scenarios in the existing methods. The performance of extensions achievable by the proposed formulation is evaluated through computer experiments with synthetic and real data.
Authors: Fares B. Mehouachi, Saif Eddin Jabari
Abstract: We introduce FlowMixer, a neural architecture that leverages constrained matrix operations to model structured spatiotemporal patterns. At its core, FlowMixer incorporates non-negative matrix mixing layers within a reversible mapping framework-applying transforms before mixing and their inverses afterward. This shape-preserving design enables a Kronecker-Koopman eigenmode framework that bridges statistical learning with dynamical systems theory, providing interpretable spatiotemporal patterns and facilitating direct algebraic manipulation of prediction horizons without retraining. Extensive experiments across diverse domains demonstrate FlowMixer's robust long-horizon forecasting capabilities while effectively modeling physical phenomena such as chaotic attractors and turbulent flows. These results suggest that architectural constraints can simultaneously enhance predictive performance and mathematical interpretability in neural forecasting systems.
Authors: Hyunjin Seo, Taewon Kim, Sihyun Yu, SungSoo Ahn
Abstract: Masked diffusion models (MDMs) have achieved notable progress in modeling discrete data, while their potential in molecular generation remains underexplored. In this work, we explore their potential and introduce the surprising result that naively applying standards MDMs severely degrades the performance. We identify the critical cause of this issue as a state-clashing problem-where the forward diffusion of distinct molecules collapse into a common state, resulting in a mixture of reconstruction targets that cannot be learned using typical reverse diffusion process with unimodal predictions. To mitigate this, we propose Masked Element-wise Learnable Diffusion (MELD) that orchestrates per-element corruption trajectories to avoid collision between distinct molecular graphs. This is achieved through a parameterized noise scheduling network that assigns distinct corruption rates to individual graph elements, i.e., atoms and bonds. Extensive experiments on diverse molecular benchmarks reveal that MELD markedly enhances overall generation quality compared to element-agnostic noise scheduling, increasing the chemical validity of vanilla MDMs on ZINC250K from 15% to 93%, Furthermore, it achieves state-of-the-art property alignment in conditional generation tasks.
Authors: Tillmann Rheude, Roland Eils, Benjamin Wild
Abstract: Real-world machine learning applications often involve data from multiple modalities that must be integrated effectively to make robust predictions. However, in many practical settings, not all modalities are available for every sample, and acquiring additional modalities can be costly. This raises the question: which samples should be prioritized for additional modality acquisition when resources are limited? While prior work has explored individual-level acquisition strategies and training-time active learning paradigms, test-time and cohort-based acquisition remain underexplored despite their importance in many real-world settings. We introduce Cohort-based Active Modality Acquisition (CAMA), a novel test-time setting to formalize the challenge of selecting which samples should receive additional modalities. We derive acquisition strategies that leverage a combination of generative imputation and discriminative modeling to estimate the expected benefit of acquiring missing modalities based on common evaluation metrics. We also introduce upper-bound heuristics that provide performance ceilings to benchmark acquisition strategies. Experiments on common multimodal datasets demonstrate that our proposed imputation-based strategies can more effectively guide the acquisition of new samples in comparison to those relying solely on unimodal information, entropy guidance, and random selections. Our work provides an effective solution for optimizing modality acquisition at the cohort level, enabling better utilization of resources in constrained settings.
Authors: Eleftherios Kalafatis, Konstantinos Mitsis, Konstantia Zarkogianni, Maria Athanasiou, Konstantina Nikita
Abstract: Serious Games (SGs) are nowadays shifting focus to include procedural content generation (PCG) in the development process as a means of offering personalized and enhanced player experience. However, the development of a framework to assess the impact of PCG techniques when integrated into SGs remains particularly challenging. This study proposes a methodology for automated evaluation of PCG integration in SGs, incorporating deep reinforcement learning (DRL) game testing agents. To validate the proposed framework, a previously introduced SG featuring card game mechanics and incorporating three different versions of PCG for nonplayer character (NPC) creation has been deployed. Version 1 features random NPC creation, while versions 2 and 3 utilize a genetic algorithm approach. These versions are used to test the impact of different dynamic SG environments on the proposed framework's agents. The obtained results highlight the superiority of the DRL game testing agents trained on Versions 2 and 3 over those trained on Version 1 in terms of win rate (i.e. number of wins per played games) and training time. More specifically, within the execution of a test emulating regular gameplay, both Versions 2 and 3 peaked at a 97% win rate and achieved statistically significant higher (p=0009) win rates compared to those achieved in Version 1 that peaked at 94%. Overall, results advocate towards the proposed framework's capability to produce meaningful data for the evaluation of procedurally generated content in SGs.
Authors: Anna Heuser, Thomas Kesselheim
Abstract: Motivated by stochastic optimization, we introduce the problem of learning from samples of contextual value distributions. A contextual value distribution can be understood as a family of real-valued distributions, where each sample consists of a context $x$ and a random variable drawn from the corresponding real-valued distribution $D_x$. By minimizing a convex surrogate loss, we learn an empirical distribution $D'_x$ for each context, ensuring a small L\'evy distance to $D_x$. We apply this result to obtain the sample complexity bounds for the learning of an $\epsilon$-optimal policy for stochastic optimization problems defined on an unknown contextual value distribution. The sample complexity is shown to be polynomial for the general case of strongly monotone and stable optimization problems, including Single-item Revenue Maximization, Pandora's Box and Optimal Stopping.
Authors: Alihan H\"uy\"uk, Finale Doshi-Velez
Abstract: Long-term planning, as in reinforcement learning (RL), involves finding strategies: actions that collectively work toward a goal rather than individually optimizing their immediate outcomes. As part of a strategy, some actions are taken at the expense of short-term benefit to enable future actions with even greater returns. These actions are only advantageous if followed up by the actions they facilitate, consequently, they would not have been taken if those follow-ups were not available. In this paper, we quantify such dependencies between planned actions with strategic link scores: the drop in the likelihood of one decision under the constraint that a follow-up decision is no longer available. We demonstrate the utility of strategic link scores through three practical applications: (i) explaining black-box RL agents by identifying strategically linked pairs among decisions they make, (ii) improving the worst-case performance of decision support systems by distinguishing whether recommended actions can be adopted as standalone improvements or whether they are strategically linked hence requiring a commitment to a broader strategy to be effective, and (iii) characterizing the planning processes of non-RL agents purely through interventions aimed at measuring strategic link scores - as an example, we consider a realistic traffic simulator and analyze through road closures the effective planning horizon of the emergent routing behavior of many drivers.
Authors: Tajamul Ashraf, Mohammed Mohsen Peerzada, Moloud Abdar, Yutong Xie, Yuyin Zhou, Xiaofeng Liu, Iqra Altaf Gillani, Janibul Bashir
Abstract: Federated Learning (FL) has emerged as a promising paradigm for collaborative model training while preserving data privacy across decentralized participants. As FL adoption grows, numerous techniques have been proposed to tackle its practical challenges. However, the lack of standardized evaluation across key dimensions hampers systematic progress and fair comparison of FL methods. In this work, we introduce ATR-Bench, a unified framework for analyzing federated learning through three foundational dimensions: Adaptation, Trust, and Reasoning. We provide an in-depth examination of the conceptual foundations, task formulations, and open research challenges associated with each theme. We have extensively benchmarked representative methods and datasets for adaptation to heterogeneous clients and trustworthiness in adversarial or unreliable environments. Due to the lack of reliable metrics and models for reasoning in FL, we only provide literature-driven insights for this dimension. ATR-Bench lays the groundwork for a systematic and holistic evaluation of federated learning with real-world relevance. We will make our complete codebase publicly accessible and a curated repository that continuously tracks new developments and research in the FL literature.
Authors: Wei Xiao, Jiacheng Liu, Zifeng Zhuang, Runze Suo, Shangke Lyu, Donglin Wang
Abstract: Improving the performance of pre-trained policies through online reinforcement learning (RL) is a critical yet challenging topic. Existing online RL fine-tuning methods require continued training with offline pretrained Q-functions for stability and performance. However, these offline pretrained Q-functions commonly underestimate state-action pairs beyond the offline dataset due to the conservatism in most offline RL methods, which hinders further exploration when transitioning from the offline to the online setting. Additionally, this requirement limits their applicability in scenarios where only pre-trained policies are available but pre-trained Q-functions are absent, such as in imitation learning (IL) pre-training. To address these challenges, we propose a method for efficient online RL fine-tuning using solely the offline pre-trained policy, eliminating reliance on pre-trained Q-functions. We introduce PORL (Policy-Only Reinforcement Learning Fine-Tuning), which rapidly initializes the Q-function from scratch during the online phase to avoid detrimental pessimism. Our method not only achieves competitive performance with advanced offline-to-online RL algorithms and online RL approaches that leverage data or policies prior, but also pioneers a new path for directly fine-tuning behavior cloning (BC) policies.
Authors: Ertu\u{g}rul Ke\c{c}eci, M\"ujde G\"uzelkaya, Tufan Kumbasar
Abstract: This paper addresses the System Identification (SYSID) problem within the framework of federated learning. We introduce a novel algorithm, Incremental Clustering-based federated learning method for SYSID (IC-SYSID), designed to tackle SYSID challenges across multiple data sources without prior knowledge. IC-SYSID utilizes an incremental clustering method, ClusterCraft (CC), to eliminate the dependency on the prior knowledge of the dataset. CC starts with a single cluster model and assigns similar local workers to the same clusters by dynamically increasing the number of clusters. To reduce the number of clusters generated by CC, we introduce ClusterMerge, where similar cluster models are merged. We also introduce enhanced ClusterCraft to reduce the generation of similar cluster models during the training. Moreover, IC-SYSID addresses cluster model instability by integrating a regularization term into the loss function and initializing cluster models with scaled Glorot initialization. It also utilizes a mini-batch deep learning approach to manage large SYSID datasets during local training. Through the experiments conducted on a real-world representing SYSID problem, where a fleet of vehicles collaboratively learns vehicle dynamics, we show that IC-SYSID achieves a high SYSID performance while preventing the learning of unstable clusters.
Authors: Ziyue Qiao, Qianyi Cai, Hao Dong, Jiawei Gu, Pengyang Wang, Meng Xiao, Xiao Luo, Hui Xiong
Abstract: This paper addresses the challenge of graph domain adaptation on evolving, multiple out-of-distribution (OOD) graphs. Conventional graph domain adaptation methods are confined to single-step adaptation, making them ineffective in handling continuous domain shifts and prone to catastrophic forgetting. This paper introduces the Graph Continual Adaptive Learning (GCAL) method, designed to enhance model sustainability and adaptability across various graph domains. GCAL employs a bilevel optimization strategy. The "adapt" phase uses an information maximization approach to fine-tune the model with new graph domains while re-adapting past memories to mitigate forgetting. Concurrently, the "generate memory" phase, guided by a theoretical lower bound derived from information bottleneck theory, involves a variational memory graph generation module to condense original graphs into memories. Extensive experimental evaluations demonstrate that GCAL substantially outperforms existing methods in terms of adaptability and knowledge retention.
Authors: Mohammed Al-Qudah, Fadi AlMahamid
Abstract: The rapid expansion of Internet of Things (IoT) devices has introduced critical security challenges, underscoring the need for accurate anomaly detection. Although numerous studies have proposed machine learning (ML) methods for this purpose, limited research systematically examines how different preprocessing steps--normalization, transformation, and feature selection--interact with distinct model architectures. To address this gap, this paper presents a multi-step evaluation framework assessing the combined impact of preprocessing choices on three ML algorithms: RNN-LSTM, autoencoder neural networks (ANN), and Gradient Boosting (GBoosting). Experiments on the IoTID20 dataset shows that GBoosting consistently delivers superior accuracy across preprocessing configurations, while RNN-LSTM shows notable gains with z-score normalization and autoencoders excel in recall, making them well-suited for unsupervised scenarios. By offering a structured analysis of preprocessing decisions and their interplay with various ML techniques, the proposed framework provides actionable guidance to enhance anomaly detection performance in IoT environments.
Authors: Can Chen, David Heurtel-Depeiges, Robert M. Vernon, Christopher James Langmead, Yoshua Bengio, Quentin Fournier
Abstract: Protein language models (pLMs) pre-trained on vast protein sequence databases excel at various downstream tasks but lack the structural knowledge essential for many biological applications. To address this, we integrate structural insights from pre-trained protein graph neural networks (pGNNs) into pLMs through a latent-level contrastive learning task. This task aligns residue representations from pLMs with those from pGNNs across multiple proteins, enriching pLMs with inter-protein structural knowledge. Additionally, we incorporate a physical-level task that infuses intra-protein structural knowledge by optimizing pLMs to predict structural tokens. The proposed dual-task framework effectively incorporates both inter-protein and intra-protein structural knowledge into pLMs. Given the variability in the quality of protein structures in PDB, we further introduce a residue loss selection module, which uses a small model trained on high-quality structures to select reliable yet challenging residue losses for the pLM to learn. Applying our structure alignment method to the state-of-the-art ESM2 and AMPLIFY results in notable performance gains across a wide range of tasks, including a 12.7% increase in ESM2 contact prediction. The data, code, and resulting SaESM2 and SaAMPLIFY models will be released on Hugging Face.
Authors: Peyman Baghershahi, Sourav Medya
Abstract: Prompt tuning methods for Graph Neural Networks (GNNs) have become popular to address the semantic gap between pre-training and fine-tuning steps. However, existing GNN prompting methods rely on labeled data and involve lightweight fine-tuning for downstream tasks. Meanwhile, in-context learning methods for Large Language Models (LLMs) have shown promising performance with no parameter updating and no or minimal labeled data. Inspired by these approaches, in this work, we first introduce a challenging problem setup to evaluate GNN prompting methods. This setup encourages a prompting function to enhance a pre-trained GNN's generalization to a target dataset under covariate shift without updating the GNN's parameters and with no labeled data. Next, we propose a fully unsupervised prompting method based on consistency regularization through pseudo-labeling. We use two regularization techniques to align the prompted graphs' distribution with the original data and reduce biased predictions. Through extensive experiments under our problem setting, we demonstrate that our unsupervised approach outperforms the state-of-the-art prompting methods that have access to labels.
Authors: Nikola Tankovic, Robert Sajina
Abstract: This paper presents a concise review of Contextual Multi-Armed Bandit (CMAB) methods and introduces an experimental framework for scalable, interpretable offer selection, addressing the challenge of fast-changing offers. The approach models context at the product category level, allowing offers to span multiple categories and enabling knowledge transfer across similar offers. This improves learning efficiency and generalization in dynamic environments. The framework extends standard CMAB methodology to support multi-category contexts, and achieves scalability through efficient feature engineering and modular design. Advanced features such as MPG (Member Purchase Gap) and MF (Matrix Factorization) capture nuanced user-offer interactions, with implementation in Python for practical deployment. A key contribution is interpretability at scale: logistic regression models yield transparent weight vectors, accessible via a large language model (LLM) interface for real-time, user-level tracking and explanation of evolving preferences. This enables the generation of detailed member profiles and identification of behavioral patterns, supporting personalized offer optimization and enhancing trust in automated decisions. By situating our prototype alongside established paradigms like Generalized Linear Models and Thompson Sampling, we demonstrate its value for both research and real-world CMAB applications.
Authors: Igor Udovichenko, Olivier Croissant, Anita Toleutaeva, Evgeny Burnaev, Alexander Korotin
Abstract: Risk-averse reinforcement learning finds application in various high-stakes fields. Unlike classical reinforcement learning, which aims to maximize expected returns, risk-averse agents choose policies that minimize risk, occasionally sacrificing expected value. These preferences can be framed through utility theory. We focus on the specific case of the exponential utility function, where we can derive the Bellman equations and employ various reinforcement learning algorithms with few modifications. However, these methods suffer from numerical instability due to the need for exponent computation throughout the process. To address this, we introduce a numerically stable and mathematically sound loss function based on the Itakura-Saito divergence for learning state-value and action-value functions. We evaluate our proposed loss function against established alternatives, both theoretically and empirically. In the experimental section, we explore multiple financial scenarios, some with known analytical solutions, and show that our loss function outperforms the alternatives.
Authors: Noah Amsel, David Persson, Christopher Musco, Robert Gower
Abstract: Computing the polar decomposition and the related matrix sign function, has been a well-studied problem in numerical analysis for decades. More recently, it has emerged as an important subroutine in deep learning, particularly within the Muon optimization framework. However, the requirements in this setting differ significantly from those of traditional numerical analysis. In deep learning, methods must be highly efficient and GPU-compatible, but high accuracy is often unnecessary. As a result, classical algorithms like Newton-Schulz (which suffers from slow initial convergence) and methods based on rational functions (which rely on QR decompositions or matrix inverses) are poorly suited to this context. In this work, we introduce Polar Express, a GPU-friendly algorithm for computing the polar decomposition. Like classical polynomial methods such as Newton-Schulz, our approach uses only matrix-matrix multiplications, making it GPU-compatible. Motivated by earlier work of Chen & Chow and Nakatsukasa & Freund, Polar Express adapts the polynomial update rule at each iteration by solving a minimax optimization problem, and we prove that it enjoys a strong worst-case optimality guarantee. This property ensures both rapid early convergence and fast asymptotic convergence. We also address finite-precision issues, making it stable in bfloat16 in practice. We apply Polar Express within the Muon optimization framework and show consistent improvements in validation loss on large-scale models such as GPT-2, outperforming recent alternatives across a range of learning rates.
Authors: Zebin You, Shen Nie, Xiaolu Zhang, Jun Hu, Jun Zhou, Zhiwu Lu, Ji-Rong Wen, Chongxuan Li
Abstract: In this work, we introduce LLaDA-V, a purely diffusion-based Multimodal Large Language Model (MLLM) that integrates visual instruction tuning with masked diffusion models, representing a departure from the autoregressive paradigms dominant in current multimodal approaches. Built upon LLaDA, a representative large language diffusion model, LLaDA-V incorporates a vision encoder and MLP connector that projects visual features into the language embedding space, enabling effective multimodal alignment. Our empirical investigation reveals several intriguing results: First, LLaDA-V demonstrates promising multimodal performance despite its language model being weaker on purely textual tasks than counterparts like LLaMA3-8B and Qwen2-7B. When trained on the same instruction data, LLaDA-V is highly competitive to LLaMA3-V across multimodal tasks with better data scalability. It also narrows the performance gap to Qwen2-VL, suggesting the effectiveness of its architecture for multimodal tasks. Second, LLaDA-V achieves state-of-the-art performance in multimodal understanding compared to existing hybrid autoregressive-diffusion and purely diffusion-based MLLMs. Our findings suggest that large language diffusion models show promise in multimodal contexts and warrant further investigation in future research. Project page and codes: https://ml-gsai.github.io/LLaDA-V-demo/.
Authors: Yizhuo Chen, Tianchen Wang, You Lyu, Yanlan Hu, Jinyang Li, Tomoyoshi Kimura, Hongjue Zhao, Yigong Hu, Denizhan Kara, Tarek Abdelzaher
Abstract: This work develops the underpinnings of self-supervised placement-aware representation learning given spatially-distributed (multi-view and multimodal) sensor observations, motivated by the need to represent external environmental state in multi-sensor IoT systems in a manner that correctly distills spatial phenomena from the distributed multi-vantage observations. The objective of sensing in IoT systems is, in general, to collectively represent an externally observed environment given multiple vantage points from which sensory observations occur. Pretraining of models that help interpret sensor data must therefore encode the relation between signals observed by sensors and the observers' vantage points in order to attain a representation that encodes the observed spatial phenomena in a manner informed by the specific placement of the measuring instruments, while allowing arbitrary placement. The work significantly advances self-supervised model pretraining from IoT signals beyond current solutions that often overlook the distinctive spatial nature of IoT data. Our framework explicitly learns the dependencies between measurements and geometric observer layouts and structural characteristics, guided by a core design principle: the duality between signals and observer positions. We further provide theoretical analyses from the perspectives of information theory and occlusion-invariant representation learning to offer insight into the rationale behind our design. Experiments on three real-world datasets--covering vehicle monitoring, human activity recognition, and earthquake localization--demonstrate the superior generalizability and robustness of our method across diverse modalities, sensor placements, application-level inference tasks, and spatial scales.
Authors: Chao Pang, Vincent Jeanselme, Young Sang Choi, Xinzhuo Jiang, Zilin Jing, Aparajita Kashyap, Yuta Kobayashi, Yanwei Li, Florent Pollet, Karthik Natarajan, Shalmali Joshi
Abstract: Foundation models hold significant promise in healthcare, given their capacity to extract meaningful representations independent of downstream tasks. This property has enabled state-of-the-art performance across several clinical applications trained on structured electronic health record (EHR) data, even in settings with limited labeled data, a prevalent challenge in healthcare. However, there is little consensus on these models' potential for clinical utility due to the lack of desiderata of comprehensive and meaningful tasks and sufficiently diverse evaluations to characterize the benefit over conventional supervised learning. To address this gap, we propose a suite of clinically meaningful tasks spanning patient outcomes, early prediction of acute and chronic conditions, including desiderata for robust evaluations. We evaluate state-of-the-art foundation models on EHR data consisting of 5 million patients from Columbia University Irving Medical Center (CUMC), a large urban academic medical center in New York City, across 14 clinically relevant tasks. We measure overall accuracy, calibration, and subpopulation performance to surface tradeoffs based on the choice of pre-training, tokenization, and data representation strategies. Our study aims to advance the empirical evaluation of structured EHR foundation models and guide the development of future healthcare foundation models.
Authors: Csaba D\'ek\'any, Stefan Balauca, Robin Staab, Dimitar I. Dimitrov, Martin Vechev
Abstract: Despite recent efforts in Large Language Models (LLMs) safety and alignment, current adversarial attacks on frontier LLMs are still able to force harmful generations consistently. Although adversarial training has been widely studied and shown to significantly improve the robustness of traditional machine learning models, its strengths and weaknesses in the context of LLMs are less understood. Specifically, while existing discrete adversarial attacks are effective at producing harmful content, training LLMs with concrete adversarial prompts is often computationally expensive, leading to reliance on continuous relaxations. As these relaxations do not correspond to discrete input tokens, such latent training methods often leave models vulnerable to a diverse set of discrete attacks. In this work, we aim to bridge this gap by introducing MixAT, a novel method that combines stronger discrete and faster continuous attacks during training. We rigorously evaluate MixAT across a wide spectrum of state-of-the-art attacks, proposing the At Least One Attack Success Rate (ALO-ASR) metric to capture the worst-case vulnerability of models. We show MixAT achieves substantially better robustness (ALO-ASR < 20%) compared to prior defenses (ALO-ASR > 50%), while maintaining a runtime comparable to methods based on continuous relaxations. We further analyze MixAT in realistic deployment settings, exploring how chat templates, quantization, low-rank adapters, and temperature affect both adversarial training and evaluation, revealing additional blind spots in current methodologies. Our results demonstrate that MixAT's discrete-continuous defense offers a principled and superior robustness-accuracy tradeoff with minimal computational overhead, highlighting its promise for building safer LLMs. We provide our code and models at https://github.com/insait-institute/MixAT.
Authors: Adnan Oomerjee, Zafeirios Fountas, Zhongwei Yu, Haitham Bou-Ammar, Jun Wang
Abstract: Despite their impressive capabilities, Large Language Models struggle with generalisation beyond their training distribution, often exhibiting sophisticated pattern interpolation rather than true abstract reasoning (extrapolation). In this work, we approach this limitation through the lens of Information Bottleneck (IB) theory, which posits that model generalisation emerges from an optimal balance between input compression and retention of predictive information in latent representations. We prove using IB theory that decoder-only Transformers are inherently constrained in their ability to form task-optimal sequence representations. We then use this result to demonstrate that periodic global transformation of the internal sequence-level representations (KV cache) is a necessary computational step for improving Transformer generalisation in reasoning tasks. Based on these theoretical insights, we propose a modification to the Transformer architecture, in the form of an additional module that globally rewrites the KV cache at periodic intervals, shifting its capacity away from memorising input prefixes and toward encoding features most useful for predicting future tokens. Our model delivers substantial gains on mathematical reasoning benchmarks, outperforming both vanilla Transformers with up to 3.5x more parameters, as well as heuristic-driven pruning mechanisms for cache compression. Our approach can be seen as a principled generalisation of existing KV-cache compression methods; whereas such methods focus solely on compressing input representations, they often do so at the expense of retaining predictive information, and thus their capabilities are inherently bounded by those of an unconstrained model. This establishes a principled framework to manipulate Transformer memory using information theory, addressing fundamental reasoning limitations that scaling alone cannot overcome.
Authors: Shengyu Feng, Weiwei Sun, Shanda Li, Ameet Talwalkar, Yiming Yang
Abstract: Machine learning (ML) has demonstrated considerable potential in supporting model design and optimization for combinatorial optimization (CO) problems. However, much of the progress to date has been evaluated on small-scale, synthetic datasets, raising concerns about the practical effectiveness of ML-based solvers in real-world, large-scale CO scenarios. Additionally, many existing CO benchmarks lack sufficient training data, limiting their utility for evaluating data-driven approaches. To address these limitations, we introduce FrontierCO, a comprehensive benchmark that covers eight canonical CO problem types and evaluates 16 representative ML-based solvers--including graph neural networks and large language model (LLM) agents. FrontierCO features challenging instances drawn from industrial applications and frontier CO research, offering both realistic problem difficulty and abundant training data. Our empirical results provide critical insights into the strengths and limitations of current ML methods, helping to guide more robust and practically relevant advances at the intersection of machine learning and combinatorial optimization. Our data is available at https://huggingface.co/datasets/CO-Bench/FrontierCO.
Authors: Young Sang Choi, Vincent Jeanselme, Pierre Elias, Shalmali Joshi
Abstract: Multimodal learning is of continued interest in artificial intelligence-based applications, motivated by the potential information gain from combining different types of data. However, modalities collected and curated during development may differ from the modalities available at deployment due to multiple factors including cost, hardware failure, or -- as we argue in this work -- the perceived informativeness of a given modality. Na{\"i}ve estimation of the information gain associated with including an additional modality without accounting for missingness may result in improper estimates of that modality's value in downstream tasks. Our work formalizes the problem of missingness in multimodal learning and demonstrates the biases resulting from ignoring this process. To address this issue, we introduce ICYM2I (In Case You Multimodal Missed It), a framework for the evaluation of predictive performance and information gain under missingness through inverse probability weighting-based correction. We demonstrate the importance of the proposed adjustment to estimate information gain under missingness on synthetic, semi-synthetic, and real-world medical datasets.
Authors: Alessandro Favero, Antonio Sclocchi, Matthieu Wyart
Abstract: Diffusion probabilistic models have become a cornerstone of modern generative AI, yet the mechanisms underlying their generalization remain poorly understood. In fact, if these models were perfectly minimizing their training loss, they would just generate data belonging to their training set, i.e., memorize, as empirically found in the overparameterized regime. We revisit this view by showing that, in highly overparameterized diffusion models, generalization in natural data domains is progressively achieved during training before the onset of memorization. Our results, ranging from image to language diffusion models, systematically support the empirical law that memorization time is proportional to the dataset size. Generalization vs. memorization is then best understood as a competition between time scales. We show that this phenomenology is recovered in diffusion models learning a simple probabilistic context-free grammar with random rules, where generalization corresponds to the hierarchical acquisition of deeper grammar rules as training time grows, and the generalization cost of early stopping can be characterized. We summarize these results in a phase diagram. Overall, our results support that a principled early-stopping criterion - scaling with dataset size - can effectively optimize generalization while avoiding memorization, with direct implications for hyperparameter transfer and privacy-sensitive applications.
Authors: Mingyang Liu, Gabriele Farina, Asuman Ozdaglar
Abstract: Post-training has demonstrated its importance in enhancing the reasoning capabilities of large language models (LLMs). The primary post-training methods can be categorized into supervised fine-tuning (SFT) and reinforcement fine-tuning (RFT). SFT is efficient and well-suited for small language models, but it may lead to overfitting and limit the reasoning abilities of larger models. In contrast, RFT generally yields better generalization but depends heavily on the strength of the base model. To address the limitations of SFT and RFT, we propose Unified Fine-Tuning (UFT), a novel post-training paradigm that unifies SFT and RFT into a single, integrated process. UFT enables the model to effectively explore solutions while incorporating informative supervision signals, bridging the gap between memorizing and thinking underlying existing methods. Notably, UFT outperforms both SFT and RFT in general, regardless of model sizes. Furthermore, we theoretically prove that UFT breaks RFT's inherent exponential sample complexity bottleneck, showing for the first time that unified training can exponentially accelerate convergence on long-horizon reasoning tasks.
Authors: Aleksandra Franz, Hao Wei, Luca Guastoni, Nils Thuerey
Abstract: Despite decades of advancements, the simulation of fluids remains one of the most challenging areas of in scientific computing. Supported by the necessity of gradient information in deep learning, differentiable simulators have emerged as an effective tool for optimization and learning in physics simulations. In this work, we present our fluid simulator PICT, a differentiable pressure-implicit solver coded in PyTorch with Graphics-processing-unit (GPU) support. We first verify the accuracy of both the forward simulation and our derived gradients in various established benchmarks like lid-driven cavities and turbulent channel flows before we show that the gradients provided by our solver can be used to learn complicated turbulence models in 2D and 3D. We apply both supervised and unsupervised training regimes using physical priors to match flow statistics. In particular, we learn a stable sub-grid scale (SGS) model for a 3D turbulent channel flow purely based on reference statistics. The low-resolution corrector trained with our solver runs substantially faster than the highly resolved references, while keeping or even surpassing their accuracy. Finally, we give additional insights into the physical interpretation of different solver gradients, and motivate a physically informed regularization technique. To ensure that the full potential of PICT can be leveraged, it is published as open source: https://github.com/tum-pbs/PICT.
Authors: Shalev Manor, Mohammad Kohandel
Abstract: Inverse problems involving differential equations often require identifying unknown parameters or functions from data. Existing approaches, such as Physics-Informed Neural Networks (PINNs), Universal Differential Equations (UDEs) and Universal Physics-Informed Neural Networks (UPINNs), are effective at isolating either parameters or functions but can face challenges when applied simultaneously due to solution non-uniqueness. In this work, we introduce a framework that addresses these limitations by establishing conditions under which unique solutions can be guaranteed. To illustrate, we apply it to examples from biological systems and ecological dynamics, demonstrating accurate and interpretable results. Our approach significantly enhances the potential of machine learning techniques in modeling complex systems in science and engineering.
Authors: Jiachen Yao, Abbas Mammadov, Julius Berner, Gavin Kerrigan, Jong Chul Ye, Kamyar Azizzadenesheli, Anima Anandkumar
Abstract: We propose a general framework for conditional sampling in PDE-based inverse problems, targeting the recovery of whole solutions from extremely sparse or noisy measurements. This is accomplished by a function-space diffusion model and plug-and-play guidance for conditioning. Our method first trains an unconditional discretization-agnostic denoising model using neural operator architectures. At inference, we refine the samples to satisfy sparse observation data via a gradient-based guidance mechanism. Through rigorous mathematical analysis, we extend Tweedie's formula to infinite-dimensional Hilbert spaces, providing the theoretical foundation for our posterior sampling approach. Our method (FunDPS) accurately captures posterior distributions in function spaces under minimal supervision and severe data scarcity. Across five PDE tasks with only 3% observation, our method achieves an average 32% accuracy improvement over state-of-the-art fixed-resolution diffusion baselines while reducing sampling steps by 4x. Furthermore, multi-resolution fine-tuning ensures strong cross-resolution generalizability. To the best of our knowledge, this is the first diffusion-based framework to operate independently of discretization, offering a practical and flexible solution for forward and inverse problems in the context of PDEs. Code is available at https://github.com/neuraloperator/FunDPS
Authors: Tim Genewein, Kevin Wenliang Li, Jordi Grau-Moya, Anian Ruoss, Laurent Orseau, Marcus Hutter
Abstract: Prompting is one of the main ways to adapt a pretrained model to target tasks. Besides manually constructing prompts, many prompt optimization methods have been proposed in the literature. Method development is mainly empirically driven, with less emphasis on a conceptual understanding of prompting. In this paper we discuss how optimal prompting can be understood through a Bayesian view, which also implies some fundamental limitations of prompting that can only be overcome by tuning weights. The paper explains in detail how meta-trained neural networks behave as Bayesian predictors over the pretraining distribution, whose hallmark feature is rapid in-context adaptation. Optimal prompting can be studied formally as conditioning these Bayesian predictors, yielding criteria for target tasks where optimal prompting is and is not possible. We support the theory with educational experiments on LSTMs and Transformers, where we compare different versions of prefix-tuning and different weight-tuning methods. We also confirm that soft prefixes, which are sequences of real-valued vectors outside the token alphabet, can lead to very effective prompts for trained and even untrained networks by manipulating activations in ways that are not achievable by hard tokens. This adds an important mechanistic aspect beyond the conceptual Bayesian theory.
Authors: Kevin Lu, Nicky Kriplani, Rohit Gandikota, Minh Pham, David Bau, Chinmay Hegde, Niv Cohen
Abstract: Concept erasure, the ability to selectively prevent a model from generating specific concepts, has attracted growing interest, with various approaches emerging to address the challenge. However, it remains unclear how thoroughly these methods erase the target concept. We begin by proposing two conceptual models for the erasure mechanism in diffusion models: (i) reducing the likelihood of generating the target concept, and (ii) interfering with the model's internal guidance mechanisms. To thoroughly assess whether a concept has been truly erased from the model, we introduce a suite of independent evaluations. Our evaluation framework includes adversarial attacks, novel probing techniques, and analysis of the model's alternative generations in place of the erased concept. Our results shed light on the tension between minimizing side effects and maintaining robustness to adversarial prompts. Broadly, our work underlines the importance of comprehensive evaluation for erasure in diffusion models.
Authors: Shuhan Tan, Kairan Dou, Yue Zhao, Philipp Kr\"ahenb\"uhl
Abstract: We introduce RIPT-VLA, a simple and scalable reinforcement-learning-based interactive post-training paradigm that fine-tunes pretrained Vision-Language-Action (VLA) models using only sparse binary success rewards. Existing VLA training pipelines rely heavily on offline expert demonstration data and supervised imitation, limiting their ability to adapt to new tasks and environments under low-data regimes. RIPT-VLA addresses this by enabling interactive post-training with a stable policy optimization algorithm based on dynamic rollout sampling and leave-one-out advantage estimation. RIPT-VLA has the following characteristics. First, it applies to various VLA models, resulting in an improvement on the lightweight QueST model by 21.2%, and the 7B OpenVLA-OFT model to an unprecedented 97.5% success rate. Second, it is computationally efficient and data-efficient: with only one demonstration, RIPT-VLA enables an unworkable SFT model (4%) to succeed with a 97% success rate within 15 iterations. Furthermore, we demonstrate that the policy learned by RIPT-VLA generalizes across different tasks and scenarios and is robust to the initial state context. These results highlight RIPT-VLA as a practical and effective paradigm for post-training VLA models through minimal supervision.
Authors: Zhaohui Yang, Shilei Jiang, Chen Hu, Linjing Li, Shihong Deng, Daxin Jiang
Abstract: Recent advances in reasoning language models have witnessed a paradigm shift from short to long CoT pattern. Given the substantial computational cost of rollouts in long CoT models, maximizing the utility of fixed training datasets becomes crucial. Our analysis reveals that negative responses contain valuable components such as self-reflection and error-correction steps, yet primary existing methods either completely discard negative samples (RFT) or apply equal penalization across all tokens (RL), failing to leverage these potential learning signals. In light of this, we propose Behavior Constrained Policy Gradient with Negative Sample Augmentation (BCPG-NSA), a fine-grained offline RL framework that encompasses three stages: 1) sample segmentation, 2) consensus-based step correctness assessment combining LLM and PRM judgers, and 3) policy optimization with NSA designed to effectively mine positive steps within negative samples. Experimental results show that BCPG-NSA outperforms baselines on several challenging math/coding reasoning benchmarks using the same training dataset, achieving improved sample efficiency and demonstrating robustness and scalability when extended to multiple iterations.
Authors: Xiaojie Gu, Guangxu Chen, Jungang Li, Jia-Chen Gu, Xuming Hu, Kai Zhang
Abstract: Lifelong learning enables large language models (LLMs) to adapt to evolving information by continually updating their internal knowledge. An ideal system should support efficient, wide-ranging updates while preserving existing capabilities and ensuring reliable deployment. Model editing stands out as a promising solution for this goal, offering a focused and efficient way to revise a model's internal knowledge. Although recent paradigms have made notable progress, they often struggle to meet the demands of practical lifelong adaptation at scale. To bridge this gap, we propose ULTRAEDIT-a fundamentally new editing solution that is training-, subject- and memory-free, making it particularly well-suited for ultra-scalable, real-world lifelong model editing. ULTRAEDIT performs editing through a self-contained process that relies solely on lightweight linear algebra operations to compute parameter shifts, enabling fast and consistent parameter modifications with minimal overhead. To improve scalability in lifelong settings, ULTRAEDIT employs a lifelong normalization strategy that continuously updates feature statistics across turns, allowing it to adapt to distributional shifts and maintain consistency over time. ULTRAEDIT achieves editing speeds over 7x faster than the previous state-of-the-art method-which was also the fastest known approach-while consuming less than 1/3 the VRAM, making it the only method currently capable of editing a 7B LLM on a 24GB consumer-grade GPU. Furthermore, we construct ULTRAEDITBENCH-the largest dataset in the field to date, with over 2M editing pairs-and demonstrate that our method supports up to 1M edits while maintaining high accuracy. Comprehensive experiments on four datasets and six models show that ULTRAEDIT consistently achieves superior performance across diverse model editing scenarios. Our code is available at: https://github.com/XiaojieGu/UltraEdit.
Authors: Jhon Lopez, Carlos Hinojosa, Henry Arguello, Bernard Ghanem
Abstract: The task of inverting real images into StyleGAN's latent space to manipulate their attributes has been extensively studied. However, existing GAN inversion methods struggle to balance high reconstruction quality, effective editability, and computational efficiency. In this paper, we introduce MambaStyle, an efficient single-stage encoder-based approach for GAN inversion and editing that leverages vision state-space models (VSSMs) to address these challenges. Specifically, our approach integrates VSSMs within the proposed architecture, enabling high-quality image inversion and flexible editing with significantly fewer parameters and reduced computational complexity compared to state-of-the-art methods. Extensive experiments show that MambaStyle achieves a superior balance among inversion accuracy, editing quality, and computational efficiency. Notably, our method achieves superior inversion and editing results with reduced model complexity and faster inference, making it suitable for real-time applications.
Authors: Mathias Schmolli, Maximilian Baronig, Robert Legenstein, Ozan \"Ozdenizci
Abstract: Deployment of deep neural networks in resource-constrained embedded systems requires innovative algorithmic solutions to facilitate their energy and memory efficiency. To further ensure the reliability of these systems against malicious actors, recent works have extensively studied adversarial robustness of existing architectures. Our work focuses on the intersection of adversarial robustness, memory- and energy-efficiency in neural networks. We introduce a neural network conversion algorithm designed to produce sparse and adversarially robust spiking neural networks (SNNs) by leveraging the sparse connectivity and weights from a robustly pretrained artificial neural network (ANN). Our approach combines the energy-efficient architecture of SNNs with a novel conversion algorithm, leading to state-of-the-art performance with enhanced energy and memory efficiency through sparse connectivity and activations. Our models are shown to achieve up to 100x reduction in the number of weights to be stored in memory, with an estimated 8.6x increase in energy efficiency compared to dense SNNs, while maintaining high performance and robustness against adversarial threats.
Authors: Aarav Lala, Kalyan Cherukuri
Abstract: As artificial intelligence continues to drive innovation in complex, decentralized environments, the need for scalable, adaptive, and privacy-preserving decision-making systems has become critical. This paper introduces a novel framework combining quantum-inspired neural networks with evolutionary algorithms to optimize real-time decision-making in multi-agent systems (MAS). The proposed Quantum-Evolutionary Neural Network (QE-NN) leverages quantum computing principles -- such as quantum superposition and entanglement -- to enhance learning speed and decision accuracy, while integrating evolutionary optimization to continually refine agent behaviors in dynamic, uncertain environments. By utilizing federated learning, QE-NN ensures privacy preservation, enabling decentralized agents to collaborate without sharing sensitive data. The framework is designed to allow agents to adapt in real-time to their environments, optimizing decision-making processes for applications in areas such as autonomous systems, smart cities, and healthcare. This research represents a breakthrough in merging quantum computing, evolutionary optimization, and privacy-preserving techniques to solve complex problems in multi-agent decision-making systems, pushing the boundaries of AI in real-world, privacy-sensitive applications.
Authors: Saining Liu, Yi Mei, Mengjie Zhang
Abstract: Manually designing (meta-)heuristics for the Vehicle Routing Problem (VRP) is a challenging task that requires significant domain expertise. Recently, data-driven approaches have emerged as a promising solution, automatically learning heuristics that perform well on training instances and generalize to unseen test cases. Such an approach learns (meta-)heuristics that can perform well on the training instances, expecting it to generalize well on the unseen test instances. A recent method, named GPGLS, uses Genetic Programming (GP) to learn the utility function in Guided Local Search (GLS) and solved large scale VRP effectively. However, the selection of appropriate training instances during the learning process remains an open question, with most existing studies including GPGLS relying on random instance selection. To address this, we propose a novel method, CL-GPGLS, which integrates Curriculum Learning (CL) into GPGLS. Our approach leverages a predefined curriculum to introduce training instances progressively, starting with simpler tasks and gradually increasing complexity, enabling the model to better adapt and optimize for large-scale VRP (LSVRP). Extensive experiments verify the effectiveness of CL-GPGLS, demonstrating significant performance improvements over three baseline methods.
Authors: Mohit Kataria, Shreyash Bhilwade, Sandeep Kumar, Jayadeva
Abstract: $\textbf{Graph Coarsening (GC)}$ is a prominent graph reduction technique that compresses large graphs to enable efficient learning and inference. However, existing GC methods generate only one coarsened graph per run and must recompute from scratch for each new coarsening ratio, resulting in unnecessary overhead. Moreover, most prior approaches are tailored to $\textit{homogeneous}$ graphs and fail to accommodate the semantic constraints of $\textit{heterogeneous}$ graphs, which comprise multiple node and edge types. To overcome these limitations, we introduce a novel framework that combines Locality Sensitive Hashing (LSH) with Consistent Hashing to enable $\textit{adaptive graph coarsening}$. Leveraging hashing techniques, our method is inherently fast and scalable. For heterogeneous graphs, we propose a $\textit{type isolated coarsening}$ strategy that ensures semantic consistency by restricting merges to nodes of the same type. Our approach is the first unified framework to support both adaptive and heterogeneous coarsening. Extensive evaluations on 23 real-world datasets including homophilic, heterophilic, homogeneous, and heterogeneous graphs demonstrate that our method achieves superior scalability while preserving the structural and semantic integrity of the original graph.
Authors: Yousuf Islam, Md. Jalal Uddin Chowdhury, Sumon Chandra Das
Abstract: Brain stroke remains one of the principal causes of death and disability worldwide, yet most tabular-data prediction models still hover below the 95% accuracy threshold, limiting real-world utility. Addressing this gap, the present work develops and validates a completely data-driven and interpretable machine-learning framework designed to predict strokes using ten routinely gathered demographic, lifestyle, and clinical variables sourced from a public cohort of 4,981 records. We employ a detailed exploratory data analysis (EDA) to understand the dataset's structure and distribution, followed by rigorous data preprocessing, including handling missing values, outlier removal, and class imbalance correction using Synthetic Minority Over-sampling Technique (SMOTE). To streamline feature selection, point-biserial correlation and random-forest Gini importance were utilized, and ten varied algorithms-encompassing tree ensembles, boosting, kernel methods, and a multilayer neural network-were optimized using stratified five-fold cross-validation. Their predictions based on probabilities helped us build the proposed model, which included Random Forest, XGBoost, LightGBM, and a support-vector classifier, with logistic regression acting as a meta-learner. The proposed model achieved an accuracy rate of 97.2% and an F1-score of 97.15%, indicating a significant enhancement compared to the leading individual model, LightGBM, which had an accuracy of 91.4%. Our study's findings indicate that rigorous preprocessing, coupled with a diverse hybrid model, can convert low-cost tabular data into a nearly clinical-grade stroke-risk assessment tool.
Authors: Bla\v{z} Bertalani\v{c}, Matej Vnu\v{c}ec, Carolina Fortuna
Abstract: In today's world, modern infrastructures are being equipped with information and communication technologies to create large IoT networks. It is essential to monitor these networks to ensure smooth operations by detecting and correcting link failures or abnormal network behaviour proactively, which can otherwise cause interruptions in business operations. This paper presents a novel method for detecting anomalies in wireless links using graph neural networks. The proposed approach involves converting time series data into graphs and training a new graph neural network architecture based on graph attention networks that successfully detects anomalies at the level of individual measurements of the time series data. The model provides competitive results compared to the state of the art while being computationally more efficient with ~171 times fewer trainable parameters.
Authors: Thai-Hoc Vu, Ngo Hoang Tu, Thien Huynh-The, Kyungchun Lee, Sunghwan Kim, Miroslav Voznak, Quoc-Viet Pham
Abstract: The transition from 5G networks to 6G highlights a significant demand for machine learning (ML). Deep learning models, in particular, have seen wide application in mobile networking and communications to support advanced services in emerging wireless environments, such as smart healthcare, smart grids, autonomous vehicles, aerial platforms, digital twins, and the metaverse. The rapid expansion of Internet-of-Things (IoT) devices, many with limited computational capabilities, has accelerated the development of tiny machine learning (TinyML) and resource-efficient ML approaches for cost-effective services. However, the deployment of large-scale machine learning (LargeML) solutions require major computing resources and complex management strategies to support extensive IoT services and ML-generated content applications. Consequently, the integration of TinyML and LargeML is projected as a promising approach for future seamless connectivity and efficient resource management. Although the integration of TinyML and LargeML shows abundant potential, several challenges persist, including performance optimization, practical deployment strategies, effective resource management, and security considerations. In this survey, we review and analyze the latest research aimed at enabling the integration of TinyML and LargeML models for the realization of smart services and applications in future 6G networks and beyond. The paper concludes by outlining critical challenges and identifying future research directions for the holistic integration of TinyML and LargeML in next-generation wireless networks.
Authors: Stephen Asiedu, David Watson
Abstract: Causal discovery in multi-omic datasets is crucial for understanding the bigger picture of gene regulatory mechanisms, but remains challenging due to high dimensionality, differentiation of direct from indirect relationships, and hidden confounders. We introduce GENESIS (GEne Network inference from Expression SIgnals and SNPs), a constraint-based algorithm that leverages the natural causal precedence of genotypes to infer ancestral relationships in transcriptomic data. Unlike traditional causal discovery methods that start with a fully connected graph, GENESIS initialises an empty ancestrality matrix and iteratively populates it with direct, indirect or non-causal relationships using a series of provably sound marginal and conditional independence tests. By integrating genotypes as fixed causal anchors, GENESIS provides a principled ``head start'' to classical causal discovery algorithms, restricting the search space to biologically plausible edges. We test GENESIS on synthetic and real-world genomic datasets. This framework offers a powerful avenue for uncovering causal pathways in complex traits, with promising applications to functional genomics, drug discovery, and precision medicine.
Authors: Nikolaos Chaidos, Angeliki Dimitriou, Maria Lymperaiou, Giorgos Stamou
Abstract: Despite the dominance of convolutional and transformer-based architectures in image-to-image retrieval, these models are prone to biases arising from low-level visual features, such as color. Recognizing the lack of semantic understanding as a key limitation, we propose a novel scene graph-based retrieval framework that emphasizes semantic content over superficial image characteristics. Prior approaches to scene graph retrieval predominantly rely on supervised Graph Neural Networks (GNNs), which require ground truth graph pairs driven from image captions. However, the inconsistency of caption-based supervision stemming from variable text encodings undermine retrieval reliability. To address these, we present SCENIR, a Graph Autoencoder-based unsupervised retrieval framework, which eliminates the dependence on labeled training data. Our model demonstrates superior performance across metrics and runtime efficiency, outperforming existing vision-based, multimodal, and supervised GNN approaches. We further advocate for Graph Edit Distance (GED) as a deterministic and robust ground truth measure for scene graph similarity, replacing the inconsistent caption-based alternatives for the first time in image-to-image retrieval evaluation. Finally, we validate the generalizability of our method by applying it to unannotated datasets via automated scene graph generation, while substantially contributing in advancing state-of-the-art in counterfactual image retrieval.
Authors: Siting Li, Xiang Gao, Simon Shaolei Du
Abstract: While an image is worth more than a thousand words, only a few provide crucial information for a given task and thus should be focused on. In light of this, ideal text-to-image (T2I) retrievers should prioritize specific visual attributes relevant to queries. To evaluate current retrievers on handling attribute-focused queries, we build COCO-Facet, a COCO-based benchmark with 9,112 queries about diverse attributes of interest. We find that CLIP-like retrievers, which are widely adopted due to their efficiency and zero-shot ability, have poor and imbalanced performance, possibly because their image embeddings focus on global semantics and subjects while leaving out other details. Notably, we reveal that even recent Multimodal Large Language Model (MLLM)-based, stronger retrievers with a larger output dimension struggle with this limitation. Hence, we hypothesize that retrieving with general image embeddings is suboptimal for performing such queries. As a solution, we propose to use promptable image embeddings enabled by these multimodal retrievers, which boost performance by highlighting required attributes. Our pipeline for deriving such embeddings generalizes across query types, image pools, and base retriever architectures. To enhance real-world applicability, we offer two acceleration strategies: Pre-processing promptable embeddings and using linear approximations. We show that the former yields a 15% improvement in Recall@5 when prompts are predefined, while the latter achieves an 8% improvement when prompts are only available during inference.
Authors: Awni Altabaa, Omar Montasser, John Lafferty
Abstract: Learning complex functions that involve multi-step reasoning poses a significant challenge for standard supervised learning from input-output examples. Chain-of-thought (CoT) supervision, which provides intermediate reasoning steps together with the final output, has emerged as a powerful empirical technique, underpinning much of the recent progress in the reasoning capabilities of large language models. This paper develops a statistical theory of learning under CoT supervision. A key characteristic of the CoT setting, in contrast to standard supervision, is the mismatch between the training objective (CoT risk) and the test objective (end-to-end risk). A central part of our analysis, distinguished from prior work, is explicitly linking those two types of risk to achieve sharper sample complexity bounds. This is achieved via the *CoT information measure* $\mathcal{I}_{\mathcal{D}, h_\star}^{\mathrm{CoT}}(\epsilon; \calH)$, which quantifies the additional discriminative power gained from observing the reasoning process. The main theoretical results demonstrate how CoT supervision can yield significantly faster learning rates compared to standard E2E supervision. Specifically, it is shown that the sample complexity required to achieve a target E2E error $\epsilon$ scales as $d/\mathcal{I}_{\mathcal{D}, h_\star}^{\mathrm{CoT}}(\epsilon; \calH)$, where $d$ is a measure of hypothesis class complexity, which can be much faster than standard $d/\epsilon$ rates. Information-theoretic lower bounds in terms of the CoT information are also obtained. Together, these results suggest that CoT information is a fundamental measure of statistical complexity for learning under chain-of-thought supervision.
Authors: Ahmed Bouajjani, Wael-Amine Boutglay, Peter Habermehl
Abstract: We address the problem of verifying automatically procedural programs manipulating parametric-size arrays of integers, encoded as a constrained Horn clauses solving problem. We propose a new algorithmic method for synthesizing loop invariants and procedure pre/post-conditions represented as universally quantified first-order formulas constraining the array elements and program variables. We adopt a data-driven approach that extends the decision tree Horn-ICE framework to handle arrays. We provide a powerful learning technique based on reducing a complex classification problem of vectors of integer arrays to a simpler classification problem of vectors of integers. The obtained classifier is generalized to get universally quantified invariants and procedure pre/post-conditions. We have implemented our method and shown its efficiency and competitiveness w.r.t. state-of-the-art tools on a significant benchmark.
Authors: Linxi Zhao, Sofian Zalouk, Christian K. Belardi, Justin Lovelace, Jin Peng Zhou, Kilian Q. Weinberger, Yoav Artzi, Jennifer J. Sun
Abstract: Neural language models are black-boxes -- both linguistic patterns and factual knowledge are distributed across billions of opaque parameters. This entangled encoding makes it difficult to reliably inspect, verify, or update specific facts. We propose a new class of language models, Large Memory Language Models (LMLM) with a pre-training recipe that stores factual knowledge in both internal weights and an external database. Our approach strategically masks externally retrieved factual values from the training loss, thereby teaching the model to perform targeted lookups rather than relying on memorization in model weights. Our experiments demonstrate that LMLMs achieve competitive performance compared to significantly larger, knowledge-dense LLMs on standard benchmarks, while offering the advantages of explicit, editable, and verifiable knowledge bases. This work represents a fundamental shift in how language models interact with and manage factual knowledge.
Authors: Matthew Lyle Olson, Musashi Hinck, Neale Ratzlaff, Changbai Li, Phillip Howard, Vasudev Lal, Shao-Yen Tseng
Abstract: The ImageNet hierarchy provides a structured taxonomy of object categories, offering a valuable lens through which to analyze the representations learned by deep vision models. In this work, we conduct a comprehensive analysis of how vision models encode the ImageNet hierarchy, leveraging Sparse Autoencoders (SAEs) to probe their internal representations. SAEs have been widely used as an explanation tool for large language models (LLMs), where they enable the discovery of semantically meaningful features. Here, we extend their use to vision models to investigate whether learned representations align with the ontological structure defined by the ImageNet taxonomy. Our results show that SAEs uncover hierarchical relationships in model activations, revealing an implicit encoding of taxonomic structure. We analyze the consistency of these representations across different layers of the popular vision foundation model DINOv2 and provide insights into how deep vision models internalize hierarchical category information by increasing information in the class token through each layer. Our study establishes a framework for systematic hierarchical analysis of vision model representations and highlights the potential of SAEs as a tool for probing semantic structure in deep networks.
Authors: Alan Ta, Nilsu Salgin, Mustafa Demir, Kala Philips Randal, Ranjana K. Mehta, Anthony McDonald, Carly McCord, Farzan Sasangohar
Abstract: College students are increasingly affected by stress, anxiety, and depression, yet face barriers to traditional mental health care. This study evaluated the efficacy of a mobile health (mHealth) intervention, Mental Health Evaluation and Lookout Program (mHELP), which integrates a smartwatch sensor and machine learning (ML) algorithms for real-time stress detection and self-management. In a 12-week randomized controlled trial (n = 117), participants were assigned to a treatment group using mHELP's full suite of interventions or a control group using the app solely for real-time stress logging and weekly psychological assessments. The primary outcome, "Moments of Stress" (MS), was assessed via physiological and self-reported indicators and analyzed using Generalized Linear Mixed Models (GLMM) approaches. Similarly, secondary outcomes of psychological assessments, including the Generalized Anxiety Disorder-7 (GAD-7) for anxiety, the Patient Health Questionnaire (PHQ-8) for depression, and the Perceived Stress Scale (PSS), were also analyzed via GLMM. The finding of the objective measure, MS, indicates a substantial decrease in MS among the treatment group compared to the control group, while no notable between-group differences were observed in subjective scores of anxiety (GAD-7), depression (PHQ-8), or stress (PSS). However, the treatment group exhibited a clinically meaningful decline in GAD-7 and PSS scores. These findings underscore the potential of wearable-enabled mHealth tools to reduce acute stress in college populations and highlight the need for extended interventions and tailored features to address chronic symptoms like depression.
Authors: Yamin Arefeen, Brett Levac, Bhairav Patel, Chang Ho, Jonathan I. Tamir
Abstract: Purpose: Magnetic Resonance Imaging (MRI) enables non-invasive assessment of brain abnormalities during early life development. Permanent magnet scanners operating in the neonatal intensive care unit (NICU) facilitate MRI of sick infants, but have long scan times due to lower signal-to-noise ratios (SNR) and limited receive coils. This work accelerates in-NICU MRI with diffusion probabilistic generative models by developing a training pipeline accounting for these challenges. Methods: We establish a novel training dataset of clinical, 1 Tesla neonatal MR images in collaboration with Aspect Imaging and Sha'are Zedek Medical Center. We propose a pipeline to handle the low quantity and SNR of our real-world dataset (1) modifying existing network architectures to support varying resolutions; (2) training a single model on all data with learned class embedding vectors; (3) applying self-supervised denoising before training; and (4) reconstructing by averaging posterior samples. Retrospective under-sampling experiments, accounting for signal decay, evaluated each item of our proposed methodology. A clinical reader study with practicing pediatric neuroradiologists evaluated our proposed images reconstructed from 1.5x under-sampled data. Results: Combining all data, denoising pre-training, and averaging posterior samples yields quantitative improvements in reconstruction. The generative model decouples the learned prior from the measurement model and functions at two acceleration rates without re-training. The reader study suggests that proposed images reconstructed from approximately 1.5x under-sampled data are adequate for clinical use. Conclusion: Diffusion probabilistic generative models applied with the proposed pipeline to handle challenging real-world datasets could reduce scan time of in-NICU neonatal MRI.
Authors: Wei Liu, Siya Qi, Xinyu Wang, Chen Qian, Yali Du, Yulan He
Abstract: Recent advances such as DeepSeek R1-Zero highlight the effectiveness of incentive training, a reinforcement learning paradigm that computes rewards solely based on the final answer part of a language model's output, thereby encouraging the generation of intermediate reasoning steps. However, these methods fundamentally rely on external verifiers, which limits their applicability to domains like mathematics and coding where such verifiers are readily available. Although reward models can serve as verifiers, they require high-quality annotated data and are costly to train. In this work, we propose NOVER, NO-VERifier Reinforcement Learning, a general reinforcement learning framework that requires only standard supervised fine-tuning data with no need for an external verifier. NOVER enables incentive training across a wide range of text-to-text tasks and outperforms the model of the same size distilled from large reasoning models such as DeepSeek R1 671B by 7.7 percent. Moreover, the flexibility of NOVER enables new possibilities for optimizing large language models, such as inverse incentive training.
Authors: Asterios Tsiourvas, Wei Sun, Georgia Perakis
Abstract: LLM routing aims to select the most appropriate model for each query, balancing competing performance metrics such as accuracy and cost across a pool of language models. Prior approaches typically adopt a decoupled strategy, where the metrics are first predicted and the model is then selected based on these estimates. This setup is prone to compounding errors and often relies on full-feedback data, where each query is evaluated by all candidate models, which is costly to obtain and maintain in practice. In contrast, we learn from observational data, which records only the outcome of the model actually deployed. We propose a causal end-to-end framework that learns routing policies by minimizing decision-making regret from observational data. To enable efficient optimization, we introduce two theoretically grounded surrogate objectives: a classification-based upper bound, and a softmax-weighted regret approximation shown to recover the optimal policy at convergence. We further extend our framework to handle heterogeneous cost preferences via an interval-conditioned architecture. Experiments on public benchmarks show that our method outperforms existing baselines, achieving state-of-the-art performance across different embedding models.
Authors: Siddhant Agarwal, Ali Can Bekar, Christian H\"uttig, David S. Greenberg, Nicola Tosi
Abstract: Mantle convection simulations are an essential tool for understanding how rocky planets evolve. However, the poorly known input parameters to these simulations, the non-linear dependence of transport properties on pressure and temperature, and the long integration times in excess of several billion years all pose a computational challenge for numerical solvers. We propose a physics-based machine learning approach that predicts creeping flow velocities as a function of temperature while conserving mass, thereby bypassing the numerical solution of the Stokes problem. A finite-volume solver then uses the predicted velocities to advect and diffuse the temperature field to the next time-step, enabling autoregressive rollout at inference. For training, our model requires temperature-velocity snapshots from a handful of simulations (94). We consider mantle convection in a two-dimensional rectangular box with basal and internal heating, pressure- and temperature-dependent viscosity. Overall, our model is up to 89 times faster than the numerical solver. We also show the importance of different components in our convolutional neural network architecture such as mass conservation, learned paddings on the boundaries, and loss scaling for the overall rollout performance. Finally, we test our approach on unseen scenarios to demonstrate some of its strengths and weaknesses.
Authors: Gowtham Premananth, Philip Resnik, Sonia Bansal, Deanna L. Kelly, Carol Espy-Wilson
Abstract: Studies on schizophrenia assessments using deep learning typically treat it as a classification task to detect the presence or absence of the disorder, oversimplifying the condition and reducing its clinical applicability. This traditional approach overlooks the complexity of schizophrenia, limiting its practical value in healthcare settings. This study shifts the focus to individual symptom severity estimation using a multimodal approach that integrates speech, video, and text inputs. We develop unimodal models for each modality and a multimodal framework to improve accuracy and robustness. By capturing a more detailed symptom profile, this approach can help in enhancing diagnostic precision and support personalized treatment, offering a scalable and objective tool for mental health assessment.
Authors: Dongze Wu, David I. Inouye, Yao Xie
Abstract: We propose PO-Flow, a novel continuous normalizing flow (CNF) framework for causal inference that jointly models potential outcomes and counterfactuals. Trained via flow matching, PO-Flow provides a unified framework for individualized potential outcome prediction, counterfactual predictions, and uncertainty-aware density learning. Among generative models, it is the first to enable density learning of potential outcomes without requiring explicit distributional assumptions (e.g., Gaussian mixtures), while also supporting counterfactual prediction conditioned on factual outcomes in general observational datasets. On benchmarks such as ACIC, IHDP, and IBM, it consistently outperforms prior methods across a range of causal inference tasks. Beyond that, PO-Flow succeeds in high-dimensional settings, including counterfactual image generation, demonstrating its broad applicability.
Authors: Renato Berlinghieri, Yunyi Shen, Jialong Jiang, Tamara Broderick
Abstract: Scientists often want to make predictions beyond the observed time horizon of "snapshot" data following latent stochastic dynamics. For example, in time course single-cell mRNA profiling, scientists have access to cellular transcriptional state measurements (snapshots) from different biological replicates at different time points, but they cannot access the trajectory of any one cell because measurement destroys the cell. Researchers want to forecast (e.g.) differentiation outcomes from early state measurements of stem cells. Recent Schr\"odinger-bridge (SB) methods are natural for interpolating between snapshots. But past SB papers have not addressed forecasting -- likely since existing methods either (1) reduce to following pre-set reference dynamics (chosen before seeing data) or (2) require the user to choose a fixed, state-independent volatility since they minimize a Kullback-Leibler divergence. Either case can lead to poor forecasting quality. In the present work, we propose a new framework, SnapMMD, that learns dynamics by directly fitting the joint distribution of both state measurements and observation time with a maximum mean discrepancy (MMD) loss. Unlike past work, our method allows us to infer unknown and state-dependent volatilities from the observed data. We show in a variety of real and synthetic experiments that our method delivers accurate forecasts. Moreover, our approach allows us to learn in the presence of incomplete state measurements and yields an $R^2$-style statistic that diagnoses fit. We also find that our method's performance at interpolation (and general velocity-field reconstruction) is at least as good as (and often better than) state-of-the-art in almost all of our experiments.
Authors: Damien Ferbach, Katie Everett, Gauthier Gidel, Elliot Paquette, Courtney Paquette
Abstract: We investigate scaling laws for stochastic momentum algorithms with small batch on the power law random features model, parameterized by data complexity, target complexity, and model size. When trained with a stochastic momentum algorithm, our analysis reveals four distinct loss curve shapes determined by varying data-target complexities. While traditional stochastic gradient descent with momentum (SGD-M) yields identical scaling law exponents to SGD, dimension-adapted Nesterov acceleration (DANA) improves these exponents by scaling momentum hyperparameters based on model size and data complexity. This outscaling phenomenon, which also improves compute-optimal scaling behavior, is achieved by DANA across a broad range of data and target complexities, while traditional methods fall short. Extensive experiments on high-dimensional synthetic quadratics validate our theoretical predictions and large-scale text experiments with LSTMs show DANA's improved loss exponents over SGD hold in a practical setting.
Authors: Yue Li, Xin Yi, Dongsheng Shi, Gerard de Melo, Xiaoling Wang, Linlin Wang
Abstract: With the increasing size of Large Vision-Language Models (LVLMs), network pruning techniques aimed at compressing models for deployment in resource-constrained environments have garnered significant attention. However, we observe that pruning often leads to a degradation in safety performance. To address this issue, we present a novel and lightweight approach, termed Hierarchical Safety Realignment (HSR). HSR operates by first quantifying the contribution of each attention head to safety, identifying the most critical ones, and then selectively restoring neurons directly within these attention heads that play a pivotal role in maintaining safety. This process hierarchically realigns the safety of pruned LVLMs, progressing from the attention head level to the neuron level. We validate HSR across various models and pruning strategies, consistently achieving notable improvements in safety performance. To our knowledge, this is the first work explicitly focused on restoring safety in LVLMs post-pruning.
Authors: Nathan Brady, David Tennyson, Thomas Vandermeulen
Abstract: In this paper, we apply both supervised and unsupervised machine learning algorithms to the study of the string landscape and swampland in 6-dimensions. Our data are the (almost) anomaly-free 6-dimensional $\mathcal{N} = (1,0)$ supergravity models, characterised by the Gram matrix of anomaly coefficients. Our work demonstrates the ability of machine learning algorithms to efficiently learn highly complex features of the landscape and swampland. Employing an autoencoder for unsupervised learning, we provide an auto-classification of these models by compressing the Gram matrix data to 2-dimensions. Through compression, similar models cluster together, and we identify prominent features of these clusters. The autoencoder also identifies outlier models which are difficult to reconstruct. One of these outliers proves to be incredibly difficult to combine with other models such that the $\text{tr}R^{4}$ anomaly vanishes, making its presence in the landscape extremely rare. Further, we utilise supervised learning to build two classifiers predicting (1) model consistency under probe string insertion (precision: 0.78, predicting consistency for 214,837 models with reasonable certainty) and (2) inconsistency under anomaly inflow (precision: 0.91, predicting inconsistency for 1,909,359 models). Notably, projecting these predictions onto the autoencoder's 2-dimensional latent layer shows consistent models clustering together, further indicating that the autoencoder has learnt interesting and complex features of the set of models and potentially offers a novel approach to mapping the landscape and swampland of 6-dimensional supergravity theories.
Authors: Menschikov Mikhail, Alexander Kharitonov, Maiia Kotyga, Vadim Porvatov, Anna Zhukovskaya, David Kagramanyan, Egor Shvetsov, Evgeny Burnaev
Abstract: Large language models exhibit positional bias -- systematic neglect of information at specific context positions -- yet its interplay with linguistic diversity remains poorly understood. We present a cross-linguistic study across five typologically distinct languages (English, Russian, German, Hindi, Vietnamese), examining how positional bias interacts with model uncertainty, syntax, and prompting. Key findings: (1) Positional bias is model-driven, with language-specific variations -- Qwen2.5-7B favors late positions, challenging assumptions of early-token bias; (2) Explicit positional guidance (e.g., correct context is at position X) reduces accuracy across languages, undermining prompt-engineering practices; (3) Aligning context with positional bias increases entropy, yet minimal entropy does not predict accuracy. (4) We further uncover that LLMs differently impose dominant word order in free-word-order languages like Hindi.
Authors: Jeffrey Seely, Yuki Imajuku, Tianyu Zhao, Edoardo Cetin, Llion Jones
Abstract: Existing reasoning benchmarks for large language models (LLMs) frequently fail to capture authentic creativity, often rewarding memorization of previously observed patterns. We address this shortcoming with Sudoku-Bench, a curated benchmark of challenging and unconventional Sudoku variants specifically selected to evaluate creative, multi-step logical reasoning. Sudoku variants form an unusually effective domain for reasoning research: each puzzle introduces unique or subtly interacting constraints, making memorization infeasible and requiring solvers to identify novel logical breakthroughs (``break-ins''). Despite their diversity, Sudoku variants maintain a common and compact structure, enabling clear and consistent evaluation. Sudoku-Bench includes a carefully chosen puzzle set, a standardized text-based puzzle representation, and flexible tools compatible with thousands of publicly available puzzles -- making it easy to extend into a general research environment. Baseline experiments show that state-of-the-art LLMs solve fewer than 15\% of puzzles unaided, highlighting significant opportunities to advance long-horizon, strategic reasoning capabilities.
Authors: Yuke Zhang
Abstract: This study introduces an interpretable machine learning (ML) framework to extract macroeconomic alpha from global news sentiment. We process the Global Database of Events, Language, and Tone (GDELT) Project's worldwide news feed using FinBERT -- a Bidirectional Encoder Representations from Transformers (BERT) based model pretrained on finance-specific language -- to construct daily sentiment indices incorporating mean tone, dispersion, and event impact. These indices drive an XGBoost classifier, benchmarked against logistic regression, to predict next-day returns for EUR/USD, USD/JPY, and 10-year U.S. Treasury futures (ZN). Rigorous out-of-sample (OOS) backtesting (5-fold expanding-window cross-validation, OOS period: c. 2017-April 2025) demonstrates exceptional, cost-adjusted performance for the XGBoost strategy: Sharpe ratios achieve 5.87 (EUR/USD), 4.65 (USD/JPY), and 4.65 (Treasuries), with respective compound annual growth rates (CAGRs) exceeding 50% in Foreign Exchange (FX) and 22% in bonds. Shapley Additive Explanations (SHAP) affirm that sentiment dispersion and article impact are key predictive features. Our findings establish that integrating domain-specific Natural Language Processing (NLP) with interpretable ML offers a potent and explainable source of macro alpha.
Authors: Arghya Datta, Philippe Gagnon, Florian Maire
Abstract: Probabilistic principal component analysis (PCA) and its Bayesian variant (BPCA) are widely used for dimension reduction in machine learning and statistics. The main advantage of probabilistic PCA over the traditional formulation is allowing uncertainty quantification. The parameters of BPCA are typically learned using mean-field variational inference, and in particular, the coordinate ascent variational inference (CAVI) algorithm. So far, the convergence speed of CAVI for BPCA has not been characterized. In our paper, we fill this gap in the literature. Firstly, we prove a precise exponential convergence result in the case where the model uses a single principal component (PC). Interestingly, this result is established through a connection with the classical $\textit{power iteration algorithm}$ and it indicates that traditional PCA is retrieved as points estimates of the BPCA parameters. Secondly, we leverage recent tools to prove exponential convergence of CAVI for the model with any number of PCs, thus leading to a more general result, but one that is of a slightly different flavor. To prove the latter result, we additionally needed to introduce a novel lower bound for the symmetric Kullback--Leibler divergence between two multivariate normal distributions, which, we believe, is of independent interest in information theory.
Authors: Zhenglin Hua, Jinghan He, Zijun Yao, Tianxu Han, Haiyun Guo, Yuheng Jia, Junfeng Fang
Abstract: Large vision-language models (LVLMs) have achieved remarkable performance on multimodal tasks such as visual question answering (VQA) and image captioning. However, they still suffer from hallucinations, generating text inconsistent with visual input, posing significant risks in real-world applications. Existing approaches to address this issue focus on incorporating external knowledge bases, alignment training, or decoding strategies, all of which require substantial computational cost and time. Recent works try to explore more efficient alternatives by adjusting LVLMs' internal representations. Although promising, these methods may cause hallucinations to be insufficiently suppressed or lead to excessive interventions that negatively affect normal semantics. In this work, we leverage sparse autoencoders (SAEs) to identify semantic directions closely associated with either hallucinations or actuality, realizing more precise and direct hallucination-related representations. Our analysis demonstrates that interventions along the faithful direction we identified can mitigate hallucinations, while those along the hallucinatory direction can exacerbate them. Building on these insights, we propose Steering LVLMs via SAE Latent Directions (SSL), a training-free method based on SAE-derived latent directions to mitigate hallucinations in LVLMs. Extensive experiments demonstrate that SSL significantly outperforms existing decoding approaches in mitigating hallucinations, while maintaining transferability across different model architectures with negligible additional time overhead.
Authors: Siu Lun Chau, Michele Caprio, Krikamol Muandet
Abstract: Quantifying differences between probability distributions is fundamental to statistics and machine learning, primarily for comparing statistical uncertainty. In contrast, epistemic uncertainty (EU) -- due to incomplete knowledge -- requires richer representations than those offered by classical probability. Imprecise probability (IP) theory offers such models, capturing ambiguity and partial belief. This has driven growing interest in imprecise probabilistic machine learning (IPML), where inference and decision-making rely on broader uncertainty models -- highlighting the need for metrics beyond classical probability. This work introduces the Integral Imprecise Probability Metric (IIPM) framework, a Choquet integral-based generalisation of classical Integral Probability Metric (IPM) to the setting of capacities -- a broad class of IP models encompassing many existing ones, including lower probabilities, probability intervals, belief functions, and more. Theoretically, we establish conditions under which IIPM serves as a valid metric and metrises a form of weak convergence of capacities. Practically, IIPM not only enables comparison across different IP models but also supports the quantification of epistemic uncertainty within a single IP model. In particular, by comparing an IP model with its conjugate, IIPM gives rise to a new class of EU measures -- Maximum Mean Imprecision -- which satisfy key axiomatic properties proposed in the Uncertainty Quantification literature. We validate MMI through selective classification experiments, demonstrating strong empirical performance against established EU measures, and outperforming them when classical methods struggle to scale to a large number of classes. Our work advances both theory and practice in IPML, offering a principled framework for comparing and quantifying epistemic uncertainty under imprecision.
Authors: Zhi Zhong, Akira Takahashi, Shuyang Cui, Keisuke Toyama, Shusuke Takahashi, Yuki Mitsufuji
Abstract: Foley synthesis aims to synthesize high-quality audio that is both semantically and temporally aligned with video frames. Given its broad application in creative industries, the task has gained increasing attention in the research community. To avoid the non-trivial task of training audio generative models from scratch, adapting pretrained audio generative models for video-synchronized foley synthesis presents an attractive direction. ControlNet, a method for adding fine-grained controls to pretrained generative models, has been applied to foley synthesis, but its use has been limited to handcrafted human-readable temporal conditions. In contrast, from-scratch models achieved success by leveraging high-dimensional deep features extracted using pretrained video encoders. We have observed a performance gap between ControlNet-based and from-scratch foley models. To narrow this gap, we propose SpecMaskFoley, a method that steers the pretrained SpecMaskGIT model toward video-synchronized foley synthesis via ControlNet. To unlock the potential of a single ControlNet branch, we resolve the discrepancy between the temporal video features and the time-frequency nature of the pretrained SpecMaskGIT via a frequency-aware temporal feature aligner, eliminating the need for complicated conditioning mechanisms widely used in prior arts. Evaluations on a common foley synthesis benchmark demonstrate that SpecMaskFoley could even outperform strong from-scratch baselines, substantially advancing the development of ControlNet-based foley synthesis models. Demo page: https://zzaudio.github.io/SpecMaskFoley_Demo/
Authors: Anton Erofeev, Balasubramanya T. Nadiga, Ilya Timofeyev
Abstract: We apply the Echo-State Networks to predict the time series and statistical properties of the competitive Lotka-Volterra model in the chaotic regime. In particular, we demonstrate that Echo-State Networks successfully learn the chaotic attractor of the competitive Lotka-Volterra model and reproduce histograms of dependent variables, including tails and rare events. We use the Generalized Extreme Value distribution to quantify the tail behavior.
Authors: Md Ashraf Uddin, Nam H. Chu, Reza Rafeh, Mutaz Barika
Abstract: Due to its nature of dynamic, mobility, and wireless data transfer, the Internet of Vehicles (IoV) is prone to various cyber threats, ranging from spoofing and Distributed Denial of Services (DDoS) attacks to malware. To safeguard the IoV ecosystem from intrusions, malicious activities, policy violations, intrusion detection systems (IDS) play a critical role by continuously monitoring and analyzing network traffic to identify and mitigate potential threats in real-time. However, most existing research has focused on developing centralized, machine learning-based IDS systems for IoV without accounting for its inherently distributed nature. Due to intensive computing requirements, these centralized systems often rely on the cloud to detect cyber threats, increasing delay of system response. On the other hand, edge nodes typically lack the necessary resources to train and deploy complex machine learning algorithms. To address this issue, this paper proposes an effective hierarchical classification framework tailored for IoV networks. Hierarchical classification allows classifiers to be trained and tested at different levels, enabling edge nodes to detect specific types of attacks independently. With this approach, edge nodes can conduct targeted attack detection while leveraging cloud nodes for comprehensive threat analysis and support. Given the resource constraints of edge nodes, we have employed the Boruta feature selection method to reduce data dimensionality, optimizing processing efficiency. To evaluate our proposed framework, we utilize the latest IoV security dataset CIC-IoV2024, achieving promising results that demonstrate the feasibility and effectiveness of our models in securing IoV networks.
Authors: Sangyong Lee, Subo Hwang, Dohoon Kim
Abstract: In this paper, we propose MADCluster, a novel model-agnostic anomaly detection framework utilizing self-supervised clustering. MADCluster is applicable to various deep learning architectures and addresses the 'hypersphere collapse' problem inherent in existing deep learning-based anomaly detection methods. The core idea is to cluster normal pattern data into a 'single cluster' while simultaneously learning the cluster center and mapping data close to this center. Also, to improve expressiveness and enable effective single clustering, we propose a new 'One-directed Adaptive loss'. The optimization of this loss is mathematically proven. MADCluster consists of three main components: Base Embedder capturing high-dimensional temporal dynamics, Cluster Distance Mapping, and Sequence-wise Clustering for continuous center updates. Its model-agnostic characteristics are achieved by applying various architectures to the Base Embedder. Experiments on four time series benchmark datasets demonstrate that applying MADCluster improves the overall performance of comparative models. In conclusion, the compatibility of MADCluster shows potential for enhancing model performance across various architectures.
Authors: Bohao Wu, Qingyun Wang, Yue Guo
Abstract: Personalizing jargon detection and explanation is essential for making technical documents accessible to readers with diverse disciplinary backgrounds. However, tailoring models to individual users typically requires substantial annotation efforts and computational resources due to user-specific finetuning. To address this, we present a systematic study of personalized jargon detection, focusing on methods that are both efficient and scalable for real-world deployment. We explore two personalization strategies: (1) lightweight fine-tuning using Low-Rank Adaptation (LoRA) on open-source models, and (2) personalized prompting, which tailors model behavior at inference time without retaining. To reflect realistic constraints, we also investigate hybrid approaches that combine limited annotated data with unsupervised user background signals. Our personalized LoRA model outperforms GPT-4 by 21.4% in F1 score and exceeds the best performing oracle baseline by 8.3%. Remarkably, our method achieves comparable performance using only 10% of the annotated training data, demonstrating its practicality for resource-constrained settings. Our study offers the first work to systematically explore efficient, low-resource personalization of jargon detection using open-source language models, offering a practical path toward scalable, user-adaptive NLP system.
Authors: Masanari Kimura, Howard Bondell
Abstract: The power prior is a class of informative priors designed to incorporate historical data alongside current data in a Bayesian framework. It includes a power parameter that controls the influence of historical data, providing flexibility and adaptability. A key property of the power prior is that the resulting posterior minimizes a linear combination of KL divergences between two pseudo-posterior distributions: one ignoring historical data and the other fully incorporating it. We extend this framework by identifying the posterior distribution as the minimizer of a linear combination of Amari's $\alpha$-divergence, a generalization of KL divergence. We show that this generalization can lead to improved performance by allowing for the data to adapt to appropriate choices of the $\alpha$ parameter. Theoretical properties of this generalized power posterior are established, including behavior as a generalized geodesic on the Riemannian manifold of probability distributions, offering novel insights into its geometric interpretation.
Authors: Masanari Kimura
Abstract: Label shift adaptation aims to recover target class priors when the labelled source distribution $P$ and the unlabelled target distribution $Q$ share $P(X \mid Y) = Q(X \mid Y)$ but $P(Y) \neq Q(Y)$. Classical black-box shift estimators invert an empirical confusion matrix of a frozen classifier, producing a brittle point estimate that ignores sampling noise and similarity among classes. We present Graph-Smoothed Bayesian BBSE (GS-B$^3$SE), a fully probabilistic alternative that places Laplacian-Gaussian priors on both target log-priors and confusion-matrix columns, tying them together on a label-similarity graph. The resulting posterior is tractable with HMC or a fast block Newton-CG scheme. We prove identifiability, $N^{-1/2}$ contraction, variance bounds that shrink with the graph's algebraic connectivity, and robustness to Laplacian misspecification. We also reinterpret GS-B$^3$SE through information geometry, showing that it generalizes existing shift estimators.
Authors: Masanari Kimura
Abstract: This study develops a higher-order asymptotic framework for test-time adaptation (TTA) of Batch Normalization (BN) statistics under distribution shift by integrating classical Edgeworth expansion and saddlepoint approximation techniques with a novel one-step M-estimation perspective. By analyzing the statistical discrepancy between training and test distributions, we derive an Edgeworth expansion for the normalized difference in BN means and obtain an optimal weighting parameter that minimizes the mean-squared error of the adapted statistic. Reinterpreting BN TTA as a one-step M-estimator allows us to derive higher-order local asymptotic normality results, which incorporate skewness and other higher moments into the estimator's behavior. Moreover, we quantify the trade-offs among bias, variance, and skewness in the adaptation process and establish a corresponding generalization bound on the model risk. The refined saddlepoint approximations further deliver uniformly accurate density and tail probability estimates for the BN TTA statistic. These theoretical insights provide a comprehensive understanding of how higher-order corrections and robust one-step updating can enhance the reliability and performance of BN layers in adapting to changing data distributions.
Authors: Sampanna Yashwant Kahu, Naman Ahuja
Abstract: Social media and online forums are increasingly becoming popular. Unfortunately, these platforms are being used for spreading hate speech. In this paper, we design black-box techniques to protect users from hate-speech on online platforms by generating perturbations that can fool state of the art deep learning based hate speech detection models thereby decreasing their efficiency. We also ensure a minimal change in the original meaning of hate-speech. Our best perturbation attack is successfully able to evade hate-speech detection for 86.8 % of hateful text.
Authors: Jiaru Zou, Yikun Ban, Zihao Li, Yunzhe Qi, Ruizhong Qiu, Ling Yang, Jingrui He
Abstract: Large language models are typically adapted to downstream tasks through supervised fine-tuning on domain-specific data. While standard fine-tuning focuses on minimizing generation loss to optimize model parameters, we take a deeper step by retaining and leveraging the model's own learning signals, analogous to how human learners reflect on past mistakes to improve future performance. We first introduce the concept of Mistake Log to systematically track the model's learning behavior and recurring errors throughout fine-tuning. Treating the original transformer-based model as the Pilot, we correspondingly design a Copilot model to refine the Pilot's inference performance via logits rectification. We name the overall Pilot-Copilot framework the Transformer Copilot, which introduces (i) a novel Copilot model design, (ii) a joint training paradigm where the Copilot continuously learns from the evolving Mistake Log alongside the Pilot, and (iii) a fused inference paradigm where the Copilot rectifies the Pilot's logits for enhanced generation. We provide both theoretical and empirical analyses on our new learning framework. Experiments on 12 benchmarks spanning commonsense, arithmetic, and recommendation tasks demonstrate that Transformer Copilot consistently improves performance by up to 34.5%, while introducing marginal computational overhead to Pilot models and exhibiting strong scalability and transferability.
Authors: Fuchun Ge, Pavlo O. Dral
Abstract: Molecular dynamics (MD) is a powerful tool for exploring the behavior of atomistic systems, but its reliance on sequential numerical integration limits simulation efficiency. We present MDtrajNet-1, a foundational AI model that directly generates MD trajectories across chemical space, bypassing force calculations and integration. This approach accelerates simulations by up to two orders of magnitude compared to traditional MD, even those enhanced by machine-learning interatomic potentials. MDtrajNet-1 combines equivariant neural networks with a Transformer-based architecture to achieve strong accuracy and transferability in predicting long-time trajectories for both known and unseen systems. Remarkably, the errors of the trajectories generated by MDtrajNet-1 for various molecular systems are close to those of the conventional ab initio MD. The model's flexible design supports diverse application scenarios, including different statistical ensembles, boundary conditions, and interaction types. By overcoming the intrinsic speed barrier of conventional MD, MDtrajNet-1 opens new frontiers in efficient and scalable atomistic simulations.
Authors: Chenzhuo Zhao, Ziqian Liu, Xingda Wang, Junting Lu, Chaoyi Ruan
Abstract: Prompt optimization offers a practical and broadly applicable alternative to fine-tuning for improving large language model (LLM) performance. However, existing methods often rely on costly output generation, self-critiquing abilities, or human-annotated preferences, which limit their scalability, especially for smaller or non-instruction-tuned models. We introduce PMPO (Probabilistic Metric Prompt Optimization), a unified framework that refines prompts using token-level cross-entropy loss as a direct, lightweight evaluation signal. PMPO identifies low-quality prompt segments by masking and measuring their impact on loss, then rewrites and selects improved variants by minimizing loss over positive and negative examples. Unlike prior methods, it requires no output sampling or human evaluation during optimization, relying only on forward passes and log-likelihoods. PMPO supports both supervised and preference-based tasks through a closely aligned loss-based evaluation strategy. Experiments show that PMPO consistently outperforms prior methods across model sizes and tasks: it achieves the highest average accuracy on BBH, performs strongly on GSM8K and AQUA-RAT, and improves AlpacaEval 2.0 win rates by over 19 points. These results highlight PMPO's effectiveness, efficiency, and broad applicability.
Authors: Marc Brooks, Gabriel Durham, Kihyuk Hong, Ambuj Tewari
Abstract: Recent advances in generative artificial intelligence (GenAI) models have enabled the generation of personalized content that adapts to up-to-date user context. While personalized decision systems are often modeled using bandit formulations, the integration of GenAI introduces new structure into otherwise classical sequential learning problems. In GenAI-powered interventions, the agent selects a query, but the environment experiences a stochastic response drawn from the generative model. Standard bandit methods do not explicitly account for this structure, where actions influence rewards only through stochastic, observed treatments. We introduce generator-mediated bandit-Thompson sampling (GAMBITTS), a bandit approach designed for this action/treatment split, using mobile health interventions with large language model-generated text as a motivating case study. GAMBITTS explicitly models both the treatment and reward generation processes, using information in the delivered treatment to accelerate policy learning relative to standard methods. We establish regret bounds for GAMBITTS by decomposing sources of uncertainty in treatment and reward, identifying conditions where it achieves stronger guarantees than standard bandit approaches. In simulation studies, GAMBITTS consistently outperforms conventional algorithms by leveraging observed treatments to more accurately estimate expected rewards.
Authors: Arjhun Swaminathan, Mete Akg\"un
Abstract: Deep neural networks for image classification remain vulnerable to adversarial examples -- small, imperceptible perturbations that induce misclassifications. In black-box settings, where only the final prediction is accessible, crafting targeted attacks that aim to misclassify into a specific target class is particularly challenging due to narrow decision regions. Current state-of-the-art methods often exploit the geometric properties of the decision boundary separating a source image and a target image rather than incorporating information from the images themselves. In contrast, we propose Targeted Edge-informed Attack (TEA), a novel attack that utilizes edge information from the target image to carefully perturb it, thereby producing an adversarial image that is closer to the source image while still achieving the desired target classification. Our approach consistently outperforms current state-of-the-art methods across different models in low query settings (nearly 70\% fewer queries are used), a scenario especially relevant in real-world applications with limited queries and black-box access. Furthermore, by efficiently generating a suitable adversarial example, TEA provides an improved target initialization for established geometry-based attacks.
Authors: P. Huijse, J. De Ridder, L. Eyer, L. Rimoldini, B. Holl, N. Chornay, J. Roquette, K. Nienartowicz, G. Jevardat de Fombelle, D. J. Fritzewski, A. Kemp, V. Vanlaer, M. Vanrespaille, H. Wang, M. I. Carnerero, C. M. Raiteri, G. Marton, M. Madar\'asz, G. Clementini, P. Gavras, C. Aerts
Abstract: Gaia Data Release 3 (DR3) published for the first time epoch photometry, BP/RP (XP) low-resolution mean spectra, and supervised classification results for millions of variable sources. This extensive dataset offers a unique opportunity to study their variability by combining multiple Gaia data products. In preparation for DR4, we propose and evaluate a machine learning methodology capable of ingesting multiple Gaia data products to achieve an unsupervised classification of stellar and quasar variability. A dataset of 4 million Gaia DR3 sources is used to train three variational autoencoders (VAE), which are artificial neural networks (ANNs) designed for data compression and generation. One VAE is trained on Gaia XP low-resolution spectra, another on a novel approach based on the distribution of magnitude differences in the Gaia G band, and the third on folded Gaia G band light curves. Each Gaia source is compressed into 15 numbers, representing the coordinates in a 15-dimensional latent space generated by combining the outputs of these three models. The learned latent representation produced by the ANN effectively distinguishes between the main variability classes present in Gaia DR3, as demonstrated through both supervised and unsupervised classification analysis of the latent space. The results highlight a strong synergy between light curves and low-resolution spectral data, emphasising the benefits of combining the different Gaia data products. A two-dimensional projection of the latent variables reveals numerous overdensities, most of which strongly correlate with astrophysical properties, showing the potential of this latent space for astrophysical discovery. We show that the properties of our novel latent representation make it highly valuable for variability analysis tasks, including classification, clustering and outlier detection.
Authors: Simone Bombari, Inbar Seroussi, Marco Mondelli
Abstract: Differentially private (DP) linear regression has received significant attention in the recent theoretical literature, with several works aimed at obtaining improved error rates. A common approach is to set the clipping constant much larger than the expected norm of the per-sample gradients. While simplifying the analysis, this is however in sharp contrast with what empirical evidence suggests to optimize performance. Our work bridges this gap between theory and practice: we provide sharper rates for DP stochastic gradient descent (DP-SGD) by crucially operating in a regime where clipping happens frequently. Specifically, we consider the setting where the data is multivariate Gaussian, the number of training samples $n$ is proportional to the input dimension $d$, and the algorithm guarantees constant-order zero concentrated DP. Our method relies on establishing a deterministic equivalent for the trajectory of DP-SGD in terms of a family of ordinary differential equations (ODEs). As a consequence, the risk of DP-SGD is bounded between two ODEs, with upper and lower bounds matching for isotropic data. By studying these ODEs when $n / d$ is large enough, we demonstrate the optimality of aggressive clipping, and we uncover the benefits of decaying learning rate and private noise scheduling.
Authors: Javad Mirzaei, Jeebak Mitra, Gwenael Poitau
Abstract: With increased 5G deployments, network densification is higher than ever to support the exponentially high throughput requirements. However, this has meant a significant increase in energy consumption, leading to higher operational expenditure (OpEx) for network operators creating an acute need for improvements in network energy savings (NES). A key determinant of operational efficacy in cellular networks is the user association (UA) policy, as it affects critical aspects like spectral efficiency, load balancing etc. and therefore impacts the overall energy consumption of the network directly. Furthermore, with cellular network topologies lending themselves well to graphical abstractions, use of graphs in network optimization has gained significant prominence. In this work, we propose and analyze a graphical abstraction based optimization for UA in cellular networks to improve NES by determining when energy saving features like cell switch off can be activated. A comparison with legacy approaches establishes the superiority of the proposed approach.
Authors: Estelle Chigot, Dennis G. Wilson, Meriem Ghrib, Thomas Oberlin
Abstract: Semantic segmentation models trained on synthetic data often perform poorly on real-world images due to domain gaps, particularly in adverse conditions where labeled data is scarce. Yet, recent foundation models enable to generate realistic images without any training. This paper proposes to leverage such diffusion models to improve the performance of vision models when learned on synthetic data. We introduce two novel techniques for semantically consistent style transfer using diffusion models: Class-wise Adaptive Instance Normalization and Cross-Attention (CACTI) and its extension with selective attention Filtering (CACTIF). CACTI applies statistical normalization selectively based on semantic classes, while CACTIF further filters cross-attention maps based on feature similarity, preventing artifacts in regions with weak cross-attention correspondences. Our methods transfer style characteristics while preserving semantic boundaries and structural coherence, unlike approaches that apply global transformations or generate content without constraints. Experiments using GTA5 as source and Cityscapes/ACDC as target domains show that our approach produces higher quality images with lower FID scores and better content preservation. Our work demonstrates that class-aware diffusion-based style transfer effectively bridges the synthetic-to-real domain gap even with minimal target domain data, advancing robust perception systems for challenging real-world applications. The source code is available at: https://github.com/echigot/cactif.
Authors: Songlin Yang, Yikang Shen, Kaiyue Wen, Shawn Tan, Mayank Mishra, Liliang Ren, Rameswar Panda, Yoon Kim
Abstract: The attention mechanism is a core primitive in modern large language models (LLMs) and AI more broadly. Since attention by itself is permutation-invariant, position encoding is essential for modeling structured domains such as language. Rotary position encoding (RoPE) has emerged as the de facto standard approach for position encoding and is part of many modern LLMs. However, in RoPE the key/query transformation between two elements in a sequence is only a function of their relative position and otherwise independent of the actual input. This limits the expressivity of RoPE-based transformers. This paper describes PaTH, a flexible data-dependent position encoding scheme based on accumulated products of Householder(like) transformations, where each transformation is data-dependent, i.e., a function of the input. We derive an efficient parallel algorithm for training through exploiting a compact representation of products of Householder matrices, and implement a FlashAttention-style blockwise algorithm that minimizes I/O cost. Across both targeted synthetic benchmarks and moderate-scale real-world language modeling experiments, we find that PaTH demonstrates superior performance compared to RoPE and other recent baselines.
Authors: Guanting Dong, Yifei Chen, Xiaoxi Li, Jiajie Jin, Hongjin Qian, Yutao Zhu, Hangyu Mao, Guorui Zhou, Zhicheng Dou, Ji-Rong Wen
Abstract: Recently, large language models (LLMs) have shown remarkable reasoning capabilities via large-scale reinforcement learning (RL). However, leveraging the RL algorithm to empower effective multi-tool collaborative reasoning in LLMs remains an open challenge. In this paper, we introduce Tool-Star, an RL-based framework designed to empower LLMs to autonomously invoke multiple external tools during stepwise reasoning. Tool-Star integrates six types of tools and incorporates systematic designs in both data synthesis and training. To address the scarcity of tool-use data, we propose a general tool-integrated reasoning data synthesis pipeline, which combines tool-integrated prompting with hint-based sampling to automatically and scalably generate tool-use trajectories. A subsequent quality normalization and difficulty-aware classification process filters out low-quality samples and organizes the dataset from easy to hard. Furthermore, we propose a two-stage training framework to enhance multi-tool collaborative reasoning by: (1) cold-start fine-tuning, which guides LLMs to explore reasoning patterns via tool-invocation feedback; and (2) a multi-tool self-critic RL algorithm with hierarchical reward design, which reinforces reward understanding and promotes effective tool collaboration. Experimental analyses on over 10 challenging reasoning benchmarks highlight the effectiveness and efficiency of Tool-Star. The code is available at https://github.com/dongguanting/Tool-Star.
Authors: Sreetama Sarkar, Yue Che, Alex Gavin, Peter A. Beerel, Souvik Kundu
Abstract: Despite their remarkable progress in multimodal understanding tasks, large vision language models (LVLMs) often suffer from "hallucinations", generating texts misaligned with the visual context. Existing methods aimed at reducing hallucinations through inference time intervention incur a significant increase in latency. To mitigate this, we present SPIN, a task-agnostic attention-guided head suppression strategy that can be seamlessly integrated during inference, without incurring any significant compute or latency overhead. We investigate whether hallucination in LVLMs can be linked to specific model components. Our analysis suggests that hallucinations can be attributed to a dynamic subset of attention heads in each layer. Leveraging this insight, for each text query token, we selectively suppress attention heads that exhibit low attention to image tokens, keeping the top-K attention heads intact. Extensive evaluations on visual question answering and image description tasks demonstrate the efficacy of SPIN in reducing hallucination scores up to 2.7x while maintaining F1, and improving throughput by 1.8x compared to existing alternatives. Code is available at https://github.com/YUECHE77/SPIN.
Authors: Ruizhe Li, Chen Chen, Yuchen Hu, Yanjun Gao, Xi Wang, Emine Yilmaz
Abstract: Retrieval-Augmented Generation (RAG) leverages large language models (LLMs) combined with external contexts to enhance the accuracy and reliability of generated responses. However, reliably attributing generated content to specific context segments, context attribution, remains challenging due to the computationally intensive nature of current methods, which often require extensive fine-tuning or human annotation. In this work, we introduce a novel Jensen-Shannon Divergence driven method to Attribute Response to Context (ARC-JSD), enabling efficient and accurate identification of essential context sentences without additional fine-tuning or surrogate modelling. Evaluations on a wide range of RAG benchmarks, such as TyDi QA, Hotpot QA, and Musique, using instruction-tuned LLMs in different scales demonstrate superior accuracy and significant computational efficiency improvements compared to the previous surrogate-based method. Furthermore, our mechanistic analysis reveals specific attention heads and multilayer perceptron (MLP) layers responsible for context attribution, providing valuable insights into the internal workings of RAG models.
Authors: Zhepei Wei, Wenlin Yao, Yao Liu, Weizhi Zhang, Qin Lu, Liang Qiu, Changlong Yu, Puyang Xu, Chao Zhang, Bing Yin, Hyokun Yun, Lihong Li
Abstract: While reinforcement learning (RL) has demonstrated remarkable success in enhancing large language models (LLMs), it has primarily focused on single-turn tasks such as solving math problems. Training effective web agents for multi-turn interactions remains challenging due to the complexity of long-horizon decision-making across dynamic web interfaces. In this work, we present WebAgent-R1, a simple yet effective end-to-end multi-turn RL framework for training web agents. It learns directly from online interactions with web environments by asynchronously generating diverse trajectories, entirely guided by binary rewards depending on task success. Experiments on the WebArena-Lite benchmark demonstrate the effectiveness of WebAgent-R1, boosting the task success rate of Qwen-2.5-3B from 6.1% to 33.9% and Llama-3.1-8B from 8.5% to 44.8%, significantly outperforming existing state-of-the-art methods and strong proprietary models such as OpenAI o3. In-depth analyses reveal the effectiveness of the thinking-based prompting strategy and test-time scaling through increased interactions for web tasks. We further investigate different RL initialization policies by introducing two variants, namely WebAgent-R1-Zero and WebAgent-R1-CoT, which highlight the importance of the warm-up training stage (i.e., behavior cloning) and provide insights on incorporating long chain-of-thought (CoT) reasoning in web agents.
Authors: Jisu Han, Jaemin Na, Wonjun Hwang
Abstract: Test-time adaptation aims to adapt to realistic environments in an online manner by learning during test time. Entropy minimization has emerged as a principal strategy for test-time adaptation due to its efficiency and adaptability. Nevertheless, it remains underexplored in continual test-time adaptation, where stability is more important. We observe that the entropy minimization method often suffers from model collapse, where the model converges to predicting a single class for all images due to a trivial solution. We propose ranked entropy minimization to mitigate the stability problem of the entropy minimization method and extend its applicability to continuous scenarios. Our approach explicitly structures the prediction difficulty through a progressive masking strategy. Specifically, it gradually aligns the model's probability distributions across different levels of prediction difficulty while preserving the rank order of entropy. The proposed method is extensively evaluated across various benchmarks, demonstrating its effectiveness through empirical results. Our code is available at https://github.com/pilsHan/rem
Authors: Mohamed S. Elmahdy, Marius Staring, Patrick J. H. de Koning, Samer Alabed, Mahan Salehi, Faisal Alandejani, Michael Sharkey, Ziad Aldabbagh, Andrew J. Swift, Rob J. van der Geest
Abstract: Accurate and efficient quantification of cardiac function is essential for the estimation of prognosis of cardiovascular diseases (CVDs). One of the most commonly used metrics for evaluating cardiac pumping performance is left ventricular ejection fraction (LVEF). However, LVEF can be affected by factors such as inter-observer variability and varying pre-load and after-load conditions, which can reduce its reproducibility. Additionally, cardiac dysfunction may not always manifest as alterations in LVEF, such as in heart failure and cardiotoxicity diseases. An alternative measure that can provide a relatively load-independent quantitative assessment of myocardial contractility is myocardial strain and strain rate. By using LVEF in combination with myocardial strain, it is possible to obtain a thorough description of cardiac function. Automated estimation of LVEF and other volumetric measures from cine-MRI sequences can be achieved through segmentation models, while strain calculation requires the estimation of tissue displacement between sequential frames, which can be accomplished using registration models. These tasks are often performed separately, potentially limiting the assessment of cardiac function. To address this issue, in this study we propose an end-to-end deep learning (DL) model that jointly estimates groupwise (GW) registration and segmentation for cardiac cine-MRI images. The proposed anatomically-guided Deep GW network was trained and validated on a large dataset of 4-chamber view cine-MRI image series of 374 subjects. A quantitative comparison with conventional GW registration using elastix and two DL-based methods showed that the proposed model improved performance and substantially reduced computation time.
Authors: Jiquan Shan, Junxiao Wang, Lifeng Zhao, Liang Cai, Hongyuan Zhang, Ioannis Liritzis
Abstract: Recently, vision transformers (ViTs) have achieved excellent performance on vision tasks by measuring the global self-attention among the image patches. Given $n$ patches, they will have quadratic complexity such as $\mathcal{O}(n^2)$ and the time cost is high when splitting the input image with a small granularity. Meanwhile, the pivotal information is often randomly gathered in a few regions of an input image, some tokens may not be helpful for the downstream tasks. To handle this problem, we introduce an anchor-based efficient vision transformer (AnchorFormer), which employs the anchor tokens to learn the pivotal information and accelerate the inference. Firstly, by estimating the bipartite attention between the anchors and tokens, the complexity will be reduced from $\mathcal{O}(n^2)$ to $\mathcal{O}(mn)$, where $m$ is an anchor number and $m < n$. Notably, by representing the anchors with the neurons in a neural layer, we can differentiable learn these distributions and approximate global self-attention through the Markov process. Moreover, we extend the proposed model to three downstream tasks including classification, detection, and segmentation. Extensive experiments show the effectiveness of our AnchorFormer, e.g., achieving up to a 9.0% higher accuracy or 46.7% FLOPs reduction on ImageNet classification, 81.3% higher mAP on COCO detection under comparable FLOPs, as compared to the current baselines.
Authors: Robbert Reijnen, Yaoxin Wu, Zaharah Bukhsh, Yingqian Zhang
Abstract: Deep reinforcement learning (DRL) has been widely used for dynamic algorithm configuration, particularly in evolutionary computation, which benefits from the adaptive update of parameters during the algorithmic execution. However, applying DRL to algorithm configuration for multi-objective combinatorial optimization (MOCO) problems remains relatively unexplored. This paper presents a novel graph neural network (GNN) based DRL to configure multi-objective evolutionary algorithms. We model the dynamic algorithm configuration as a Markov decision process, representing the convergence of solutions in the objective space by a graph, with their embeddings learned by a GNN to enhance the state representation. Experiments on diverse MOCO challenges indicate that our method outperforms traditional and DRL-based algorithm configuration methods in terms of efficacy and adaptability. It also exhibits advantageous generalizability across objective types and problem sizes, and applicability to different evolutionary computation methods.
Authors: Huitong Yang, Zhuoxiao Chen, Fengyi Zhang, Zi Huang, Yadan Luo
Abstract: Maintaining robust 3D perception under dynamic and unpredictable test-time conditions remains a critical challenge for autonomous driving systems. Existing test-time adaptation (TTA) methods often fail in high-variance tasks like 3D object detection due to unstable optimization and sharp minima. While recent model merging strategies based on linear mode connectivity (LMC) offer improved stability by interpolating between fine-tuned checkpoints, they are computationally expensive, requiring repeated checkpoint access and multiple forward passes. In this paper, we introduce CodeMerge, a lightweight and scalable model merging framework that bypasses these limitations by operating in a compact latent space. Instead of loading full models, CodeMerge represents each checkpoint with a low-dimensional fingerprint derived from the source model's penultimate features and constructs a key-value codebook. We compute merging coefficients using ridge leverage scores on these fingerprints, enabling efficient model composition without compromising adaptation quality. Our method achieves strong performance across challenging benchmarks, improving end-to-end 3D detection 14.9% NDS on nuScenes-C and LiDAR-based detection by over 7.6% mAP on nuScenes-to-KITTI, while benefiting downstream tasks such as online mapping, motion prediction and planning even without training. Code and pretrained models are released in the supplementary material.
Authors: Vignesh Gopinathan, Urs Zimmermann, Michael Arnold, Matthias Rottmann
Abstract: Video captioning models have seen notable advancements in recent years, especially with regard to their ability to capture temporal information. While many research efforts have focused on architectural advancements, such as temporal attention mechanisms, there remains a notable gap in understanding how models capture and utilize temporal semantics for effective temporal feature extraction, especially in the context of Advanced Driver Assistance Systems. We propose an automated LiDAR-based captioning procedure that focuses on the temporal dynamics of traffic participants. Our approach uses a rule-based system to extract essential details such as lane position and relative motion from object tracks, followed by a template-based caption generation. Our findings show that training SwinBERT, a video captioning model, using only front camera images and supervised with our template-based captions, specifically designed to encapsulate fine-grained temporal behavior, leads to improved temporal understanding consistently across three datasets. In conclusion, our results clearly demonstrate that integrating LiDAR-based caption supervision significantly enhances temporal understanding, effectively addressing and reducing the inherent visual/static biases prevalent in current state-of-the-art model architectures.
Authors: Daniel Scalena, Gabriele Sarti, Arianna Bisazza, Elisabetta Fersini, Malvina Nissim
Abstract: High-quality machine translation systems based on large language models (LLMs) have simplified the production of personalized translations reflecting specific stylistic constraints. However, these systems still struggle in settings where stylistic requirements are less explicit and might be harder to convey via prompting. We explore various strategies for personalizing LLM-generated translations in low-resource settings, focusing on the challenging literary translation domain. We explore prompting strategies and inference-time interventions for steering model generations towards a personalized style, and propose a contrastive framework exploiting latent concepts extracted from sparse autoencoders to identify salient personalization properties. Our results show that steering achieves strong personalization while preserving translation quality. We further examine the impact of steering on LLM representations, finding model layers with a relevant impact for personalization are impacted similarly by multi-shot prompting and our steering method, suggesting similar mechanism at play.
Authors: Zhaomin Wu, Ziyang Wang, Bingsheng He
Abstract: Tabular data, ubiquitous and rich in informational value, is an increasing focus for deep representation learning, yet progress is hindered by studies centered on single tables or isolated databases, which limits model capabilities due to data scale. While collaborative learning approaches such as federated learning, transfer learning, split learning, and tabular foundation models aim to learn from multiple correlated databases, they are challenged by a scarcity of real-world interconnected tabular resources. Current data lakes and corpora largely consist of isolated databases lacking defined inter-database correlations. To overcome this, we introduce WikiDBGraph, a large-scale graph of 100,000 real-world tabular databases from WikiData, interconnected by 17 million edges and characterized by 13 node and 12 edge properties derived from its database schema and data distribution. WikiDBGraph's weighted edges identify both instance- and feature-overlapped databases. Experiments on these newly identified databases confirm that collaborative learning yields superior performance, thereby offering considerable promise for structured foundation model training while also exposing key challenges and future directions for learning from interconnected tabular data.
Authors: Wenjie Yang, Mao Zheng, Mingyang Song, Zheng Li
Abstract: Large language models (LLMs) have recently demonstrated remarkable capabilities in machine translation (MT). However, most advanced MT-specific LLMs heavily rely on external supervision signals during training, such as human-annotated reference data or trained reward models (RMs), which are often expensive to obtain and challenging to scale. To overcome this limitation, we propose a Simple Self-Rewarding (SSR) Reinforcement Learning (RL) framework for MT that is reference-free, fully online, and relies solely on self-judging rewards. Training with SSR using 13K monolingual examples and Qwen-2.5-7B as the backbone, our model SSR-Zero-7B outperforms existing MT-specific LLMs, e.g., TowerInstruct-13B and GemmaX-28-9B, as well as larger general LLMs like Qwen2.5-32B-Instruct in English $\leftrightarrow$ Chinese translation tasks from WMT23, WMT24, and Flores200 benchmarks. Furthermore, by augmenting SSR with external supervision from COMET, our strongest model, SSR-X-Zero-7B, achieves state-of-the-art performance in English $\leftrightarrow$ Chinese translation, surpassing all existing open-source models under 72B parameters and even outperforming closed-source models, e.g., GPT-4o and Gemini 1.5 Pro. Our analysis highlights the effectiveness of the self-rewarding mechanism compared to the external LLM-as-a-judge approach in MT and demonstrates its complementary benefits when combined with trained RMs. Our findings provide valuable insight into the potential of self-improving RL methods. We have publicly released our code, data and models.
Authors: Stephen Y. Zhang, Michael P H Stumpf
Abstract: We consider the Schr\"odinger bridge problem which, given ensemble measurements of the initial and final configurations of a stochastic dynamical system and some prior knowledge on the dynamics, aims to reconstruct the "most likely" evolution of the system compatible with the data. Most existing literature assume Brownian reference dynamics and are implicitly limited to potential-driven dynamics. We depart from this regime and consider reference processes described by a multivariate Ornstein-Uhlenbeck process with generic drift matrix $\mathbf{A} \in \mathbb{R}^{d \times d}$. When $\mathbf{A}$ is asymmetric, this corresponds to a non-equilibrium system with non-conservative forces at play: this is important for applications to biological systems, which are naturally exist out-of-equilibrium. In the case of Gaussian marginals, we derive explicit expressions that characterise the solution of both the static and dynamic Schr\"odinger bridge. For general marginals, we propose mvOU-OTFM, a simulation-free algorithm based on flow and score matching for learning the Schr\"odinger bridge. In application to a range of problems based on synthetic and real single cell data, we demonstrate that mvOU-OTFM achieves higher accuracy compared to competing methods, whilst being significantly faster to train.
Authors: Feilong Tang, Chengzhi Liu, Zhongxing Xu, Ming Hu, Zelin Peng, Zhiwei Yang, Jionglong Su, Minquan Lin, Yifan Peng, Xuelian Cheng, Imran Razzak, Zongyuan Ge
Abstract: Recent advancements in multimodal large language models (MLLMs) have significantly improved performance in visual question answering. However, they often suffer from hallucinations. In this work, hallucinations are categorized into two main types: initial hallucinations and snowball hallucinations. We argue that adequate contextual information can be extracted directly from the token interaction process. Inspired by causal inference in the decoding strategy, we propose to leverage causal masks to establish information propagation between multimodal tokens. The hypothesis is that insufficient interaction between those tokens may lead the model to rely on outlier tokens, overlooking dense and rich contextual cues. Therefore, we propose to intervene in the propagation process by tackling outlier tokens to enhance in-context inference. With this goal, we present FarSight, a versatile plug-and-play decoding strategy to reduce attention interference from outlier tokens merely by optimizing the causal mask. The heart of our method is effective token propagation. We design an attention register structure within the upper triangular matrix of the causal mask, dynamically allocating attention to capture attention diverted to outlier tokens. Moreover, a positional awareness encoding method with a diminishing masking rate is proposed, allowing the model to attend to further preceding tokens, especially for video sequence tasks. With extensive experiments, FarSight demonstrates significant hallucination-mitigating performance across different MLLMs on both image and video benchmarks, proving its effectiveness.
Authors: Shogo Nakakita
Abstract: We examine the concentration of uniform generalization errors around their expectation in binary linear classification problems via an isoperimetric argument. In particular, we establish Poincar\'{e} and log-Sobolev inequalities for the joint distribution of the output labels and the label-weighted input vectors, which we apply to derive concentration bounds. The derived concentration bounds are sharp up to moderate multiplicative constants by those under well-balanced labels. In asymptotic analysis, we also show that almost sure convergence of uniform generalization errors to their expectation occurs in very broad settings, such as proportionally high-dimensional regimes. Using this convergence, we establish uniform laws of large numbers under dimension-free conditions.
Authors: Hai-Feng Zhang, Zhao-Yun Chen, Peng Wang, Liang-Liang Guo, Tian-Le Wang, Xiao-Yan Yang, Ren-Ze Zhao, Ze-An Zhao, Sheng Zhang, Lei Du, Hao-Ran Tao, Zhi-Long Jia, Wei-Cheng Kong, Huan-Yu Liu, Athanasios V. Vasilakos, Yang Yang, Yu-Chun Wu, Ji Guan, Peng Duan, Guo-Ping Guo
Abstract: Quantum machine learning (QML) models, like their classical counterparts, are vulnerable to adversarial attacks, hindering their secure deployment. Here, we report the first systematic experimental robustness benchmark for 20-qubit quantum neural network (QNN) classifiers executed on a superconducting processor. Our benchmarking framework features an efficient adversarial attack algorithm designed for QNNs, enabling quantitative characterization of adversarial robustness and robustness bounds. From our analysis, we verify that adversarial training reduces sensitivity to targeted perturbations by regularizing input gradients, significantly enhancing QNN's robustness. Additionally, our analysis reveals that QNNs exhibit superior adversarial robustness compared to classical neural networks, an advantage attributed to inherent quantum noise. Furthermore, the empirical upper bound extracted from our attack experiments shows a minimal deviation ($3 \times 10^{-3}$) from the theoretical lower bound, providing strong experimental confirmation of the attack's effectiveness and the tightness of fidelity-based robustness bounds. This work establishes a critical experimental framework for assessing and improving quantum adversarial robustness, paving the way for secure and reliable QML applications.
Authors: Moritz Stargalla, Christoph Hertrich, Daniel Reichman
Abstract: An established measure of the expressive power of a given ReLU neural network is the number of linear regions into which it partitions the input space. There exist many different, non-equivalent definitions of what a linear region actually is. We systematically assess which papers use which definitions and discuss how they relate to each other. We then analyze the computational complexity of counting the number of such regions for the various definitions. Generally, this turns out to be an intractable problem. We prove NP- and #P-hardness results already for networks with one hidden layer and strong hardness of approximation results for two or more hidden layers. Finally, on the algorithmic side, we demonstrate that counting linear regions can at least be achieved in polynomial space for some common definitions.
Authors: Thibaud Gloaguen, Robin Staab, Nikola Jovanovi\'c, Martin Vechev
Abstract: As open-source language models (OSMs) grow more capable and are widely shared and finetuned, ensuring model provenance, i.e., identifying the origin of a given model instance, has become an increasingly important issue. At the same time, existing backdoor-based model fingerprinting techniques often fall short of achieving key requirements of real-world model ownership detection. In this work, we build on the observation that while current open-source model watermarks fail to achieve reliable content traceability, they can be effectively adapted to address the challenge of model provenance. To this end, we introduce the concept of domain-specific watermarking for model fingerprinting. Rather than watermarking all generated content, we train the model to embed watermarks only within specified subdomains (e.g., particular languages or topics). This targeted approach ensures detection reliability, while improving watermark durability and quality under a range of real-world deployment settings. Our evaluations show that domain-specific watermarking enables model fingerprinting with strong statistical guarantees, controllable false positive rates, high detection power, and preserved generation quality. Moreover, we find that our fingerprints are inherently stealthy and naturally robust to real-world variability across deployment scenarios.
Authors: Florentin Beck, William Rudman, Carsten Eickhoff
Abstract: Large Language Models (LLMs) present significant computational and memory challenges due to their extensive size, making pruning essential for their efficient deployment. Existing one-shot pruning methods often apply uniform sparsity constraints across layers or within each layer, resulting in suboptimal performance, especially at high sparsity ratios. This work introduces TRIM (Targeted Row-wise Iterative Metric-driven pruning), a novel approach that applies varying sparsity ratios to individual output dimensions (rows) within each layer. TRIM employs an iterative adjustment process guided by quality metrics to optimize dimension-wise sparsity allocation, focusing on reducing variance in quality retention across outputs to preserve critical information. TRIM can be seamlessly integrated with existing layer-wise pruning strategies. Our evaluations on perplexity and zero-shot tasks across diverse LLM families (Qwen2.5, LLaMA-2, and OPT) and sparsity levels demonstrate that TRIM achieves new state-of-the-art results and enhances stability. For instance, at 80% sparsity, TRIM reduces perplexity by 48% for Qwen2.5-14B and over 90% for OPT-13B compared to baseline methods. We conclude that fine-grained, dimension-wise sparsity adaptation is crucial for pushing the limits of extreme LLM compression. Code available at: https://github.com/flobk/TRIM
Authors: Punya Syon Pandey, Samuel Simko, Kellin Pelrine, Zhijing Jin
Abstract: As large language models gain popularity, their vulnerability to adversarial attacks remains a primary concern. While fine-tuning models on domain-specific datasets is often employed to improve model performance, it can introduce vulnerabilities within the underlying model. In this work, we investigate Accidental Misalignment, unexpected vulnerabilities arising from characteristics of fine-tuning data. We begin by identifying potential correlation factors such as linguistic features, semantic similarity, and toxicity within our experimental datasets. We then evaluate the adversarial performance of these fine-tuned models and assess how dataset factors correlate with attack success rates. Lastly, we explore potential causal links, offering new insights into adversarial defense strategies and highlighting the crucial role of dataset design in preserving model alignment. Our code is available at https://github.com/psyonp/accidental_misalignment.
Authors: Ziming liu, Bryan Liu, Alvaro Valcarce, Xiaoli Chu
Abstract: Integrating large AI models (LAMs) into 6G mobile networks promises to redefine protocol design and control-plane intelligence by enabling autonomous, cognitive network operations. While industry concepts, such as ETSI's Experiential Networked Intelligence (ENI), envision LAM-driven agents for adaptive network slicing and intent-based management, practical implementations still face challenges in protocol literacy and real-world deployment. This paper presents an end-to-end demonstration of a LAM that generates standards-compliant, ASN.1-encoded Radio Resource Control (RRC) messages as part of control-plane procedures inside a gNB. We treat RRC messaging as a domain-specific language and fine-tune a decoder-only transformer model (LLaMA class) using parameter-efficient Low-Rank Adaptation (LoRA) on RRC messages linearized to retain their ASN.1 syntactic structure before standard byte-pair encoding tokenization. This enables combinatorial generalization over RRC protocol states while minimizing training overhead. On 30k field-test request-response pairs, our 8 B model achieves a median cosine similarity of 0.97 with ground-truth messages on an edge GPU -- a 61 % relative gain over a zero-shot LLaMA-3 8B baseline -- indicating substantially improved structural and semantic RRC fidelity. Overall, our results show that LAMs, when augmented with Radio Access Network (RAN)-specific reasoning, can directly orchestrate control-plane procedures, representing a stepping stone toward the AI-native air-interface paradigm. Beyond RRC emulation, this work lays the groundwork for future AI-native wireless standards.
Authors: Xiaoyu Xu, Xiang Yue, Yang Liu, Qingqing Ye, Haibo Hu, Minxin Du
Abstract: Unlearning in large language models (LLMs) is intended to remove the influence of specific data, yet current evaluations rely heavily on token-level metrics such as accuracy and perplexity. We show that these metrics can be misleading: models often appear to forget, but their original behavior can be rapidly restored with minimal fine-tuning, revealing that unlearning may obscure information rather than erase it. To diagnose this phenomenon, we introduce a representation-level evaluation framework using PCA-based similarity and shift, centered kernel alignment, and Fisher information. Applying this toolkit across six unlearning methods, three domains (text, code, math), and two open-source LLMs, we uncover a critical distinction between reversible and irreversible forgetting. In reversible cases, models suffer token-level collapse yet retain latent features; in irreversible cases, deeper representational damage occurs. We further provide a theoretical account linking shallow weight perturbations near output layers to misleading unlearning signals, and show that reversibility is modulated by task type and hyperparameters. Our findings reveal a fundamental gap in current evaluation practices and establish a new diagnostic foundation for trustworthy unlearning in LLMs. We provide a unified toolkit for analyzing LLM representation changes under unlearning and relearning: https://github.com/XiaoyuXU1/Representational_Analysis_Tools.git.
URLs: https://github.com/XiaoyuXU1/Representational_Analysis_Tools.git.
Authors: Haonian Ji, Shi Qiu, Siyang Xin, Siwei Han, Zhaorun Chen, Hongyi Wang, Dake Zhang, Huaxiu Yao
Abstract: While foundation models (FMs), such as diffusion models and large vision-language models (LVLMs), have been widely applied in educational contexts, their ability to generate pedagogically effective visual explanations remains limited. Most existing approaches focus primarily on textual reasoning, overlooking the critical role of structured and interpretable visualizations in supporting conceptual understanding. To better assess the visual reasoning capabilities of FMs in educational settings, we introduce EduVisBench, a multi-domain, multi-level benchmark. EduVisBench features diverse STEM problem sets requiring visually grounded solutions, along with a fine-grained evaluation rubric informed by pedagogical theory. Our empirical analysis reveals that existing models frequently struggle with the inherent challenge of decomposing complex reasoning and translating it into visual representations aligned with human cognitive processes. To address these limitations, we propose EduVisAgent, a multi-agent collaborative framework that coordinates specialized agents for instructional planning, reasoning decomposition, metacognitive prompting, and visualization design. Experimental results show that EduVisAgent substantially outperforms all baselines, achieving a 40.2% improvement and delivering more educationally aligned visualizations. EduVisBench and EduVisAgent are available at https://github.com/aiming-lab/EduVisBench and https://github.com/aiming-lab/EduVisAgent.
URLs: https://github.com/aiming-lab/EduVisBench, https://github.com/aiming-lab/EduVisAgent.
Authors: Zhehao Huang, Yuhang Liu, Yixin Lou, Zhengbao He, Mingzhen He, Wenxing Zhou, Tao Li, Kehan Li, Zeyi Huang, Xiaolin Huang
Abstract: Continual post-training adapts a single text-to-image diffusion model to learn new tasks without incurring the cost of separate models, but naive post-training causes forgetting of pretrained knowledge and undermines zero-shot compositionality. We observe that the absence of a standardized evaluation protocol hampers related research for continual post-training. To address this, we introduce T2I-ConBench, a unified benchmark for continual post-training of text-to-image models. T2I-ConBench focuses on two practical scenarios, item customization and domain enhancement, and analyzes four dimensions: (1) retention of generality, (2) target-task performance, (3) catastrophic forgetting, and (4) cross-task generalization. It combines automated metrics, human-preference modeling, and vision-language QA for comprehensive assessment. We benchmark ten representative methods across three realistic task sequences and find that no approach excels on all fronts. Even joint "oracle" training does not succeed for every task, and cross-task generalization remains unsolved. We release all datasets, code, and evaluation tools to accelerate research in continual post-training for text-to-image models.
Authors: Hannah Sansford, Nick Whiteley, Patrick Rubin-Delanchy
Abstract: We present a generalised Hanson-Wright inequality and use it to establish new statistical insights into the geometry of data point-clouds. In the setting of a general random function model of data, we clarify the roles played by three notions of dimensionality: ambient intrinsic dimension $p_{\mathrm{int}}$, which measures total variability across orthogonal feature directions; correlation rank, which measures functional complexity across samples; and latent intrinsic dimension, which is the dimension of manifold structure hidden in data. Our analysis shows that in order for persistence diagrams to reveal latent homology and for manifold structure to emerge it is sufficient that $p_{\mathrm{int}}\gg \log n$, where $n$ is the sample size. Informed by these theoretical perspectives, we revisit the ground-breaking neuroscience discovery of toroidal structure in grid-cell activity made by Gardner et al. (Nature, 2022): our findings reveal, for the first time, evidence that this structure is in fact isometric to physical space, meaning that grid cell activity conveys a geometrically faithful representation of the real world.
Authors: Nour Jedidi, Yung-Sung Chuang, James Glass, Jimmy Lin
Abstract: With the growing success of reasoning models across complex natural language tasks, researchers in the Information Retrieval (IR) community have begun exploring how similar reasoning capabilities can be integrated into passage rerankers built on Large Language Models (LLMs). These methods typically employ an LLM to produce an explicit, step-by-step reasoning process before arriving at a final relevance prediction. But, does reasoning actually improve reranking accuracy? In this paper, we dive deeper into this question, studying the impact of the reasoning process by comparing reasoning-based pointwise rerankers (ReasonRR) to standard, non-reasoning pointwise rerankers (StandardRR) under identical training conditions, and observe that StandardRR generally outperforms ReasonRR. Building on this observation, we then study the importance of reasoning to ReasonRR by disabling its reasoning process (ReasonRR-NoReason), and find that ReasonRR-NoReason is surprisingly more effective than ReasonRR. Examining the cause of this result, our findings reveal that reasoning-based rerankers are limited by the LLM's reasoning process, which pushes it toward polarized relevance scores and thus fails to consider the partial relevance of passages, a key factor for the accuracy of pointwise rerankers.
Authors: Shuichi Nishino, Tomohiro Shiraishi, Teruyuki Katsuoka, Ichiro Takeuchi
Abstract: Graph Neural Networks (GNNs) have gained prominence for their ability to process graph-structured data across various domains. However, interpreting GNN decisions remains a significant challenge, leading to the adoption of saliency maps for identifying influential nodes and edges. Despite their utility, the reliability of GNN saliency maps has been questioned, particularly in terms of their robustness to noise. In this study, we propose a statistical testing framework to rigorously evaluate the significance of saliency maps. Our main contribution lies in addressing the inflation of the Type I error rate caused by double-dipping of data, leveraging the framework of Selective Inference. Our method provides statistically valid $p$-values while controlling the Type I error rate, ensuring that identified salient subgraphs contain meaningful information rather than random artifacts. To demonstrate the effectiveness of our method, we conduct experiments on both synthetic and real-world datasets, showing its effectiveness in assessing the reliability of GNN interpretations.
Authors: Jintian Shao, Hongyi Huang, Jiayi Wu, Beiwen Zhang, ZhiYu Wu, You Shan, MingKai Zheng
Abstract: During the finetuning stage of text generation tasks, standard cross-entropy loss treats all tokens equally. This can lead models to overemphasize high-frequency, low-information tokens, neglecting lower-frequency tokens crucial for specificity and informativeness in generated content. This paper introduces a novel loss function, Power-Law Decay Loss (PDL), specifically designed to optimize the finetuning process for text generation. The core motivation for PDL stems from observations in information theory and linguistics: the informativeness of a token is often inversely proportional to its frequency of occurrence. PDL re-weights the contribution of each token in the standard cross-entropy loss based on its frequency in the training corpus, following a power-law decay. Specifically, the weights for high-frequency tokens are reduced, while low-frequency, information-dense tokens are assigned higher weights. This mechanism guides the model during finetuning to focus more on learning and generating tokens that convey specific and unique information, thereby enhancing the quality, diversity, and informativeness of the generated text. We theoretically elaborate on the motivation and construction of PDL and discuss its potential applications and advantages across various text generation finetuning tasks, such as abstractive summarization, dialogue systems, and style transfer.
Authors: Hongyuan Tao, Ying Zhang, Zhenhao Tang, Hongen Peng, Xukun Zhu, Bingchang Liu, Yingguang Yang, Ziyin Zhang, Zhaogui Xu, Haipeng Zhang, Linchao Zhu, Rui Wang, Hang Yu, Jianguo Li, Peng Di
Abstract: Recent advances in Large Language Models (LLMs) have shown promise in function-level code generation, yet repository-level software engineering tasks remain challenging. Current solutions predominantly rely on proprietary LLM agents, which introduce unpredictability and limit accessibility, raising concerns about data privacy and model customization. This paper investigates whether open-source LLMs can effectively address repository-level tasks without requiring agent-based approaches. We demonstrate this is possible by enabling LLMs to comprehend functions and files within codebases through their semantic information and structural dependencies. To this end, we introduce Code Graph Models (CGMs), which integrate repository code graph structures into the LLM's attention mechanism and map node attributes to the LLM's input space using a specialized adapter. When combined with an agentless graph RAG framework, our approach achieves a 43.00% resolution rate on the SWE-bench Lite benchmark using the open-source Qwen2.5-72B model. This performance ranks first among open weight models, second among methods with open-source systems, and eighth overall, surpassing the previous best open-source model-based method by 12.33%.
Authors: Yuhui Zhang, Dongshen Wu, Yuichiro Wada, Takafumi Kanamori
Abstract: A reliable uncertainty estimation method is the foundation of many modern out-of-distribution (OOD) detectors, which are critical for safe deployments of deep learning models in the open world. In this work, we propose TULiP, a theoretically-driven post-hoc uncertainty estimator for OOD detection. Our approach considers a hypothetical perturbation applied to the network before convergence. Based on linearized training dynamics, we bound the effect of such perturbation, resulting in an uncertainty score computable by perturbing model parameters. Ultimately, our approach computes uncertainty from a set of sampled predictions. We visualize our bound on synthetic regression and classification datasets. Furthermore, we demonstrate the effectiveness of TULiP using large-scale OOD detection benchmarks for image classification. Our method exhibits state-of-the-art performance, particularly for near-distribution samples.
Authors: Keshav Ramji, Tahira Naseem, Ram\'on Fernandez Astudillo
Abstract: When language model (LM) users aim to improve the quality of its generations, it is crucial to specify concrete behavioral attributes that the model should strive to reflect. However, curating such principles across many domains, even non-exhaustively, requires a labor-intensive annotation process. To automate this process, we propose eliciting these latent attributes guiding model reasoning towards human-preferred responses by explicitly modeling them in a self-correction setting. Our approach mines new principles from the LM itself and compresses the discovered elements to an interpretable set via clustering. Specifically, we employ an approximation of posterior-regularized Monte Carlo Expectation-Maximization to both identify a condensed set of the most effective latent principles and teach the LM to strategically invoke them in order to intrinsically refine its responses. We demonstrate that bootstrapping our algorithm over multiple iterations enables smaller language models (7-8B parameters) to self-improve, achieving +8-10% in AlpacaEval win-rate, an average of +0.3 on MT-Bench, and +19-23% in principle-following win-rate on IFEval. We also show that clustering the principles yields interpretable and diverse model-generated constitutions while retaining model performance. The gains our method achieves highlight the potential of automated, principle-driven post-training recipes toward continual self-improvement.
Authors: Bosung Kim, Prithviraj Ammanabrolu
Abstract: We introduce $\infty$-THOR, a new framework for long-horizon embodied tasks that advances long-context understanding in embodied AI. $\infty$-THOR provides: (1) a generation framework for synthesizing scalable, reproducible, and unlimited long-horizon trajectories; (2) a novel embodied QA task, Needle(s) in the Embodied Haystack, where multiple scattered clues across extended trajectories test agents' long-context reasoning ability; and (3) a long-horizon dataset and benchmark suite featuring complex tasks that span hundreds of environment steps, each paired with ground-truth action sequences. To enable this capability, we explore architectural adaptations, including interleaved Goal-State-Action modeling, context extension techniques, and Context Parallelism, to equip LLM-based agents for extreme long-context reasoning and interaction. Experimental results and analyses highlight the challenges posed by our benchmark and provide insights into training strategies and model behaviors under long-horizon conditions. Our work provides a foundation for the next generation of embodied AI systems capable of robust, long-term reasoning and planning.
Authors: Karlis Martins Briedis, Markus Gross, Christopher Schroers
Abstract: Recent optical flow estimation methods often employ local cost sampling from a dense all-pairs correlation volume. This results in quadratic computational and memory complexity in the number of pixels. Although an alternative memory-efficient implementation with on-demand cost computation exists, this is slower in practice and therefore prior methods typically process images at reduced resolutions, missing fine-grained details. To address this, we propose a more efficient implementation of the all-pairs correlation volume sampling, still matching the exact mathematical operator as defined by RAFT. Our approach outperforms on-demand sampling by up to 90% while maintaining low memory usage, and performs on par with the default implementation with up to 95% lower memory usage. As cost sampling makes up a significant portion of the overall runtime, this can translate to up to 50% savings for the total end-to-end model inference in memory-constrained environments. Our evaluation of existing methods includes an 8K ultra-high-resolution dataset and an additional inference-time modification of the recent SEA-RAFT method. With this, we achieve state-of-the-art results at high resolutions both in accuracy and efficiency.
Authors: Sanjana Chalavadi, Andrei Pastor, Terry Leitch
Abstract: This study analyzes tract-level real estate ownership patterns in New York State (NYS) and New York City (NYC) to uncover racial disparities. We use an advanced race/ethnicity imputation model (LSTM+Geo with XGBoost filtering, validated at 89.2% accuracy) to compare the predicted racial composition of property owners to the resident population from census data. We examine both a Full Model (statewide) and a Name-Only LSTM Model (NYC) to assess how incorporating geospatial context affects our predictions and disparity estimates. The results reveal significant inequities: White individuals hold a disproportionate share of properties and property value relative to their population, while Black, Hispanic, and Asian communities are underrepresented as property owners. These disparities are most pronounced in minority-majority neighborhoods, where ownership is predominantly White despite a predominantly non-White population. Corporate ownership (LLCs, trusts, etc.) exacerbates these gaps by reducing owner-occupied opportunities in urban minority communities. We provide a breakdown of ownership vs. population by race for majority-White, -Black, -Hispanic, and -Asian tracts, identify those with extreme ownership disparities, and compare patterns in urban, suburban, and rural contexts. The findings underscore persistent racial inequity in property ownership, reflecting broader historical and socio-economic forces, and highlight the importance of data-driven approaches to address these issues.
Authors: Fengyi Li, Kayhan Behdin, Natesh Pillai, Xiaofeng Wang, Zhipeng Wang, Ercan Yildiz
Abstract: Text segmentation based on the semantic meaning of sentences is a fundamental task with broad utility in many downstream applications. In this paper, we propose a graphical model-based unsupervised learning approach, named BP-Seg for efficient text segmentation. Our method not only considers local coherence, capturing the intuition that adjacent sentences are often more related, but also effectively groups sentences that are distant in the text yet semantically similar. This is achieved through belief propagation on the carefully constructed graphical models. Experimental results on both an illustrative example and a dataset with long-form documents demonstrate that our method performs favorably compared to competing approaches.
Authors: Ahmed Heakl, Sarim Hashmi, Gustavo Bertolo Stahl, Seung Hun Eddie Han, Salman Khan, Abdulrahman Mahmoud
Abstract: We introduce \texttt{CASS}, the first large-scale dataset and model suite for cross-architecture GPU code transpilation, targeting both source-level (CUDA~$\leftrightarrow$~HIP) and assembly-level (Nvidia SASS~$\leftrightarrow$~AMD RDNA3) translation. The dataset comprises 70k verified code pairs across host and device, addressing a critical gap in low-level GPU code portability. Leveraging this resource, we train the \texttt{CASS} family of domain-specific language models, achieving 95\% source translation accuracy and 37.5\% assembly translation accuracy, substantially outperforming commercial baselines such as GPT-4o, Claude, and Hipify. Our generated code matches native performance in over 85\% of test cases, preserving runtime and memory behavior. To support rigorous evaluation, we introduce \texttt{CASS-Bench}, a curated benchmark spanning 16 GPU domains with ground-truth execution. All data, models, and evaluation tools are released as open source to foster progress in GPU compiler tooling, binary compatibility, and LLM-guided hardware translation. Dataset and benchmark are on \href{https://huggingface.co/datasets/MBZUAI/cass}{\textcolor{blue}{HuggingFace}}, with code at \href{https://github.com/GustavoStahl/CASS}{\textcolor{blue}{GitHub}}.
URLs: https://huggingface.co/datasets/MBZUAI/cass, https://github.com/GustavoStahl/CASS
Authors: Moru Liu, Hao Dong, Jessica Kelly, Olga Fink, Mario Trapp
Abstract: Out-of-distribution (OOD) detection and segmentation are crucial for deploying machine learning models in safety-critical applications such as autonomous driving and robot-assisted surgery. While prior research has primarily focused on unimodal image data, real-world applications are inherently multimodal, requiring the integration of multiple modalities for improved OOD detection. A key challenge is the lack of supervision signals from unknown data, leading to overconfident predictions on OOD samples. To address this challenge, we propose Feature Mixing, an extremely simple and fast method for multimodal outlier synthesis with theoretical support, which can be further optimized to help the model better distinguish between in-distribution (ID) and OOD data. Feature Mixing is modality-agnostic and applicable to various modality combinations. Additionally, we introduce CARLA-OOD, a novel multimodal dataset for OOD segmentation, featuring synthetic OOD objects across diverse scenes and weather conditions. Extensive experiments on SemanticKITTI, nuScenes, CARLA-OOD datasets, and the MultiOOD benchmark demonstrate that Feature Mixing achieves state-of-the-art performance with a $10 \times$ to $370 \times$ speedup. Our source code and dataset will be available at https://github.com/mona4399/FeatureMixing.
Authors: Guillem Bras\'o, Aljo\v{s}a O\v{s}ep, Laura Leal-Taix\'e
Abstract: Uniform downsampling remains the de facto standard for reducing spatial resolution in vision backbones. In this work, we propose an alternative design built around a content-aware spatial grouping layer, that dynamically assigns tokens to a reduced set based on image boundaries and their semantic content. Stacking our grouping layer across consecutive backbone stages results in hierarchical segmentation that arises natively in the feature extraction process, resulting in our coined Native Segmentation Vision Transformer. We show that a careful design of our architecture enables the emergence of strong segmentation masks solely from grouping layers, that is, without additional segmentation-specific heads. This sets the foundation for a new paradigm of native, backbone-level segmentation, which enables strong zero-shot results without mask supervision, as well as a minimal and efficient standalone model design for downstream segmentation tasks. Our project page is https://research.nvidia.com/labs/dvl/projects/native-segmentation.
URLs: https://research.nvidia.com/labs/dvl/projects/native-segmentation.
Authors: Simmaco Di Lillo
Abstract: This work investigates the expected number of critical points of random neural networks with different activation functions as the depth increases in the infinite-width limit. Under suitable regularity conditions, we derive precise asymptotic formulas for the expected number of critical points of fixed index and those exceeding a given threshold. Our analysis reveals three distinct regimes depending on the value of the first derivative of the covariance evaluated at 1: the expected number of critical points may converge, grow polynomially, or grow exponentially with depth. The theoretical predictions are supported by numerical experiments. Moreover, we provide numerical evidence suggesting that, when the regularity condition is not satisfied (e.g. for neural networks with ReLU as activation function), the number of critical points increases as the map resolution increases, indicating a potential divergence in the number of critical points.
Authors: Nanda H. Krishna, Colin Bredenberg, Daniel Levenstein, Blake A. Richards, Guillaume Lajoie
Abstract: During periods of quiescence, such as sleep, neural activity in many brain circuits resembles that observed during periods of task engagement. However, the precise conditions under which task-optimized networks can autonomously reactivate the same network states responsible for online behavior is poorly understood. In this study, we develop a mathematical framework that outlines sufficient conditions for the emergence of neural reactivation in circuits that encode features of smoothly varying stimuli. We demonstrate mathematically that noisy recurrent networks optimized to track environmental state variables using change-based sensory information naturally develop denoising dynamics, which, in the absence of input, cause the network to revisit state configurations observed during periods of online activity. We validate our findings using numerical experiments on two canonical neuroscience tasks: spatial position estimation based on self-motion cues, and head direction estimation based on angular velocity cues. Overall, our work provides theoretical support for modeling offline reactivation as an emergent consequence of task optimization in noisy neural circuits.
Authors: Chengzhuo Tong, Ziyu Guo, Renrui Zhang, Wenyu Shan, Xinyu Wei, Zhenghao Xing, Hongsheng Li, Pheng-Ann Heng
Abstract: Recent advancements underscore the significant role of Reinforcement Learning (RL) in enhancing the Chain-of-Thought (CoT) reasoning capabilities of large language models (LLMs). Two prominent RL algorithms, Direct Preference Optimization (DPO) and Group Relative Policy Optimization (GRPO), are central to these developments, showcasing different pros and cons. Autoregressive image generation, also interpretable as a sequential CoT reasoning process, presents unique challenges distinct from LLM-based CoT reasoning. These encompass ensuring text-image consistency, improving image aesthetic quality, and designing sophisticated reward models, rather than relying on simpler rule-based rewards. While recent efforts have extended RL to this domain, these explorations typically lack an in-depth analysis of the domain-specific challenges and the characteristics of different RL strategies. To bridge this gap, we provide the first comprehensive investigation of the GRPO and DPO algorithms in autoregressive image generation, evaluating their in-domain performance and out-of-domain generalization, while scrutinizing the impact of different reward models on their respective capabilities. Our findings reveal that GRPO and DPO exhibit distinct advantages, and crucially, that reward models possessing stronger intrinsic generalization capabilities potentially enhance the generalization potential of the applied RL algorithms. Furthermore, we systematically explore three prevalent scaling strategies to enhance both their in-domain and out-of-domain proficiency, deriving unique insights into efficiently scaling performance for each paradigm. We hope our study paves a new path for inspiring future work on developing more effective RL algorithms to achieve robust CoT reasoning in the realm of autoregressive image generation. Code is released at https://github.com/ZiyuGuo99/Image-Generation-CoT
Authors: Chengqi Duan, Rongyao Fang, Yuqing Wang, Kun Wang, Linjiang Huang, Xingyu Zeng, Hongsheng Li, Xihui Liu
Abstract: Visual generation models have made remarkable progress in creating realistic images from text prompts, yet struggle with complex prompts that specify multiple objects with precise spatial relationships and attributes. Effective handling of such prompts requires explicit reasoning about the semantic content and spatial layout. We present GoT-R1, a framework that applies reinforcement learning to enhance semantic-spatial reasoning in visual generation. Building upon the Generation Chain-of-Thought approach, GoT-R1 enables models to autonomously discover effective reasoning strategies beyond predefined templates through carefully designed reinforcement learning. To achieve this, we propose a dual-stage multi-dimensional reward framework that leverages MLLMs to evaluate both the reasoning process and final output, enabling effective supervision across the entire generation pipeline. The reward system assesses semantic alignment, spatial accuracy, and visual quality in a unified approach. Experimental results demonstrate significant improvements on T2I-CompBench benchmark, particularly in compositional tasks involving precise spatial relationships and attribute binding. GoT-R1 advances the state-of-the-art in image generation by successfully transferring sophisticated reasoning capabilities to the visual generation domain. To facilitate future research, we make our code and pretrained models publicly available at https://github.com/gogoduan/GoT-R1.
Authors: Saber Malekmohammadi, Kiarash Shaloudegi, Zeou Hu, Yaoliang Yu
Abstract: Over the past few years, the federated learning ($\texttt{FL}$) community has witnessed a proliferation of new $\texttt{FL}$ algorithms. However, our understating of the theory of $\texttt{FL}$ is still fragmented, and a thorough, formal comparison of these algorithms remains elusive. Motivated by this gap, we show that many of the existing $\texttt{FL}$ algorithms can be understood from an operator splitting point of view. This unification allows us to compare different algorithms with ease, to refine previous convergence results and to uncover new algorithmic variants. In particular, our analysis reveals the vital role played by the step size in $\texttt{FL}$ algorithms. The unification also leads to a streamlined and economic way to accelerate $\texttt{FL}$ algorithms, without incurring any communication overhead. We perform numerical experiments on both convex and nonconvex models to validate our findings.
Authors: Hossein Zakerinia, Shayan Talaei, Giorgi Nadiradze, Dan Alistarh
Abstract: Federated Learning (FL) enables large-scale distributed training of machine learning models, while still allowing individual nodes to maintain data locally. However, executing FL at scale comes with inherent practical challenges: 1) heterogeneity of the local node data distributions, 2) heterogeneity of node computational speeds (asynchrony), but also 3) constraints in the amount of communication between the clients and the server. In this work, we present the first variant of the classic federated averaging (FedAvg) algorithm which, at the same time, supports data heterogeneity, partial client asynchrony, and communication compression. Our algorithm comes with a novel, rigorous analysis showing that, in spite of these system relaxations, it can provide similar convergence to FedAvg in interesting parameter regimes. Experimental results in the rigorous LEAF benchmark on setups of up to 300 nodes show that our algorithm ensures fast convergence for standard federated tasks, improving upon prior quantized and asynchronous approaches.
Authors: Peng Lu, Ahmad Rashid, Ivan Kobyzev, Mehdi Rezagholizadeh, Philippe Langlais
Abstract: Regularization techniques are crucial to improving the generalization performance and training efficiency of deep neural networks. Many deep learning algorithms rely on weight decay, dropout, batch/layer normalization to converge faster and generalize. Label Smoothing (LS) is another simple, versatile and efficient regularization which can be applied to various supervised classification tasks. Conventional LS, however, regardless of the training instance assumes that each non-target class is equally likely. In this work, we present a general framework for training with label regularization, which includes conventional LS but can also model instance-specific variants. Based on this formulation, we propose an efficient way of learning LAbel regularization by devising a Bi-level Optimization (LABO) problem. We derive a deterministic and interpretable solution of the inner loop as the optimal label smoothing without the need to store the parameters or the output of a trained model. Finally, we conduct extensive experiments and demonstrate our LABO consistently yields improvement over conventional label regularization on various fields, including seven machine translation and three image classification tasks across various
Authors: Yong Qing, Ke Li, Peng-Fei Zhou, Shi-Ju Ran
Abstract: Neural network (NN) designed for challenging machine learning tasks is in general a highly nonlinear mapping that contains massive variational parameters. High complexity of NN, if unbounded or unconstrained, might unpredictably cause severe issues including \R{overfitting}, loss of generalization power, and unbearable cost of hardware. In this work, we propose a general compression scheme that significantly reduces the variational parameters of NN's, despite of their specific types (linear, convolutional, \textit{etc}), by encoding them to deep \R{automatically differentiable} tensor network (ADTN) that contains exponentially-fewer free parameters. Superior compression performance of our scheme is demonstrated on several widely-recognized NN's (FC-2, LeNet-5, AlextNet, ZFNet and VGG-16) and datasets (MNIST, CIFAR-10 and CIFAR-100). For instance, we compress two linear layers in VGG-16 with approximately $10^{7}$ parameters to two ADTN's with just 424 parameters, improving the testing accuracy on CIFAR-10 from $90.17\%$ to $91.74\%$. We argue that the deep structure of ADTN is an essential reason for the remarkable compression performance of ADTN, compared to existing compression schemes that are mainly based on tensor decompositions/factorization and shallow tensor networks. Our work suggests deep TN as an exceptionally efficient mathematical structure for representing the variational parameters of NN's, which exhibits superior compressibility over the commonly-used matrices and multi-way arrays.
Authors: Wenjian Hao, Paulo C. Heredia, Shaoshuai Mou
Abstract: This paper presents a data-driven method for finding a closed-loop optimal controller, which minimizes a specified infinite-horizon cost function for systems with unknown dynamics given any arbitrary initial state. Suppose the closed-loop optimal controller can be parameterized by a given class of functions, hereafter referred to as the policy. The proposed method introduces a novel gradient estimation framework, which approximates the gradient of the cost function with respect to the policy parameters via integrating the Koopman operator with the classical concept of actor-critic. This enables the policy parameters to be tuned iteratively using gradient descent to achieve an optimal controller, leveraging the linearity of the Koopman operator. The convergence analysis of the proposed framework is provided. The effectiveness of the method is demonstrated through comparisons with a model-free reinforcement learning approach, and its control performance is further evaluated through simulations against model-based optimal control methods that solve the same optimal control problem utilizing the exact system dynamics.
Authors: Camila Kolling, Till Speicher, Vedant Nanda, Mariya Toneva, Krishna P. Gummadi
Abstract: Machine learning (ML) algorithms can often exhibit discriminatory behavior, negatively affecting certain populations across protected groups. To address this, numerous debiasing methods, and consequently evaluation measures, have been proposed. Current evaluation measures for debiasing methods suffer from two main limitations: (1) they primarily provide a global estimate of unfairness, failing to provide a more fine-grained analysis, and (2) they predominantly analyze the model output on a specific task, failing to generalize the findings to other tasks. In this work, we introduce Pointwise Normalized Kernel Alignment (PNKA), a pointwise representational similarity measure that addresses these limitations by measuring how debiasing measures affect the intermediate representations of individuals. On tabular data, the use of PNKA reveals previously unknown insights: while group fairness predominantly influences a small subset of the population, maintaining high representational similarity for the majority, individual fairness constraints uniformly impact representations across the entire population, altering nearly every data point. We show that by evaluating representations using PNKA, we can reliably predict the behavior of ML models trained on these representations. Moreover, applying PNKA to language embeddings shows that existing debiasing methods may not perform as intended, failing to remove biases from stereotypical words and sentences. Our findings suggest that current evaluation measures for debiasing methods are insufficient, highlighting the need for a deeper understanding of the effects of debiasing methods, and show how pointwise representational similarity metrics can help with fairness audits.
Authors: Binghui Li, Yuanzhi Li
Abstract: Similar to surprising performance in the standard deep learning, deep nets trained by adversarial training also generalize well for unseen clean data (natural data). However, despite adversarial training can achieve low robust training error, there exists a significant robust generalization gap. We call this phenomenon the Clean Generalization and Robust Overfitting (CGRO). In this work, we study the CGRO phenomenon in adversarial training from two views: representation complexity and training dynamics. Specifically, we consider a binary classification setting with $N$ separated training data points. First, we prove that, based on the assumption that we assume there is $\operatorname{poly}(D)$-size clean classifier (where $D$ is the data dimension), ReLU net with only $O(N D)$ extra parameters is able to leverages robust memorization to achieve the CGRO, while robust classifier still requires exponential representation complexity in worst case. Next, we focus on a structured-data case to analyze training dynamics, where we train a two-layer convolutional network with $O(N D)$ width against adversarial perturbation. We then show that a three-stage phase transition occurs during learning process and the network provably converges to robust memorization regime, which thereby results in the CGRO. Besides, we also empirically verify our theoretical analysis by experiments in real-image recognition datasets.
Authors: Peng Wang, Xiao Li, Can Yaras, Zhihui Zhu, Laura Balzano, Wei Hu, Qing Qu
Abstract: Over the past decade, deep learning has proven to be a highly effective tool for learning meaningful features from raw data. However, it remains an open question how deep networks perform hierarchical feature learning across layers. In this work, we attempt to unveil this mystery by investigating the structures of intermediate features. Motivated by our empirical findings that linear layers mimic the roles of deep layers in nonlinear networks for feature learning, we explore how deep linear networks transform input data into output by investigating the output (i.e., features) of each layer after training in the context of multi-class classification problems. Toward this goal, we first define metrics to measure within-class compression and between-class discrimination of intermediate features, respectively. Through theoretical analysis of these two metrics, we show that the evolution of features follows a simple and quantitative pattern from shallow to deep layers when the input data is nearly orthogonal and the network weights are minimum-norm, balanced, and approximate low-rank: Each layer of the linear network progressively compresses within-class features at a geometric rate and discriminates between-class features at a linear rate with respect to the number of layers that data have passed through. To the best of our knowledge, this is the first quantitative characterization of feature evolution in hierarchical representations of deep linear networks. Empirically, our extensive experiments not only validate our theoretical results numerically but also reveal a similar pattern in deep nonlinear networks which aligns well with recent empirical studies. Moreover, we demonstrate the practical implications of our results in transfer learning. Our code is available at https://github.com/Heimine/PNC_DLN.
Authors: Biyan Zhou, Pao-Sheng Vincent Sun, Arindam Basu
Abstract: While it is important to make implantable brain-machine interfaces (iBMI) wireless to increase patient comfort and safety, the trend of increased channel count in recent neural probes poses a challenge due to the concomitant increase in the data rate. Extracting information from raw data at the source by using edge computing is a promising solution to this problem, with integrated intention decoders providing the best compression ratio. Recent benchmarking efforts have shown recurrent neural networks to be the best solution. Spiking Neural Networks (SNN) emerge as a promising solution for resource efficient neural decoding while Long Short Term Memory (LSTM) networks achieve the best accuracy. In this work, we show that combining traditional signal processing techniques, namely signal filtering, with SNNs improve their decoding performance significantly for regression tasks, closing the gap with LSTMs, at little added cost. Results with different filters are shown with Bessel filters providing best performance. Two block-bidirectional Bessel filters have been used--one for low latency and another for high accuracy. Adding the high accuracy variant of the Bessel filters to the output of ANN, SNN and variants provided statistically significant benefits with maximum gains of $\approx 5\%$ and $8\%$ in $R^2$ for two SNN topologies (SNN\_Streaming and SNN\_3D). Our work presents state of the art results for this dataset and paves the way for decoder-integrated-implants of the future.
Authors: Fabio De Sousa Ribeiro, Ben Glocker
Abstract: Despite the growing interest in diffusion models, gaining a deep understanding of the model class remains an elusive endeavour, particularly for the uninitiated in non-equilibrium statistical physics. Thanks to the rapid rate of progress in the field, most existing work on diffusion models focuses on either applications or theoretical contributions. Unfortunately, the theoretical material is often inaccessible to practitioners and new researchers, leading to a risk of superficial understanding in ongoing research. Given that diffusion models are now an indispensable tool, a clear and consolidating perspective on the model class is needed to properly contextualize recent advances in generative modelling and lower the barrier to entry for new researchers. To that end, we revisit predecessors to diffusion models like hierarchical latent variable models and synthesize a holistic perspective using only directed graphical modelling and variational inference principles. The resulting narrative is easier to follow as it imposes fewer prerequisites on the average reader relative to the view from non-equilibrium thermodynamics or stochastic differential equations.
Authors: Cheng Yang, Jixi Liu, Yunhe Yan, Chuan Shi
Abstract: Despite the remarkable success of graph neural networks (GNNs) in modeling graph-structured data, like other machine learning models, GNNs are also susceptible to making biased predictions based on sensitive attributes, such as race and gender. For fairness consideration, recent state-of-the-art (SOTA) methods propose to filter out sensitive information from inputs or representations, e.g., edge dropping or feature masking. However, we argue that such filtering-based strategies may also filter out some non-sensitive feature information, leading to a sub-optimal trade-off between predictive performance and fairness. To address this issue, we unveil an innovative neutralization-based paradigm, where additional Fairness-facilitating Features (F3) are incorporated into node features or representations before message passing. The F3 are expected to statistically neutralize the sensitive bias in node representations and provide additional nonsensitive information. We also provide theoretical explanations for our rationale, concluding that F3 can be realized by emphasizing the features of each node's heterogeneous neighbors (neighbors with different sensitive attributes). We name our method as FairSIN, and present three implementation variants from both data-centric and model-centric perspectives. Experimental results on five benchmark datasets with three different GNN backbones show that FairSIN significantly improves fairness metrics while maintaining high prediction accuracies.
Authors: Arash Badie-Modiri, Chiara Boldrini, Lorenzo Valerio, J\'anos Kert\'esz, M\'arton Karsai
Abstract: Fully decentralised federated learning enables collaborative training of individual machine learning models on a distributed network of communicating devices while keeping the training data localised on each node. This approach avoids central coordination, enhances data privacy and eliminates the risk of a single point of failure. Our research highlights that the effectiveness of decentralised federated learning is significantly influenced by the network topology of connected devices and the learning models' initial conditions. We propose a strategy for uncoordinated initialisation of the artificial neural networks based on the distribution of eigenvector centralities of the underlying communication network, leading to a radically improved training efficiency. Additionally, our study explores the scaling behaviour and the choice of environmental parameters under our proposed initialisation strategy. This work paves the way for more efficient and scalable artificial neural network training in a distributed and uncoordinated environment, offering a deeper understanding of the intertwining roles of network structure and learning dynamics.
Authors: Akira Kitaoka
Abstract: We consider the inverse optimization problem of estimating the weights of the objective function such that the given solution is an optimal solution for a mixed integer linear program (MILP). In this inverse optimization problem, the known methods exhibit inefficient convergence. Specifically, if $d$ denotes the dimension of the weights and $k$ the number of iterations, then the error of the weights is bounded by $O(k^{-1/(d-1)})$, leading to slow convergence as $d$ increases.We propose a projected subgradient method with a step size of $k^{-1/2}$ based on suboptimality loss. We theoretically show and demonstrate that the proposed method efficiently learns the weights. In particular, we show that there exists a constant $\gamma > 0$ such that the distance between the learned and true weights is bounded by $ O\left(k^{-1/(1+\gamma)} \exp\left(-\frac{\gamma k^{1/2}}{2+\gamma}\right)\right), $ or the optimal solution is exactly recovered. Furthermore, experiments demonstrate that the proposed method solves the inverse optimization problems of MILP using fewer than $1/7$ the number of MILP calls required by known methods, and converges within a finite number of iterations.
Authors: Jerry Huang
Abstract: Long sequences occur in abundance within real-world scenarios, hence properly modelling them opens numerous down-stream use-cases. Deep neural networks, however, have often struggled with these for a variety of reasons. Recent advances, both in system engineering as well as model design, have enabled the scaling up of model that are purported to support extended context length. In particular, the state-space and linear recurrent neural network families of models hypothetically can entend to infinite sequence lenth. However, is this too good to be true? We conduct an evaluation to show that while such claims may be sound theoretically, there remain large practical gaps that are empirically observed. In particular, recurrent models still suffer in the same settings as long-context LLMs with attention. We further show that different inductive biases have inconsistent extrapolation capabilities, highlighting the need to further study such paradigms and investigate why long-context models seemingly fail to behave as one might expect.
Authors: Emanuele Carlini, Domenico Di Gangi, Vinicius Monteiro de Lira, Hanna Kavalionak, Amilcar Soares, Gabriel Spadon
Abstract: Seaports play a crucial role in the global economy, and researchers have sought to understand their significance through various studies. In this paper, we aim to explore the common characteristics shared by important ports by analyzing the network of connections formed by vessel movement among them. To accomplish this task, we adopt a bottom-up network construction approach that combines three years' worth of AIS (Automatic Identification System) data from around the world, constructing a Ports Network that represents the connections between different ports. Through this representation, we utilize machine learning to assess the relative significance of various port features. Our model examined such features and revealed that geographical characteristics and the port's depth are indicators of a port's importance to the Ports Network. Accordingly, this study employs a data-driven approach and utilizes machine learning to provide a comprehensive understanding of the factors contributing to the extent of ports. Our work aims to inform decision-making processes related to port development, resource allocation, and infrastructure planning within the industry.
Authors: Andrija Djurisic, Rosanne Liu, Mladen Nikolic
Abstract: The safe deployment of machine learning and AI models in open-world settings hinges critically on the ability to detect out-of-distribution (OOD) data accurately, data samples that contrast vastly from what the model was trained with. Current approaches to OOD detection often require further training the model, and/or statistics about the training data which may no longer be accessible. Additionally, many existing OOD detection methods struggle to maintain performance when transferred across different architectures. Our research tackles these issues by proposing a simple, post-hoc method that does not require access to the training data distribution, keeps a trained network intact, and holds strong performance across a variety of architectures. Our method, Logit Scaling (LTS), as the name suggests, simply scales the logits in a manner that effectively distinguishes between in-distribution (ID) and OOD samples. We tested our method on benchmarks across various scales, including CIFAR-10, CIFAR-100, ImageNet and OpenOOD. The experiments cover 3 ID and 14 OOD datasets, as well as 9 model architectures. Overall, we demonstrate state-of-the-art performance, robustness and adaptability across different architectures, paving the way towards a universally applicable solution for advanced OOD detection.
Authors: Maria-Florina Balcan, Anh Tuan Nguyen, Dravyansh Sharma
Abstract: Data-driven algorithm design automatically adapts algorithms to specific application domains, achieving better performance. In the context of parameterized algorithms, this approach involves tuning the algorithm's hyperparameters using problem instances drawn from the problem distribution of the target application domain. This can be achieved by maximizing empirical utilities that measure the algorithms' performance as a function of their hyperparameters, using problem instances. While empirical evidence supports the effectiveness of data-driven algorithm design, providing theoretical guarantees for several parameterized families remains challenging. This is due to the intricate behaviors of their corresponding utility functions, which typically admit piecewise discontinuous structures. In this work, we present refined frameworks for providing learning guarantees for parameterized data-driven algorithm design problems in both distributional and online learning settings. For the distributional learning setting, we introduce the \textit{Pfaffian GJ framework}, an extension of the classical \textit{GJ framework}, that is capable of providing learning guarantees for function classes for which the computation involves Pfaffian functions. Unlike the GJ framework, which is limited to function classes with computation characterized by rational functions, our proposed framework can deal with function classes involving Pfaffian functions, which are much more general and widely applicable. We then show that for many parameterized algorithms of interest, their utility function possesses a \textit{refined piecewise structure}, which automatically translates to learning guarantees using our proposed framework.
Authors: Haochen Yuan, Yutong Wang, Yihong Chen, Yunbo Wang, Xiaokang Yang
Abstract: Time series forecasting, particularly in few-shot learning scenarios, is challenging due to the limited availability of high-quality training data. To address this, we present a pilot study on using reinforcement learning (RL) for time series data augmentation. Our method, ReAugment, tackles three critical questions: which parts of the training set should be augmented, how the augmentation should be performed, and what advantages RL brings to the process. Specifically, our approach maintains a forecasting model zoo, and by measuring prediction diversity across the models, we identify samples with higher probabilities for overfitting and use them as the anchor points for augmentation. Leveraging RL, our method adaptively transforms the overfit-prone samples into new data that not only enhances training set diversity but also directs the augmented data to target regions where the forecasting models are prone to overfitting. We validate the effectiveness of ReAugment across a wide range of base models, showing its advantages in both standard time series forecasting and few-shot learning tasks.
Authors: Hayden Helm, Aranyak Acharyya, Brandon Duderstadt, Youngser Park, Carey E. Priebe
Abstract: Generative models are capable of producing human-expert level content across a variety of topics and domains. As the impact of generative models grows, it is necessary to develop statistical methods to understand collections of available models. These methods are particularly important in settings where the user may not have access to information related to a model's pre-training data, weights, or other relevant model-level covariates. In this paper we extend recent results on representations of black-box generative models to model-level statistical inference tasks. We demonstrate that the model-level representations are effective for multiple inference tasks.
Authors: Jaeyoung Park, Minsu Kim, Steven Euijong Whang
Abstract: Model fairness is becoming important in class-incremental learning for Trustworthy AI. While accuracy has been a central focus in class-incremental learning, fairness has been relatively understudied. However, naively using all the samples of the current task for training results in unfair catastrophic forgetting for certain sensitive groups including classes. We theoretically analyze that forgetting occurs if the average gradient vector of the current task data is in an "opposite direction" compared to the average gradient vector of a sensitive group, which means their inner products are negative. We then propose a fair class-incremental learning framework that adjusts the training weights of current task samples to change the direction of the average gradient vector and thus reduce the forgetting of underperforming groups and achieve fairness. For various group fairness measures, we formulate optimization problems to minimize the overall losses of sensitive groups while minimizing the disparities among them. We also show the problems can be solved with linear programming and propose an efficient Fairness-aware Sample Weighting (FSW) algorithm. Experiments show that FSW achieves better accuracy-fairness tradeoff results than state-of-the-art approaches on real datasets.
Authors: Shoaib Ahmed Siddiqui, Radhika Gaonkar, Boris K\"opf, David Krueger, Andrew Paverd, Ahmed Salem, Shruti Tople, Lukas Wutschitz, Menglin Xia, Santiago Zanella-B\'eguelin
Abstract: Large Language Models (LLMs) are rapidly becoming commodity components of larger software systems. This poses natural security and privacy problems: poisoned data retrieved from one component can change the model's behavior and compromise the entire system, including coercing the model to spread confidential data to untrusted components. One promising approach is to tackle this problem at the system level via dynamic information flow (aka taint) tracking. Unfortunately, this approach of propagating the most restrictive input label to the output is too conservative for applications where LLMs operate on inputs retrieved from diverse sources. In this paper, we propose a novel, more permissive approach to propagate information flow labels through LLM queries. The key idea behind our approach is to propagate only the labels of the samples that were influential in generating the model output and to eliminate the labels of unnecessary inputs. We implement and investigate the effectiveness of two variations of this approach, based on (i) prompt-based retrieval augmentation, and (ii) a $k$-nearest-neighbors language model. We compare these with a baseline that uses introspection to predict the output label. Our experimental results in an LLM agent setting show that the permissive label propagator improves over the baseline in more than 85% of the cases, which underscores the practicality of our approach.
Authors: Junpeng Yue, Xinrun Xu, B\"orje F. Karlsson, Zongqing Lu
Abstract: MLLM agents demonstrate potential for complex embodied tasks by retrieving multimodal task-relevant trajectory data. However, current retrieval methods primarily focus on surface-level similarities of textual or visual cues in trajectories, neglecting their effectiveness for the specific task at hand. To address this issue, we propose a novel method, MLLM As ReTriever (MART), which enhances the performance of embodied agents by utilizing interaction data to fine-tune an MLLM retriever based on preference learning, such that the retriever fully considers the effectiveness of trajectories and prioritizes them for unseen tasks. We also introduce Trajectory Abstraction, a mechanism that leverages MLLMs' summarization capabilities to represent trajectories with fewer tokens while preserving key information, enabling agents to better comprehend milestones in the trajectory. Experimental results across various environments demonstrate our method significantly improves task success rates in unseen scenes compared to baseline methods. This work presents a new paradigm for multimodal retrieval in embodied agents, by fine-tuning a general-purpose MLLM as the retriever to assess trajectory effectiveness. All the code for benchmark tasks, simulator modifications, and the MLLM retriever is available at https://github.com/PKU-RL/MART.
Authors: Max Weltevrede, Caroline Horsch, Matthijs T. J. Spaan, Wendelin B\"ohmer
Abstract: In the zero-shot policy transfer (ZSPT) setting for contextual Markov decision processes (MDP), agents train on a fixed set of contexts and must generalise to new ones. Recent work has argued and demonstrated that increased exploration can improve this generalisation, by training on more states in the training contexts. In this paper, we demonstrate that training on more states can indeed improve generalisation, but can come at a cost of reducing the accuracy of the learned value function which should not benefit generalisation. We hypothesise and demonstrate that using exploration to increase the agent's coverage while also increasing the accuracy improves generalisation even more. Inspired by this, we propose a method Explore-Go that implements an exploration phase at the beginning of each episode, which can be combined with existing on- and off-policy RL algorithms and significantly improves generalisation even in partially observable MDPs. We demonstrate the effectiveness of Explore-Go when combined with several popular algorithms and show an increase in generalisation performance across several environments. With this, we hope to provide practitioners with a simple modification that can improve the generalisation of their agents.
Authors: Udita Ghosh, Dripta S. Raychaudhuri, Jiachen Li, Konstantinos Karydis, Amit K. Roy-Chowdhury
Abstract: Offline imitation learning enables learning a policy solely from a set of expert demonstrations, without any environment interaction. To alleviate the issue of distribution shift arising due to the small amount of expert data, recent works incorporate large numbers of auxiliary demonstrations alongside the expert data. However, the performance of these approaches rely on assumptions about the quality and composition of the auxiliary data, and they are rarely successful when those assumptions do not hold. To address this limitation, we propose Robust Offline Imitation from Diverse Auxiliary Data (ROIDA). ROIDA first identifies high-quality transitions from the entire auxiliary dataset using a learned reward function. These high-reward samples are combined with the expert demonstrations for weighted behavioral cloning. For lower-quality samples, ROIDA applies temporal difference learning to steer the policy towards high-reward states, improving long-term returns. This two-pronged approach enables our framework to effectively leverage both high and low-quality data without any assumptions. Extensive experiments validate that ROIDA achieves robust and consistent performance across multiple auxiliary datasets with diverse ratios of expert and non-expert demonstrations. ROIDA effectively leverages unlabeled auxiliary data, outperforming prior methods reliant on specific data assumptions. Our code is available at https://github.com/uditaghosh/roida.
Authors: Wen Ye, Wei Yang, Defu Cao, Yizhou Zhang, Lumingyuan Tang, Jie Cai, Yan Liu
Abstract: Real-world time series inference requires more than point forecasting. It demands multi-step reasoning, constraint handling, domain knowledge incorporation, and domain-specific workflow assembly. Existing time series foundation models are limited to narrow tasks and lack flexibility to generalize across diverse scenarios. On the other hand, large language models (LLMs) struggle with numerical precision. To address these limitations, we introduce TS-Reasoner, a Domain-Oriented Time Series Agent that integrates natural language reasoning with precise numerical execution. TS-Reasoner decomposes natural language instructions into structured workflows composed of statistical, logical, and domain-specific operators, and incorporates a self-refinement mechanism for adaptive execution. We evaluate its capabilities through two axes: basic time series understanding and complex multi-step inference, using the TimeSeriesExam benchmark and a newly constructed dataset. Experimental results show that TS-Reasoner significantly outperforms general-purpose LLMs, highlighting the promise of domain-specialized agents for robust and interpretable time series reasoning.
Authors: Qi Liu, Wanjing Ma
Abstract: In this paper, we identify and analyze a recurring training loss pattern, which we term the \textit{Epochal Sawtooth Effect (ESE)}, commonly observed during training with adaptive gradient-based optimizers, particularly Adam optimizer. This pattern is characterized by a sharp drop in loss at the beginning of each epoch, followed by a gradual increase, resulting in a sawtooth-shaped loss curve. Through empirical observations, we demonstrate that while this effect is most pronounced with Adam, it persists, although less severely, with other optimizers such as RMSProp. We provide an in-depth explanation of the underlying mechanisms that lead to the Epochal Sawtooth Effect. The influences of factors like $\beta$, batch size, data shuffling on this pattern have been studied. We quantify the influence of $beta_2$ on the shape of the loss curve, showing that higher values of $\beta_2$ result in a nearly linear increase in loss, while lower values create a concave upward trend. Our analysis reveals that this behavior stems from the adaptive learning rate controlled by the second moment estimate, with $\beta_1$ playing a minimal role when $\beta_2$ is large. To support our analysis, we replicate this phenomenon through a controlled quadratic minimization task. By incrementally solving a series of quadratic optimization problems using Adam, we demonstrate that the Epochal Sawtooth Effect can emerge even in simple optimization scenarios, reinforcing the generality of this pattern. This paper provides both theoretical insights and quantitative analysis, offering a comprehensive understanding of this ubiquitous phenomenon in modern optimization techniques.
Authors: Zhaomin Wu, Jizhou Guo, Junyi Hou, Bingsheng He, Lixin Fan, Qiang Yang
Abstract: Prominent Large Language Model (LLM) services from providers like OpenAI and Google excel at general tasks but often underperform on domain-specific applications. Current customization services for these LLMs typically require users to upload data for fine-tuning, posing significant privacy risks. While differentially private (DP) data synthesis presents a potential alternative, its application commonly results in low effectiveness due to the introduction of excessive noise on data for DP. To overcome this, we introduce Llamdex, a novel framework that facilitates LLM customization as a service, where the client uploads pre-trained domain-specific models rather than data. This client-uploaded model, optionally protected by DP with much lower noise, is inserted into the base LLM via connection modules. Significantly, these connecting modules are trained without requiring sensitive domain data, enabling clients to customize LLM services while preserving data privacy. Experiments demonstrate that Llamdex improves domain-specific accuracy by up to 26\% over state-of-the-art private data synthesis methods under identical privacy constraints and, by obviating the need for users to provide domain context within queries, maintains inference efficiency comparable to the original LLM service.
Authors: Darin Tsui, Aryan Musharaf, Yigit Efe Erginbas, Justin Singh Kang, Amirali Aghazadeh
Abstract: The growing adoption of machine learning models for biological sequences has intensified the need for interpretable predictions, with Shapley values emerging as a theoretically grounded standard for model explanation. While effective for local explanations of individual input sequences, scaling Shapley-based interpretability to extract global biological insights requires evaluating thousands of sequences--incurring exponential computational cost per query. We introduce SHAP zero, a novel algorithm that amortizes the cost of Shapley value computation across large-scale biological datasets. After a one-time model sketching step, SHAP zero enables near-zero marginal cost for future queries by uncovering an underexplored connection between Shapley values, high-order feature interactions, and the sparse Fourier transform of the model. Applied to models of guide RNA efficacy, DNA repair outcomes, and protein fitness, SHAP zero explains predictions orders of magnitude faster than existing methods, recovering rich combinatorial interactions previously inaccessible at scale. This work opens the door to principled, efficient, and scalable interpretability for black-box sequence models in biology.
Authors: Nicol\'as Nieto, Simon B. Eickhoff, Christian Jung, Martin Reuter, Kersten Diers, Malte Kelm, Artur Lichtenberg, Federico Raimondo, Kaustubh R. Patil
Abstract: Machine learning (ML) models benefit from large datasets. Collecting data in biomedical domains is costly and challenging, hence, combining datasets has become a common practice. However, datasets obtained under different conditions could present undesired site-specific variability. Data harmonization methods aim to remove site-specific variance while retaining biologically relevant information. This study evaluates the effectiveness of popularly used ComBat-based methods for harmonizing data in scenarios where the class balance is not equal across sites. We find that these methods struggle with data leakage issues. To overcome this problem, we propose a novel approach PrettYharmonize, designed to harmonize data by pretending the target labels. We validate our approach using controlled datasets designed to benchmark the utility of harmonization. Finally, using real-world MRI and clinical data, we compare leakage-prone methods with PrettYharmonize and show that it achieves comparable performance while avoiding data leakage, particularly in site-target-dependence scenarios.
Authors: Quan Zhou, Changhua Pei, Fei Sun, Jing Han, Zhengwei Gao, Dan Pei, Haiming Zhang, Gaogang Xie, Jianhui Li
Abstract: Time series anomaly detection (TSAD) underpins real-time monitoring in cloud services and web systems, allowing rapid identification of anomalies to prevent costly failures. Most TSAD methods driven by forecasting models tend to overfit by emphasizing minor fluctuations. Our analysis reveals that effective TSAD should focus on modeling "normal" behavior through smooth local patterns. To achieve this, we reformulate time series modeling as approximating the series with smooth univariate functions. The local smoothness of each univariate function ensures that the fitted time series remains resilient against local disturbances. However, a direct KAN implementation proves susceptible to these disturbances due to the inherently localized characteristics of B-spline functions. We thus propose KAN-AD, replacing B-splines with truncated Fourier expansions and introducing a novel lightweight learning mechanism that emphasizes global patterns while staying robust to local disturbances. On four popular TSAD benchmarks, KAN-AD achieves an average 15% improvement in detection accuracy (with peaks exceeding 27%) over state-of-the-art baselines. Remarkably, it requires fewer than 1,000 trainable parameters, resulting in a 50% faster inference speed compared to the original KAN, demonstrating the approach's efficiency and practical viability.
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: Jincheng Huang, Yujie Mo, Xiaoshuang Shi, Lei Feng, Xiaofeng Zhu
Abstract: The message-passing mechanism of graph convolutional networks (i.e., GCNs) enables label information to reach more unlabeled neighbors, thereby increasing the utilization of labels. However, the additional label information does not always contribute positively to the GCN. To address this issue, we propose a new two-step framework called ELU-GCN. In the first stage, ELU-GCN conducts graph learning to learn a new graph structure (i.e., ELU-graph), which allows the additional label information to positively influence the predictions of GCN. In the second stage, we design a new graph contrastive learning on the GCN framework for representation learning by exploring the consistency and mutually exclusive information between the learned ELU graph and the original graph. Moreover, we theoretically demonstrate that the proposed method can ensure the generalization ability of GCNs. Extensive experiments validate the superiority of our method.
Authors: Marlon Tobaben, Talal Alrawajfeh, Marcus Klasson, Mikko Heikkil\"a, Arno Solin, Antti Honkela
Abstract: The goal of continual learning (CL) is to retain knowledge across tasks, but this conflicts with strict privacy required for sensitive training data that prevents storing or memorising individual samples. To address that, we combine CL and differential privacy (DP). We highlight that failing to account for privacy leakage through the set of labels a model can output can break the privacy of otherwise valid DP algorithms. This is especially relevant in CL. We show that mitigating the issue with a data-independent overly large label space can have minimal negative impact on utility when fine-tuning a pre-trained model under DP, while learning the labels with a separate DP mechanism risks losing small classes.
Authors: Joseph Liu, Joshua Geddes, Ziyu Guo, Haomiao Jiang, Mahesh Kumar Nandwana
Abstract: Diffusion Transformers (DiT) have emerged as powerful generative models for various tasks, including image, video, and speech synthesis. However, their inference process remains computationally expensive due to the repeated evaluation of resource-intensive attention and feed-forward modules. To address this, we introduce SmoothCache, a model-agnostic inference acceleration technique for DiT architectures. SmoothCache leverages the observed high similarity between layer outputs across adjacent diffusion timesteps. By analyzing layer-wise representation errors from a small calibration set, SmoothCache adaptively caches and reuses key features during inference. Our experiments demonstrate that SmoothCache achieves 8% to 71% speed up while maintaining or even improving generation quality across diverse modalities. We showcase its effectiveness on DiT-XL for image generation, Open-Sora for text-to-video, and Stable Audio Open for text-to-audio, highlighting its potential to enable real-time applications and broaden the accessibility of powerful DiT models.
Authors: Ivan Kralj, Lodovico Giaretta, Gordan Je\v{z}i\'c, Ivana Podnar \v{Z}arko, \v{S}ar\=unas Girdzijauskas
Abstract: In smart mobility, large networks of geographically distributed sensors produce vast amounts of high-frequency spatio-temporal data that must be processed in real time to avoid major disruptions. Traditional centralized approaches are increasingly unsuitable to this task, as they struggle to scale with expanding sensor networks, and reliability issues in central components can easily affect the whole deployment. To address these challenges, we explore and adapt semi-decentralized training techniques for Spatio-Temporal Graph Neural Networks (ST-GNNs) in smart mobility domain. We implement a simulation framework where sensors are grouped by proximity into multiple cloudlets, each handling a subgraph of the traffic graph, fetching node features from other cloudlets to train its own local ST-GNN model, and exchanging model updates with other cloudlets to ensure consistency, enhancing scalability and removing reliance on a centralized aggregator. We perform extensive comparative evaluation of four different ST-GNN training setups -- centralized, traditional FL, server-free FL, and Gossip Learning -- on large-scale traffic datasets, the METR-LA and PeMS-BAY datasets, for short-, mid-, and long-term vehicle speed predictions. Experimental results show that semi-decentralized setups are comparable to centralized approaches in performance metrics, while offering advantages in terms of scalability and fault tolerance. In addition, we highlight often overlooked issues in existing literature for distributed ST-GNNs, such as the variation in model performance across different geographical areas due to region-specific traffic patterns, and the significant communication overhead and computational costs that arise from the large receptive field of GNNs, leading to substantial data transfers and increased computation of partial embeddings.
Authors: Kale-ab Abebe Tessera, Arrasy Rahman, Amos Storkey, Stefano V. Albrecht
Abstract: Adaptability to specialised or homogeneous behaviours is critical in cooperative multi-agent reinforcement learning (MARL). Parameter sharing (PS) techniques, common for efficient adaptation, often limit behavioural diversity due to cross-agent gradient interference, which we show can be exacerbated by the coupling of observations and agent IDs. Current remedies typically add complexity through altered objectives, manual preset diversity levels, or sequential updates. We ask: can shared policies adapt without these complexities? We propose HyperMARL, a PS approach using hypernetworks for dynamic agent-specific parameters, without altering the RL objective or requiring preset diversity levels. HyperMARL's explicit decoupling of observation- and agent-conditioned gradients empirically reduces policy gradient variance, facilitates shared-policy adaptation (including specialisation), and helps mitigate cross-agent interference. Across diverse MARL benchmarks (up to 20 agents), requiring homogeneous, heterogeneous, or mixed behaviours, HyperMARL achieves competitive performance against key baselines -- fully shared, non-parameter sharing, and three diversity-promoting methods -- while preserving behavioural diversity comparable to non-parameter sharing. These findings establish HyperMARL as a versatile approach for adaptive MARL. The code is publicly available at https://github.com/KaleabTessera/HyperMARL.
Authors: Jiahai Feng, Stuart Russell, Jacob Steinhardt
Abstract: Pretrained language models (LMs) can generalize to implications of facts that they are finetuned on. For example, if finetuned on ``John Doe lives in Tokyo," LMs can correctly answer ``What language do the people in John Doe's city speak?'' with ``Japanese''. However, little is known about the mechanisms that enable this generalization or how they are learned during pretraining. We introduce extractive structures as a framework for describing how components in LMs (e.g., MLPs or attention heads) coordinate to enable this generalization. The structures consist of informative components that store training facts as weight changes, and upstream and downstream extractive components that query and process the stored information to produce the correct implication. We hypothesize that extractive structures are learned during pretraining when encountering implications of previously known facts. This yields two predictions: a data ordering effect where extractive structures can be learned only if facts precede their implications, and a weight grafting effect where extractive structures can be transferred to predict counterfactual implications. We empirically demonstrate these phenomena in the OLMo-7b, Llama 3-8b, Gemma 2-9b, and Qwen 2-7b models. Of independent interest, our results also indicate that fact learning can occur at both early and late layers, which lead to different forms of generalization.
Authors: Yinsicheng Jiang, Yao Fu, Yeqi Huang, Ping Nie, Zhan Lu, Leyang Xue, Congjie He, Man-Kit Sit, Jilong Xue, Li Dong, Ziming Miao, Dayou Du, Tairan Xu, Kai Zou, Edoardo Ponti, Luo Mai
Abstract: The sparse Mixture-of-Experts (MoE) architecture is increasingly favored for scaling Large Language Models (LLMs) efficiently, but it depends on heterogeneous compute and memory resources. These factors jointly affect system Cost, Accuracy, and Performance (CAP), making trade-offs inevitable. Existing benchmarks often fail to capture these trade-offs accurately, complicating practical deployment decisions. To address this, we introduce MoE-CAP, a benchmark specifically designed for MoE systems. Our analysis reveals that achieving an optimal balance across CAP is difficult with current hardware; MoE systems typically optimize two of the three dimensions at the expense of the third-a dynamic we term the MoE-CAP trade-off. To visualize this, we propose the CAP Radar Diagram. We further introduce sparsity-aware performance metrics-Sparse Memory Bandwidth Utilization (S-MBU) and Sparse Model FLOPS Utilization (S-MFU)-to enable accurate performance benchmarking of MoE systems across diverse hardware platforms and deployment scenarios.
Authors: Ya-Wei Eileen Lin, Ronald R. Coifman, Gal Mishne, Ronen Talmon
Abstract: High-dimensional data often exhibit hierarchical structures in both modes: samples and features. Yet, most existing approaches for hierarchical representation learning consider only one mode at a time. In this work, we propose an unsupervised method for jointly learning hierarchical representations of samples and features via Tree-Wasserstein Distance (TWD). Our method alternates between the two data modes. It first constructs a tree for one mode, then computes a TWD for the other mode based on that tree, and finally uses the resulting TWD to build the second mode's tree. By repeatedly alternating through these steps, the method gradually refines both trees and the corresponding TWDs, capturing meaningful hierarchical representations of the data. We provide a theoretical analysis showing that our method converges. We show that our method can be integrated into hyperbolic graph convolutional networks as a pre-processing technique, improving performance in link prediction and node classification tasks. In addition, our method outperforms baselines in sparse approximation and unsupervised Wasserstein distance learning tasks on word-document and single-cell RNA-sequencing datasets.
Authors: Artavazd Maranjyan, Alexander Tyurin, Peter Richt\'arik
Abstract: Asynchronous Stochastic Gradient Descent (Asynchronous SGD) is a cornerstone method for parallelizing learning in distributed machine learning. However, its performance suffers under arbitrarily heterogeneous computation times across workers, leading to suboptimal time complexity and inefficiency as the number of workers scales. While several Asynchronous SGD variants have been proposed, recent findings by Tyurin & Richt\'arik (NeurIPS 2023) reveal that none achieve optimal time complexity, leaving a significant gap in the literature. In this paper, we propose Ringmaster ASGD, a novel Asynchronous SGD method designed to address these limitations and tame the inherent challenges of Asynchronous SGD. We establish, through rigorous theoretical analysis, that Ringmaster ASGD achieves optimal time complexity under arbitrarily heterogeneous and dynamically fluctuating worker computation times. This makes it the first Asynchronous SGD method to meet the theoretical lower bounds for time complexity in such scenarios.
Authors: Jianyu Xu, Xuan Wang, Yu-Xiang Wang, Jiashuo Jiang
Abstract: We study an online learning problem on dynamic pricing and resource allocation, where we make joint pricing and inventory decisions to maximize the overall net profit. We consider the stochastic dependence of demands on the price, which complicates the resource allocation process and introduces significant non-convexity and non-smoothness to the problem. To solve this problem, we develop an efficient algorithm that utilizes a "Lower-Confidence Bound (LCB)" meta-strategy over multiple OCO agents. Our algorithm achieves $\tilde{O}(\sqrt{Tmn})$ regret (for $m$ suppliers and $n$ consumers), which is optimal with respect to the time horizon $T$. Our results illustrate an effective integration of statistical learning methodologies with complex operations research problems.
Authors: Bartosz Cywi\'nski, Kamil Deja
Abstract: Diffusion models, while powerful, can inadvertently generate harmful or undesirable content, raising significant ethical and safety concerns. Recent machine unlearning approaches offer potential solutions but often lack transparency, making it difficult to understand the changes they introduce to the base model. In this work, we introduce SAeUron, a novel method leveraging features learned by sparse autoencoders (SAEs) to remove unwanted concepts in text-to-image diffusion models. First, we demonstrate that SAEs, trained in an unsupervised manner on activations from multiple denoising timesteps of the diffusion model, capture sparse and interpretable features corresponding to specific concepts. Building on this, we propose a feature selection method that enables precise interventions on model activations to block targeted content while preserving overall performance. Our evaluation shows that SAeUron outperforms existing approaches on the UnlearnCanvas benchmark for concepts and style unlearning, and effectively eliminates nudity when evaluated with I2P. Moreover, we show that with a single SAE, we can remove multiple concepts simultaneously and that in contrast to other methods, SAeUron mitigates the possibility of generating unwanted content under adversarial attack. Code and checkpoints are available at https://github.com/cywinski/SAeUron.
Authors: Artavazd Maranjyan, El Mehdi Saad, Peter Richt\'arik, Francesco Orabona
Abstract: Asynchronous methods are fundamental for parallelizing computations in distributed machine learning. They aim to accelerate training by fully utilizing all available resources. However, their greedy approach can lead to inefficiencies using more computation than required, especially when computation times vary across devices. If the computation times were known in advance, training could be fast and resource-efficient by assigning more tasks to faster workers. The challenge lies in achieving this optimal allocation without prior knowledge of the computation time distributions. In this paper, we propose ATA (Adaptive Task Allocation), a method that adapts to heterogeneous and random distributions of worker computation times. Through rigorous theoretical analysis, we show that ATA identifies the optimal task allocation and performs comparably to methods with prior knowledge of computation times. Experimental results further demonstrate that ATA is resource-efficient, significantly reducing costs compared to the greedy approach, which can be arbitrarily expensive depending on the number of workers.
Authors: Erpai Luo, Xinran Wei, Lin Huang, Yunyang Li, Han Yang, Zaishuo Xia, Zun Wang, Chang Liu, Bin Shao, Jia Zhang
Abstract: Hamiltonian matrix prediction is pivotal in computational chemistry, serving as the foundation for determining a wide range of molecular properties. While SE(3) equivariant graph neural networks have achieved remarkable success in this domain, their substantial computational cost--driven by high-order tensor product (TP) operations--restricts their scalability to large molecular systems with extensive basis sets. To address this challenge, we introduce SPHNet, an efficient and scalable equivariant network, that incorporates adaptive SParsity into Hamiltonian prediction. SPHNet employs two innovative sparse gates to selectively constrain non-critical interaction combinations, significantly reducing tensor product computations while maintaining accuracy. To optimize the sparse representation, we develop a Three-phase Sparsity Scheduler, ensuring stable convergence and achieving high performance at sparsity rates of up to 70%. Extensive evaluations on QH9 and PubchemQH datasets demonstrate that SPHNet achieves state-of-the-art accuracy while providing up to a 7x speedup over existing models. Beyond Hamiltonian prediction, the proposed sparsification techniques also hold significant potential for improving the efficiency and scalability of other SE(3) equivariant networks, further broadening their applicability and impact. Our code can be found at https://github.com/microsoft/SPHNet.
Authors: Zeyu Wang, Yao-Hui Li, Xin Li, Hongyu Zang, Romain Laroche, Riashat Islam
Abstract: Multi-View Reinforcement Learning (MVRL) seeks to provide agents with multi-view observations, enabling them to perceive environment with greater effectiveness and precision. Recent advancements in MVRL focus on extracting latent representations from multiview observations and leveraging them in control tasks. However, it is not straightforward to learn compact and task-relevant representations, particularly in the presence of redundancy, distracting information, or missing views. In this paper, we propose Multi-view Fusion State for Control (MFSC), firstly incorporating bisimulation metric learning into MVRL to learn task-relevant representations. Furthermore, we propose a multiview-based mask and latent reconstruction auxiliary task that exploits shared information across views and improves MFSC's robustness in missing views by introducing a mask token. Extensive experimental results demonstrate that our method outperforms existing approaches in MVRL tasks. Even in more realistic scenarios with interference or missing views, MFSC consistently maintains high performance.
Authors: Yijun Dong, Yicheng Li, Yunai Li, Jason D. Lee, Qi Lei
Abstract: Weak-to-strong (W2S) generalization is a type of finetuning (FT) where a strong (large) student model is trained on pseudo-labels generated by a weak teacher. Surprisingly, W2S FT often outperforms the weak teacher. We seek to understand this phenomenon through the observation that FT often occurs in intrinsically low-dimensional spaces. Leveraging the low intrinsic dimensionality of FT, we analyze W2S in the ridgeless regression setting from a variance reduction perspective. For a strong student-weak teacher pair with sufficiently expressive low-dimensional feature subspaces $\mathcal{V}_s, \mathcal{V}_w$, we provide an exact characterization of the variance that dominates the generalization error of W2S. This unveils a virtue of discrepancy between the strong and weak models in W2S: the variance of the weak teacher is inherited by the strong student in $\mathcal{V}_s \cap \mathcal{V}_w$, while reduced by a factor of $\dim(\mathcal{V}_s)/N$ in the subspace of discrepancy $\mathcal{V}_w \setminus \mathcal{V}_s$ with $N$ pseudo-labels for W2S. Our analysis further casts light on the sample complexities and the scaling of performance gap recovery in W2S. The analysis is supported by experiments on synthetic regression problems, as well as real vision and NLP tasks.
Authors: Brandon Trabucco, Gunnar Sigurdsson, Robinson Piramuthu, Ruslan Salakhutdinov
Abstract: The predominant approach for training web navigation agents is to gather human demonstrations for a set of popular websites and hand-written tasks, but it is becoming clear that human data is an inefficient resource. We develop a pipeline to facilitate internet-scale training for agents without laborious human annotations. In the first stage, an LLM annotates 150k sites with agentic tasks. In the next stage, LLM agents complete tasks and produce trajectories. In the final stage, an LLM filters trajectories by judging their success. Language models are powerful data curation tools, identifying harmful content with an accuracy of 97%, judging successful trajectories with an accuracy of 82.6%, and producing effective data. We train agents based on Qwen 3 1.7B that are competitive with frontier LLMs as web agents, while being smaller and faster. Our top agent reaches a success rate of 56.9%, outperforming the data collection policy Qwen 3 235B, a 235 times larger Llama 4 Maverick, and reaching 94.7% of the performance of Gemini 2.5 Flash. We are releasing code, models and data at: https://data-for-agents.github.io.
Authors: Zhicong Wang, Zicheng Ma, Ziqiang Cao, Changlong Zhou, Jun Zhang, Yiqin Gao
Abstract: Motivation: Proteins are of great significance in living organisms. However, understanding their functions encounters numerous challenges, such as insufficient integration of multimodal information, a large number of training parameters, limited flexibility of classification-based methods, and the lack of systematic evaluation metrics for protein Q&A systems. To tackle these issues, we propose the Prot2Chat framework. Results: We modified ProteinMPNN to encode protein sequence and structural information in a unified way. We used a large language model (LLM) to encode questions into vectors and developed a protein-text adapter to compress protein information into virtual tokens based on these vectors, achieving the early fusion of text and protein information. Finally, the same LLM reads the virtual tokens and the questions to generate answers. To optimize training efficiency, we froze the encoder and employed Low-Rank Adaptation (LoRA) techniques for the LLM. Experiments on two datasets show that both automated metrics and expert evaluations demonstrate the superior performance of our model, and zero-shot prediction results highlight its generalization ability. The models and codes are available at https://github.com/ wangzc1233/Prot2Chat. Contact: zqcao@suda.edu.cn or wangzc025@163.com Key words: Protein Q&A, Early-Fusion, LLM
URLs: https://github.com/
Authors: Daniel Barzilai, Guy Kornowski, Ohad Shamir
Abstract: In recent years, there has been much interest in understanding the generalization behavior of interpolating predictors, which overfit on noisy training data. Whereas standard analyses are concerned with whether a method is consistent or not, recent observations have shown that even inconsistent predictors can generalize well. In this work, we revisit the classic interpolating Nadaraya-Watson (NW) estimator (also known as Shepard's method), and study its generalization capabilities through this modern viewpoint. In particular, by varying a single bandwidth-like hyperparameter, we prove the existence of multiple overfitting behaviors, ranging non-monotonically from catastrophic, through benign, to tempered. Our results highlight how even classical interpolating methods can exhibit intricate generalization behaviors. In addition, for the purpose of tuning the hyperparameter, the results suggest that over-estimating the intrinsic dimension of the data is less harmful than under-estimating it. Numerical experiments complement our theory, demonstrating the same phenomena.
Authors: Daniel Palenicek, Florian Vogt, Joe Watson, Jan Peters
Abstract: Reinforcement learning has achieved significant milestones, but sample efficiency remains a bottleneck for real-world applications. Recently, CrossQ has demonstrated state-of-the-art sample efficiency with a low update-to-data (UTD) ratio of 1. In this work, we explore CrossQ's scaling behavior with higher UTD ratios. We identify challenges in the training dynamics, which are emphasized by higher UTD ratios. To address these, we integrate weight normalization into the CrossQ framework, a solution that stabilizes training, has been shown to prevent potential loss of plasticity and keeps the effective learning rate constant. Our proposed approach reliably scales with increasing UTD ratios, achieving competitive performance across 25 challenging continuous control tasks on the DeepMind Control Suite and Myosuite benchmarks, notably the complex dog and humanoid environments. This work eliminates the need for drastic interventions, such as network resets, and offers a simple yet robust pathway for improving sample efficiency and scalability in model-free reinforcement learning.
Authors: Krunoslav Lehman Pavasovic, Jakob Verbeek, Giulio Biroli, Marc Mezard
Abstract: Classifier-Free Guidance (CFG) is a widely adopted technique in diffusion and flow-based generative models, enabling high-quality conditional generation. A key theoretical challenge is characterizing the distribution induced by CFG, particularly in high-dimensional settings relevant to real-world data. Previous works have shown that CFG modifies the target distribution, steering it towards a distribution sharper than the target one, more shifted towards the boundary of the class. In this work, we provide a high-dimensional analysis of CFG, showing that these distortions vanish as the data dimension grows. We present a blessing-of-dimensionality result demonstrating that in sufficiently high and infinite dimensions, CFG accurately reproduces the target distribution. Using our high-dimensional theory, we show that there is a large family of guidances enjoying this property, in particular non-linear CFG generalizations. We study a simple non-linear power-law version, for which we demonstrate improved robustness, sample fidelity and diversity. Our findings are validated with experiments on class-conditional and text-to-image generation using state-of-the-art diffusion and flow-matching models.
Authors: Abdelkrim Alahyane (LAAS-SARA, LAAS), C\'eline Comte (CNRS, LAAS-SARA, LAAS), Matthieu Jonckheere (CNRS, LAAS-SARA, LAAS), \'Eric Moulines (X)
Abstract: Synchronous federated learning (FL) scales poorly with the number of clients due to the straggler effect. Algorithms like FedAsync and GeneralizedFedAsync address this limitation by enabling asynchronous communication between clients and the central server. In this work, we rely on stochastic modeling and analysis to better understand the impact of design choices in asynchronous FL algorithms, such as the concurrency level and routing probabilities, and we leverage this knowledge to optimize loss. Compared to most existing studies, we account for the joint impact of heterogeneous and variable service speeds and heterogeneous datasets at the clients. We characterize in particular a fundamental trade-off for optimizing asynchronous FL: minimizing gradient estimation errors by avoiding model parameter staleness, while also speeding up the system by increasing the throughput of model updates. Our two main contributions can be summarized as follows. First, we prove a discrete variant of Little's law to derive a closed-form expression for relative delay, a metric that quantifies staleness. This allows us to efficiently minimize the average loss per model update, which has been the gold standard in literature to date. Second, we observe that naively optimizing this metric leads us to slow down the system drastically by overemphazing staleness at the detriment of throughput. This motivates us to introduce an alternative metric that also takes system speed into account, for which we derive a tractable upper-bound that can be minimized numerically. Extensive numerical results show that these optimizations enhance accuracy by 10% to 30%.
Authors: Jiaqi Wang, Yuhang Zhou, Zhixiong Zhang, Qiguang Chen, Yongqiang Chen, James Cheng
Abstract: Out-of-distribution generalization is a common problem that expects the model to perform well in the different distributions even far from the train data. A popular approach to addressing this issue is invariant learning (IL), in which the model is compiled to focus on invariant features instead of spurious features by adding strong constraints during training. However, there are some potential pitfalls of strong invariant constraints. Due to the limited number of diverse environments and over-regularization in the feature space, it may lead to a loss of important details in the invariant features while alleviating the spurious correlations, namely the over-invariance, which can also degrade the generalization performance. We theoretically define the over-invariance and observe that this issue occurs in various classic IL methods. To alleviate this issue, we propose a simple approach Diverse Invariant Learning (DivIL) by adding the unsupervised contrastive learning and the random masking mechanism compensatory for the invariant constraints, which can be applied to various IL methods. Furthermore, we conduct experiments across multiple modalities across 12 datasets and 6 classic models, verifying our over-invariance insight and the effectiveness of our DivIL framework. Our code is available at https://github.com/kokolerk/DivIL.
Authors: Gallil Maimon, Avishai Elmakies, Yossi Adi
Abstract: We introduce Slam, a recipe for training high-quality Speech Language Models (SLMs) on a single academic GPU in 24 hours. We do so through empirical analysis of model initialisation and architecture, synthetic training data, preference optimisation with synthetic data and tweaking all other components. We empirically demonstrate that this training recipe also scales well with more compute getting results on par with leading SLMs in a fraction of the compute cost. We hope these insights will make SLM training and research more accessible. In the context of SLM scaling laws, our results far outperform predicted compute optimal performance, giving an optimistic view to SLM feasibility. See code, data, models, samples at - https://pages.cs.huji.ac.il/adiyoss-lab/slamming .
Authors: Alireza Kazemipour, Simone Parisi, Matthew E. Taylor, Michael Bowling
Abstract: A tenet of reinforcement learning is that the agent always observes rewards. However, this is not true in many realistic settings, e.g., a human observer may not always be available to provide rewards, sensors may be limited or malfunctioning, or rewards may be inaccessible during deployment. Monitored Markov decision processes (Mon-MDPs) have recently been proposed to model such settings. However, existing Mon-MDP algorithms have several limitations: they do not fully exploit the problem structure, cannot leverage a known monitor, lack worst-case guarantees for 'unsolvable' Mon-MDPs without specific initialization, and offer only asymptotic convergence proofs. This paper makes three contributions. First, we introduce a model-based algorithm for Mon-MDPs that addresses these shortcomings. The algorithm employs two instances of model-based interval estimation: one to ensure that observable rewards are reliably captured, and another to learn the minimax-optimal policy. Second, we empirically demonstrate the advantages. We show faster convergence than prior algorithms in over four dozen benchmarks, and even more dramatic improvement when the monitoring process is known. Third, we present the first finite-sample bound on performance. We show convergence to a minimax-optimal policy even when some rewards are never observable.
Authors: Allen Schmaltz
Abstract: We address the neural network robustness problem by adding Similarity (i.e., correctly predicted depth-matches into training)-awareness and Distance-to-training-distribution-awareness to the existing output Magnitude (i.e., decision-boundary)-awareness of the softmax function. The resulting SDM activation function provides strong signals of the relative epistemic (reducible) predictive uncertainty. We use this novel behavior to further address the complementary HCI problem of mapping the output to human-interpretable summary statistics over relevant partitions of a held-out calibration set. Estimates of prediction-conditional uncertainty are obtained via a parsimonious learned transform over the class-conditional empirical CDFs of the output of a final-layer SDM activation function. For decision-making and as an intrinsic model check, estimates of class-conditional accuracy are obtained by further partitioning the high-probability regions of this calibrated output into class-conditional, region-specific CDFs. The uncertainty estimates from SDM calibration are remarkably robust to test-time distribution shifts and out-of-distribution inputs; incorporate awareness of the effective sample size; provide estimates of uncertainty from the learning and data splitting processes; and are well-suited for selective classification and conditional branching for additional test-time compute based on the predictive uncertainty, as for selective LLM generation, routing, and composition over multiple models and retrieval. Finally, we construct SDM networks, LLMs with uncertainty-aware verification and interpretability-by-exemplar as intrinsic properties. We provide open-source software implementing these results.
Authors: Guanghan Wang, Yair Schiff, Subham Sekhar Sahoo, Volodymyr Kuleshov
Abstract: Part of the success of diffusion models stems from their ability to perform iterative refinement, i.e., repeatedly correcting outputs during generation. However, modern masked discrete diffusion lacks this capability: when a token is generated, it cannot be updated again, even when it introduces an error. Here, we address this limitation by introducing the remasking diffusion model (ReMDM) sampler, a method that can be applied to pretrained masked diffusion models in a principled way and that is derived from a discrete diffusion model with a custom remasking backward process. Most interestingly, ReMDM endows discrete diffusion with a form of inference-time compute scaling. By increasing the number of sampling steps, ReMDM generates natural language outputs that approach the quality of autoregressive models, whereas when the computation budget is limited, ReMDM better maintains quality. ReMDM also improves sample quality of masked diffusion models for discretized images, and in scientific domains such as molecule design, ReMDM facilitates diffusion guidance and pushes the Pareto frontier of controllability relative to classical masking and uniform noise diffusion. We provide the code along with a blog post on the project page: https://remdm.github.io
URLs: https://remdm.github.io
Authors: Wenda Chu, Zihui Wu, Yifan Chen, Yang Song, Yisong Yue
Abstract: We study the problem of posterior sampling in discrete-state spaces using discrete diffusion models. While posterior sampling methods for continuous diffusion models have achieved remarkable progress, analogous methods for discrete diffusion models remain challenging. In this work, we introduce a principled plug-and-play discrete diffusion posterior sampling algorithm based on split Gibbs sampling, which we call SGDD. Our algorithm enables reward-guided generation and solving inverse problems in discrete-state spaces. We demonstrate the convergence of SGDD to the target posterior distribution and verify this through controlled experiments on synthetic benchmarks. Our method enjoys state-of-the-art posterior sampling performance on a range of benchmarks for discrete data, including DNA sequence design, discrete image inverse problems, and music infilling, achieving more than 30% improved performance compared to existing baselines.
Authors: Zekang Wang, Zhe He, Borong Zhang, Edan Toledo, Steven Morad
Abstract: We present the POPGym Arcade, a collection of hardware-accelerated, pixel-based environments with shared observation and action spaces. Each environment includes fully and partially observable variants, enabling counterfactual studies on partial observability. We also introduce mathematical tools for analyzing policies under partial observability, which reveal how agents recall past information to make decisions. Our analysis shows (1) that controlling for partial observability is critical and (2) that agents with long-term memory learn brittle policies that struggle to generalize. Finally, we demonstrate that recurrent policies can be "poisoned" by old, out-of-distribution observations, with implications for sim-to-real transfer, imitation learning, and offline reinforcement learning.
Authors: Seongheon Park, Xuefeng Du, Min-Hsuan Yeh, Haobo Wang, Yixuan Li
Abstract: Hallucinations in LLMs pose a significant concern to their safe deployment in real-world applications. Recent approaches have leveraged the latent space of LLMs for hallucination detection, but their embeddings, optimized for linguistic coherence rather than factual accuracy, often fail to clearly separate truthful and hallucinated content. To this end, we propose the Truthfulness Separator Vector (TSV), a lightweight and flexible steering vector that reshapes the LLM's representation space during inference to enhance the separation between truthful and hallucinated outputs, without altering model parameters. Our two-stage framework first trains TSV on a small set of labeled exemplars to form compact and well-separated clusters. It then augments the exemplar set with unlabeled LLM generations, employing an optimal transport-based algorithm for pseudo-labeling combined with a confidence-based filtering process. Extensive experiments demonstrate that TSV achieves state-of-the-art performance with minimal labeled data, exhibiting strong generalization across datasets and providing a practical solution for real-world LLM applications.
Authors: Afnan Sultan, Max Rausch-Dupont, Shahrukh Khan, Olga Kalinina, Dietrich Klakow, Andrea Volkamer
Abstract: Over the past six years, molecular transformer models have become key tools in drug discovery. Most existing models are pre-trained on large, unlabeled datasets such as ZINC or ChEMBL. However, the extent to which large-scale pre-training improves molecular property prediction remains unclear. This study evaluates transformer models for this task while addressing their limitations. We explore how pre-training dataset size and chemically informed objectives impact performance. Our results show that increasing the dataset beyond approximately 400K to 800K molecules from large-scale unlabeled databases does not enhance performance across seven datasets covering five ADME endpoints: lipophilicity, permeability, solubility (two datasets), microsomal stability (two datasets), and plasma protein binding. In contrast, domain adaptation on a small, domain-specific dataset (less than or equal 4K molecules) using multi-task regression of physicochemical properties significantly boosts performance (P-value less than 0.001). A model pre-trained on 400K molecules and adapted with domain-specific data outperforms larger models such as MolFormer and performs comparably to MolBERT. Benchmarks against Random Forest (RF) baselines using descriptors and Morgan fingerprints show that chemically and physically informed features consistently yield better performance across model types. While RF remains a strong baseline, we identify concrete practices to enhance transformer performance. Aligning pre-training and adaptation with chemically meaningful tasks and domain-relevant data presents a promising direction for molecular property prediction. Our models are available on HuggingFace for easy use and adaptation.
Authors: Chaitanya K. Joshi, Xiang Fu, Yi-Lun Liao, Vahe Gharakhanyan, Benjamin Kurt Miller, Anuroop Sriram, Zachary W. Ulissi
Abstract: Diffusion models are the standard toolkit for generative modelling of 3D atomic systems. However, for different types of atomic systems -- such as molecules and materials -- the generative processes are usually highly specific to the target system despite the underlying physics being the same. We introduce the All-atom Diffusion Transformer (ADiT), a unified latent diffusion framework for jointly generating both periodic materials and non-periodic molecular systems using the same model: (1) An autoencoder maps a unified, all-atom representations of molecules and materials to a shared latent embedding space; and (2) A diffusion model is trained to generate new latent embeddings that the autoencoder can decode to sample new molecules or materials. Experiments on MP20, QM9 and GEOM-DRUGS datasets demonstrate that jointly trained ADiT generates realistic and valid molecules as well as materials, obtaining state-of-the-art results on par with molecule and crystal-specific models. ADiT uses standard Transformers with minimal inductive biases for both the autoencoder and diffusion model, resulting in significant speedups during training and inference compared to equivariant diffusion models. Scaling ADiT up to half a billion parameters predictably improves performance, representing a step towards broadly generalizable foundation models for generative chemistry. Open source code: https://github.com/facebookresearch/all-atom-diffusion-transformer
URLs: https://github.com/facebookresearch/all-atom-diffusion-transformer
Authors: Shwai He, Weilin Cai, Jiayi Huang, Ang Li
Abstract: The Mixture of Experts (MoE) is an effective architecture for scaling large language models by leveraging sparse expert activation, optimizing the trade-off between performance and efficiency. However, under expert parallelism, MoE suffers from inference inefficiencies due to imbalanced token-to-expert assignment, where some experts are overloaded while others remain underutilized. This imbalance leads to poor resource utilization and increased latency, as the most burdened expert dictates the overall delay, a phenomenon we define as the \textbf{\textit{Straggler Effect}}. To mitigate this, we propose Capacity-Aware Inference, including two key techniques: (1) \textbf{\textit{Capacity-Aware Token Drop}}, which discards overloaded tokens to regulate the maximum latency of MoE, and (2) \textbf{\textit{Capacity-Aware Token Reroute}}, which reallocates overflowed tokens to underutilized experts, balancing the token distribution. These techniques collectively optimize both high-load and low-load expert utilization, leading to a more efficient MoE inference pipeline. Extensive experiments demonstrate the effectiveness of our methods, showing significant improvements in inference efficiency, e.g., 0.2\% average performance increase and a 1.94$\times$ inference speedup on Mixtral-8$\times$7B-Instruct.
Authors: Petr Ry\v{s}av\'y, Pavel Ryt\'i\v{r}, Xiaoyu He, Georgios Korpas, Jakub Mare\v{c}ek
Abstract: In mixed graphs, there are both directed and undirected edges. An extension of acyclicity to this mixed-graph setting is known as maximally ancestral graphs. This extension is of considerable interest in causal learning in the presence of confounders. There, directed edges represent a clear direction of causality, while undirected edges represent confounding. We propose a score-based branch-and-cut algorithm for learning maximally ancestral graphs. The algorithm produces more accurate results than state-of-the-art methods, while being faster to run on small and medium-sized synthetic instances.
Authors: Jiaming Ji, Xinyu Chen, Rui Pan, Conghui Zhang, Han Zhu, Jiahao Li, Donghai Hong, Boyuan Chen, Jiayi Zhou, Kaile Wang, Juntao Dai, Chi-Min Chan, Yida Tang, Sirui Han, Yike Guo, Yaodong Yang
Abstract: Multimodal large language models (MLLMs) are essential for building general-purpose AI assistants; however, they pose increasing safety risks. How can we ensure safety alignment of MLLMs to prevent undesired behaviors? Going further, it is critical to explore how to fine-tune MLLMs to preserve capabilities while meeting safety constraints. Fundamentally, this challenge can be formulated as a min-max optimization problem. However, existing datasets have not yet disentangled single preference signals into explicit safety constraints, hindering systematic investigation in this direction. Moreover, it remains an open question whether such constraints can be effectively incorporated into the optimization process for multi-modal models. In this work, we present the first exploration of the Safe RLHF-V -- the first multimodal safety alignment framework. The framework consists of: $\mathbf{(I)}$ BeaverTails-V, the first open-source dataset featuring dual preference annotations for helpfulness and safety, supplemented with multi-level safety labels (minor, moderate, severe); $\mathbf{(II)}$ Beaver-Guard-V, a multi-level guardrail system to proactively defend against unsafe queries and adversarial attacks. Applying the guard model over five rounds of filtering and regeneration significantly enhances the precursor model's overall safety by an average of 40.9%. $\mathbf{(III)}$ Based on dual preference, we initiate the first exploration of multi-modal safety alignment within a constrained optimization. Experimental results demonstrate that Safe RLHF effectively improves both model helpfulness and safety. Specifically, Safe RLHF-V enhances model safety by 34.2% and helpfulness by 34.3%.
Authors: Mahade Hasan, Farhana Yasmin, Xue Yu
Abstract: Heart disease remains a leading cause of mortality and morbidity worldwide, necessitating the development of accurate and reliable predictive models to facilitate early detection and intervention. While state of the art work has focused on various machine learning approaches for predicting heart disease, but they could not able to achieve remarkable accuracy. In response to this need, we applied nine machine learning algorithms XGBoost, logistic regression, decision tree, random forest, k-nearest neighbors (KNN), support vector machine (SVM), gaussian na\"ive bayes (NB gaussian), adaptive boosting, and linear regression to predict heart disease based on a range of physiological indicators. Our approach involved feature selection techniques to identify the most relevant predictors, aimed at refining the models to enhance both performance and interpretability. The models were trained, incorporating processes such as grid search hyperparameter tuning, and cross-validation to minimize overfitting. Additionally, we have developed a novel voting system with feature selection techniques to advance heart disease classification. Furthermore, we have evaluated the models using key performance metrics including accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic curve (ROC AUC). Among the models, XGBoost demonstrated exceptional performance, achieving 99% accuracy, precision, F1-Score, 98% recall, and 100% ROC AUC. This study offers a promising approach to early heart disease diagnosis and preventive healthcare.
Authors: Kwangjun Ahn, Byron Xu, Natalie Abreu, John Langford
Abstract: Recent work has shown that orthonormal matrix updates speed up neural network optimization, improve training stability, and offer better hyperparameter transfer across model sizes. Applying these updates efficiently when model weights and optimizer states are sharded across a large-scale distributed LLM training system remains a major challenge. We introduce Dion (DIstributed OrthoNormalization), a scalable and communication-efficient orthonormalizing optimizer. Dion leverages low-rank approximation and decoupled momentum buffers, eliminating the need for full gradient synchronization while producing numerically equivalent results. It is compatible with simultaneous DDP, FSDP, and TP parallelism, and it computes an orthonormalized update without unsharding a full parameter matrix on any single device. We evaluate Dion on language models from 120M to 3B parameters and find that its benefits improve with increasing model size and batch size.
Authors: Yuxuan Li, Yicheng Gao, Ning Yang, Stephen Xia
Abstract: Episodic tasks in Reinforcement Learning (RL) often pose challenges due to sparse reward signals and high-dimensional state spaces, which hinder efficient learning. Additionally, these tasks often feature hidden "trap states" -- irreversible failures that prevent task completion but do not provide explicit negative rewards to guide agents away from repeated errors. To address these issues, we propose Time-Weighted Contrastive Reward Learning (TW-CRL), an Inverse Reinforcement Learning (IRL) framework that leverages both successful and failed demonstrations. By incorporating temporal information, TW-CRL learns a dense reward function that identifies critical states associated with success or failure. This approach not only enables agents to avoid trap states but also encourages meaningful exploration beyond simple imitation of expert trajectories. Empirical evaluations on navigation tasks and robotic manipulation benchmarks demonstrate that TW-CRL surpasses state-of-the-art methods, achieving improved efficiency and robustness.
Authors: Thomas Chen, Chun-Kai Kevin Chien, Patricia Mu\~noz Ewald, Andrew G. Moore
Abstract: We prove that overparametrized neural networks are able to generalize with a test error that is independent of the level of overparametrization, and independent of the Vapnik-Chervonenkis (VC) dimension. We prove explicit bounds that only depend on the metric geometry of the test and training sets, on the regularity properties of the activation function, and on the operator norms of the weights and norms of biases. For overparametrized deep ReLU networks with a training sample size bounded by the input space dimension, we explicitly construct zero loss minimizers without use of gradient descent, and prove that the generalization error is independent of the network architecture.
Authors: Marcus R\"ub, Daniel Konegen, Patrick Selle, Axel Sikora, Daniel Mueller-Gritschneder
Abstract: Selecting data points for model training is critical in machine learning. Effective selection methods can reduce the labeling effort, optimize on-device training for embedded systems with limited data storage, and enhance the model performance. This paper introduces a novel algorithm that uses Grad-CAM to make online decisions about retaining or discarding data points. Optimized for embedded devices, the algorithm computes a unique DRIP Score to quantify the importance of each data point. This enables dynamic decision-making on whether a data point should be stored for potential retraining or discarded without compromising model performance. Experimental evaluations on four benchmark datasets demonstrate that our approach can match or even surpass the accuracy of models trained on the entire dataset, all while achieving storage savings of up to 39\%. To our knowledge, this is the first algorithm that makes online decisions about data point retention without requiring access to the entire dataset.
Authors: Michal Balcerak, Tamaz Amiranashvili, Antonio Terpin, Suprosanna Shit, Lea Bogensperger, Sebastian Kaltenbach, Petros Koumoutsakos, Bjoern Menze
Abstract: The most widely used generative models map noise and data distributions by matching flows or scores. However, they struggle to incorporate partial observations and additional priors--something energy-based models (EBMs) handle elegantly by simply adding corresponding scalar energy terms. We address this issue by proposing Energy Matching, a framework that endows flow-based approaches with the flexibility of EBMs. Far from the data manifold, samples move along curl-free, optimal transport paths from noise to data. As they approach the data manifold, an entropic energy term guides the system into a Boltzmann equilibrium distribution, explicitly capturing the underlying likelihood structure of the data. We parameterize this dynamic with a single time-independent scalar field, which serves as both a powerful generator and a flexible prior for effective regularization of inverse problems. Our method substantially outperforms existing EBMs on CIFAR-10 and ImageNet generation in terms of fidelity, while retaining simulation-free training of transport-based approaches away from the data manifold. Furthermore, we leverage the method's flexibility to introduce an interaction energy that supports diverse mode exploration, which we demonstrate in a controlled protein-generation setting. Our approach focuses on learning a scalar potential energy--without time-conditioning, auxiliary generators, or additional networks--which marks a significant departure from recent EBM methods. We believe that this simplified framework significantly advances EBMs capabilities and paves the way for their wider adoption in generative modeling across diverse domains.
Authors: Assaf Gerner, Netta Madvil, Nadav Barak, Alex Zaikman, Jonatan Liberman, Liron Hamra, Rotem Brazilay, Shay Tsadok, Yaron Friedman, Neal Harow, Noam Bressler, 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" - a member of Deepchecks' ORION (Output Reasoning-based InspectiON) family of lightweight evaluation models. It is our framework for hallucination detection, 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: Xiaolu Chen, Chenghao Huang, Yanru Zhang, Hao Wang
Abstract: The rapid expansion of distributed photovoltaic (PV) installations worldwide, many being behind-the-meter systems, has significantly challenged energy management and grid operations, as unobservable PV generation further complicates the supply-demand balance. Therefore, estimating this generation from net load, known as PV disaggregation, is critical. Given privacy concerns and the need for large training datasets, federated learning becomes a promising approach, but statistical heterogeneity, arising from geographical and behavioral variations among prosumers, poses new challenges to PV disaggregation. To overcome these challenges, a privacy-preserving distributed PV disaggregation framework is proposed using Personalized Federated Learning (PFL). The proposed method employs a two-level framework that combines local and global modeling. At the local level, a transformer-based PV disaggregation model is designed to generate solar irradiance embeddings for representing local PV conditions. A novel adaptive local aggregation mechanism is adopted to mitigate the impact of statistical heterogeneity on the local model, extracting a portion of global information that benefits the local model. At the global level, a central server aggregates information uploaded from multiple data centers, preserving privacy while enabling cross-center knowledge sharing. Experiments on real-world data demonstrate the effectiveness of this proposed framework, showing improved accuracy and robustness compared to benchmark methods.
Authors: Pouria Fatemi, Ehsan Sharifian, Mohammad Hossein Yassaee
Abstract: Counterfactual explanations enhance interpretability by identifying alternative inputs that produce different outputs, offering localized insights into model decisions. However, traditional methods often neglect causal relationships, leading to unrealistic examples. While newer approaches integrate causality, they are computationally expensive. To address these challenges, we propose an efficient method called BRACE based on backtracking counterfactuals that incorporates causal reasoning to generate actionable explanations. We first examine the limitations of existing methods and then introduce our novel approach and its features. We also explore the relationship between our method and previous techniques, demonstrating that it generalizes them in specific scenarios. Finally, experiments show that our method provides deeper insights into model outputs.
Authors: Bojian Yin, Federico Corradi
Abstract: Backpropagation (BP) is the cornerstone of deep learning, but its reliance on global gradient synchronization limits scalability and imposes significant memory overhead. We propose Stochastic Variational Propagation (SVP), a scalable alternative that reframes training as hierarchical variational inference. SVP treats layer activations as latent variables and optimizes local Evidence Lower Bounds (ELBOs), enabling independent, local updates while preserving global coherence. However, directly applying KL divergence in layer-wise ELBOs risks inter-layer's representation collapse due to excessive compression. To prevent this, SVP projects activations into low-dimensional spaces via fixed random matrices, ensuring information preservation and representational diversity. Combined with a feature alignment loss for inter-layer consistency, SVP achieves competitive accuracy with BP across diverse architectures (MLPs, CNNs, Transformers) and datasets (MNIST to ImageNet), reduces memory usage by up to 4x, and significantly improves scalability. More broadly, SVP introduces a probabilistic perspective to deep representation learning, opening pathways toward more modular and interpretable neural network design.
Authors: Lennart Justen
Abstract: This study systematically evaluates 27 frontier Large Language Models on eight biology benchmarks spanning molecular biology, genetics, cloning, virology, and biosecurity. Models from major AI developers released between November 2022 and April 2025 were assessed through ten independent runs per benchmark. The findings reveal dramatic improvements in biological capabilities. Top model performance increased more than 4-fold on the challenging text-only subset of the Virology Capabilities Test over the study period, with OpenAI's o3 now performing twice as well as expert virologists. Several models now match or exceed expert-level performance on other challenging benchmarks, including the biology subsets of GPQA and WMDP and LAB-Bench CloningScenarios. Contrary to expectations, chain-of-thought did not substantially improve performance over zero-shot evaluation, while extended reasoning features in o3-mini and Claude 3.7 Sonnet typically improved performance as predicted by inference scaling. Benchmarks such as PubMedQA and the MMLU and WMDP biology subsets exhibited performance plateaus well below 100%, suggesting benchmark saturation and errors in the underlying benchmark data. The analysis highlights the need for more sophisticated evaluation methodologies as AI systems continue to advance.
Authors: John Bencina, Erkut Aykutlug, Yue Chen, Zerui Zhang, Stephanie Sorenson, Shao Tang, Changshuai Wei
Abstract: Data Driven Attribution, which assigns conversion credits to marketing interactions based on causal patterns learned from data, is the foundation of modern marketing intelligence and vital to any marketing businesses and advertising platform. In this paper, we introduce a unified transformer-based attribution approach that can handle member-level data, aggregate-level data, and integration of external macro factors. We detail the large scale implementation of the approach at LinkedIn, showcasing significant impact. We also share learning and insights that are broadly applicable to the marketing and ad tech fields.
Authors: Nero Z. Li, Xuehao Zhai, Zhichao Shi, Boshen Shi, Xuhui Jiang
Abstract: While Graph Contrastive Learning (GCL) has attracted considerable attention in the field of graph self-supervised learning, its performance heavily relies on data augmentations that are expected to generate semantically consistent positive pairs. Existing strategies typically resort to random perturbations or local structure preservation, yet lack explicit control over global structural consistency between augmented views. To address this limitation, we propose Fractal Graph Contrastive Learning (FractalGCL), a theory-driven framework that leverages fractal self-similarity to enforce global topological coherence. FractalGCL introduces two key innovations: a renormalisation-based augmentation that generates structurally aligned positive views via box coverings; and a fractal-dimension-aware contrastive loss that aligns graph embeddings according to their fractal dimensions. While combining the two innovations markedly boosts graph-representation quality, it also adds non-trivial computational overhead. To mitigate the computational overhead of fractal dimension estimation, we derive a one-shot estimator by proving that the dimension discrepancy between original and renormalised graphs converges weakly to a centred Gaussian distribution. This theoretical insight enables a reduction in dimension computation cost by an order of magnitude, cutting overall training time by approximately 61%. The experiments show that FractalGCL not only delivers state-of-the-art results on standard benchmarks but also outperforms traditional baselines on traffic networks by an average margin of about remarkably 7%. Codes are available at (https://anonymous.4open.science/r/FractalGCL-0511).
Authors: Yuhang Liu, Yingxue Zhang, Xin Zhang, Ling Tian, 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: 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. The code is available here: https://github.com/c0510gy/AFLS-AED.
Authors: Dennis Frauen, Valentyn Melnychuk, Jonas Schweisthal, Mihaela van der Schaar, 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: Guoming Li, Jian Yang, Yifan Chen
Abstract: Filtering-based graph neural networks (GNNs) constitute a distinct class of GNNs that employ graph filters to handle graph-structured data, achieving notable success in various graph-related tasks. Conventional methods adopt a graph-wise filtering paradigm, imposing a uniform filter across all nodes, yet recent findings suggest that this rigid paradigm struggles with heterophilic graphs. To overcome this, recent works have introduced node-wise filtering, which assigns distinct filters to individual nodes, offering enhanced adaptability. However, a fundamental gap remains: a comprehensive framework unifying these two strategies is still absent, limiting theoretical insights into the filtering paradigms. Moreover, through the lens of Contextual Stochastic Block Model, we reveal that a synthesis of graph-wise and node-wise filtering provides a sufficient solution for classification on graphs exhibiting both homophily and heterophily, suggesting the risk of excessive parameterization and potential overfitting with node-wise filtering. To address the limitations, this paper introduces Coarsening-guided Partition-wise Filtering (CPF). CPF innovates by performing filtering on node partitions. The method begins with structure-aware partition-wise filtering, which filters node partitions obtained via graph coarsening algorithms, and then performs feature-aware partition-wise filtering, refining node embeddings via filtering on clusters produced by $k$-means clustering over features. In-depth analysis is conducted for each phase of CPF, showing its superiority over other paradigms. Finally, benchmark node classification experiments, along with a real-world graph anomaly detection application, validate CPF's efficacy and practical utility.
Authors: Mattes Mollenhauer, Nicole M\"ucke, Dimitri Meunier, Arthur Gretton
Abstract: This paper examines the performance of ridge regression in reproducing kernel Hilbert spaces in the presence of noise that exhibits a finite number of higher moments. We establish excess risk bounds consisting of subgaussian and polynomial terms based on the well known integral operator framework. The dominant subgaussian component allows to achieve convergence rates that have previously only been derived under subexponential noise - a prevalent assumption in related work from the last two decades. These rates are optimal under standard eigenvalue decay conditions, demonstrating the asymptotic robustness of regularized least squares against heavy-tailed noise. Our derivations are based on a Fuk-Nagaev inequality for Hilbert-space valued random variables.
Authors: Zhangchen Xu, Yuetai Li, Fengqing Jiang, Bhaskar Ramasubramanian, Luyao Niu, Bill Yuchen Lin, Radha Poovendran
Abstract: Reinforcement Learning (RL) has become a powerful tool for enhancing the reasoning abilities of large language models (LLMs) by optimizing their policies with reward signals. Yet, RL's success relies on the reliability of rewards, which are provided by verifiers. In this paper, we expose and analyze a widespread problem--false negatives--where verifiers wrongly reject correct model outputs. Our in-depth study of the Big-Math-RL-Verified dataset reveals that over 38% of model-generated responses suffer from false negatives, where the verifier fails to recognize correct answers. We show, both empirically and theoretically, that these false negatives severely impair RL training by depriving the model of informative gradient signals and slowing convergence. To mitigate this, we propose tinyV, a lightweight LLM-based verifier that augments existing rule-based methods, which dynamically identifies potential false negatives and recovers valid responses to produce more accurate reward estimates. Across multiple math-reasoning benchmarks, integrating TinyV boosts pass rates by up to 10% and accelerates convergence relative to the baseline. Our findings highlight the critical importance of addressing verifier false negatives and offer a practical approach to improve RL-based fine-tuning of LLMs. Our code is available at https://github.com/uw-nsl/TinyV.
Authors: Qingyu Song, Peiyu Liao, Wenqian Zhao, Yiwen Wang, Shoubo Hu, Hui-Ling Zhen, Ning Jiang, Mingxuan Yuan
Abstract: The increasing deployment of Large Language Models (LLMs) on edge devices, driven by model advancements and hardware improvements, offers significant privacy benefits. However, these on-device LLMs inherently face performance limitations due to reduced model capacity and necessary compression techniques. To address this, we introduce a systematic methodology -- encompassing model capability, development efficiency, and system resources -- for evaluating on-device LLMs. Our comprehensive evaluation, encompassing models from 0.5B to 14B parameters and seven post-training quantization (PTQ) methods on commodity laptops, yields several critical insights: 1) System-level metrics exhibit near-linear scaling with effective bits-per-weight (BPW). 2) A practical threshold exists around $\sim$3.5 effective BPW, larger models subjected to low-bit quantization consistently outperform smaller models utilizing higher bit-precision. 3) Quantization with low BPW incurs marginal accuracy loss but significant memory savings. 4) Determined by low-level implementation specifics power consumption on CPU, where computation-intensive operations spend more power than memory-intensive ones. These findings offer crucial insights and practical guidelines for the efficient deployment and optimized configuration of LLMs on resource-constrained edge devices. Our codebase is available at https://github.com/simmonssong/LLMOnDevice.
Authors: Haruki Settai, Naoya Takeishi, Takehisa Yairi
Abstract: For continuous systems modeled by dynamical equations such as ODEs and SDEs, Bellman's principle of optimality takes the form of the Hamilton-Jacobi-Bellman (HJB) equation, which provides the theoretical target of reinforcement learning (RL). Although recent advances in RL successfully leverage this formulation, the existing methods typically assume the underlying dynamics are known a priori because they need explicit access to the coefficient functions of dynamical equations to update the value function following the HJB equation. We address this inherent limitation of HJB-based RL; we propose a model-free approach still targeting the HJB equation and propose the corresponding temporal difference method. We demonstrate its potential advantages over transition kernel-based formulations, both qualitatively and empirically. The proposed formulation paves the way toward bridging stochastic optimal control and model-free reinforcement learning.
Authors: Kanan Kiguchi, Yunhao Tu, Katsuhiro Ajito, Fady Alnajjar, Kazuyuki Murase
Abstract: We propose a novel framework for integrating fragmented multi-modal data in Alzheimer's disease (AD) research using large language models (LLMs) and knowledge graphs. While traditional multimodal analysis requires matched patient IDs across datasets, our approach demonstrates population-level integration of MRI, gene expression, biomarkers, EEG, and clinical indicators from independent cohorts. Statistical analysis identified significant features in each modality, which were connected as nodes in a knowledge graph. LLMs then analyzed the graph to extract potential correlations and generate hypotheses in natural language. This approach revealed several novel relationships, including a potential pathway linking metabolic risk factors to tau protein abnormalities via neuroinflammation (r>0.6, p<0.001), and unexpected correlations between frontal EEG channels and specific gene expression profiles (r=0.42-0.58, p<0.01). Cross-validation with independent datasets confirmed the robustness of major findings, with consistent effect sizes across cohorts (variance <15%). The reproducibility of these findings was further supported by expert review (Cohen's k=0.82) and computational validation. Our framework enables cross modal integration at a conceptual level without requiring patient ID matching, offering new possibilities for understanding AD pathology through fragmented data reuse and generating testable hypotheses for future research.
Authors: Adam N. Elmachtoub, Henry Lam, Haofeng Zhang, Yunfan Zhao
Abstract: In data-driven stochastic optimization, model parameters of the underlying distribution need to be estimated from data in addition to the optimization task. Recent literature considers integrating the estimation and optimization processes by selecting model parameters that lead to the best empirical objective performance. This integrated approach, which we call integrated-estimation-optimization (IEO), can be readily shown to outperform simple estimate-then-optimize (ETO) when the model is misspecified. In this paper, we show that a reverse behavior appears when the model class is well-specified and there is sufficient data. Specifically, for a general class of nonlinear stochastic optimization problems, we show that simple ETO outperforms IEO asymptotically when the model class covers the ground truth, in the strong sense of stochastic dominance of the regret. Namely, the entire distribution of the regret, not only its mean or other moments, is always better for ETO compared to IEO. Our results also apply to constrained, contextual optimization problems where the decision depends on observed features. Whenever applicable, we also demonstrate how standard sample average approximation (SAA) performs the worst when the model class is well-specified in terms of regret, and best when it is misspecified. Finally, we provide experimental results to support our theoretical comparisons and illustrate when our insights hold in finite-sample regimes and under various degrees of misspecification.
Authors: Meng Wei, Charlie Budd, Luis C. Garcia-Peraza-Herrera, Reuben Dorent, Miaojing Shi, Tom Vercauteren
Abstract: Surgical instrument segmentation is recognised as a key enabler in providing advanced surgical assistance and improving computer-assisted interventions. In this work, we propose SegMatch, a semi-supervised learning method to reduce the need for expensive annotation for laparoscopic and robotic surgical images. SegMatch builds on FixMatch, a widespread semi supervised classification pipeline combining consistency regularization and pseudo-labelling, and adapts it for the purpose of segmentation. In our proposed SegMatch, the unlabelled images are first weakly augmented and fed to the segmentation model to generate pseudo-labels. In parallel, images are fed to a strong augmentation branch and consistency between the branches is used as an unsupervised loss. To increase the relevance of our strong augmentations, we depart from using only handcrafted augmentations and introduce a trainable adversarial augmentation strategy. Our FixMatch adaptation for segmentation tasks further includes carefully considering the equivariance and invariance properties of the augmentation functions we rely on. For binary segmentation tasks, our algorithm was evaluated on the MICCAI Instrument Segmentation Challenge datasets, Robust-MIS 2019 and EndoVis 2017. For multi-class segmentation tasks, we relied on the recent CholecInstanceSeg dataset. Our results show that SegMatch outperforms fully-supervised approaches by incorporating unlabelled data, and surpasses a range of state-of-the-art semi-supervised models across different labelled to unlabelled data ratios.
Authors: Yuan Wang, Surya T. Sathujoda, Krzysztof Sawicki, Kanishk Gandhi, Angelica I Aviles-Rivero, Pavlos G. Lagoudakis
Abstract: A plethora of next-generation all-optical devices based on exciton-polaritons have been proposed in latest years, including prototypes of transistors, switches, analogue quantum simulators and others. However, for such systems consisting of multiple polariton condensates, it is still challenging to predict their properties in a fast and accurate manner. The condensate physics is conventionally described by Gross-Pitaevskii equations (GPEs). While GPU-based solvers currently exist, we propose a significantly more efficient machine-learning-based Fourier neural operator approach to find the solution to the GPE coupled with exciton rate equations, trained on both numerical and experimental datasets. The proposed method predicts solutions almost three orders of magnitude faster than CUDA-based solvers in numerical studies, maintaining the high degree of accuracy. Our method not only accelerates simulations but also opens the door to faster, more scalable designs for all-optical chips and devices, offering profound implications for quantum computing, neuromorphic systems, and beyond.
Authors: Mahmud Hasan, Mathias Muia
Abstract: Information Maximizing Generative Adversarial Network (infoGAN) can be understood as a minimax problem involving two neural networks: discriminators and generators with mutual information functions. The infoGAN incorporates various components, including latent variables, mutual information, and objective function. This research demonstrates the Generalization error property of infoGAN as the discriminator and generator sample size approaches infinity. This research explores the generalization error property of InfoGAN as the sample sizes of the discriminator and generator approach infinity. To establish this property, the study considers the difference between the empirical and population versions of the objective function. The error bound is derived from the Rademacher complexity of the discriminator and generator function classes. Additionally, the bound is proven for a two-layer network, where both the discriminator and generator utilize Lipschitz and non-decreasing activation functions.
Authors: Chengzu Li, Han Zhou, Goran Glava\v{s}, Anna Korhonen, Ivan Vuli\'c
Abstract: When adapting ICL with or without fine-tuning, we are curious about whether the instruction-tuned language model is able to achieve well-calibrated results without suffering from the problem of overconfidence (i.e., miscalibration) considering its strong instruction following ability, especially in such limited data setups. In this work, we deliver an in-depth analysis of the behavior across different choices of learning methods from the perspective of both performance and calibration. Through extensive controlled experiments, we observe that the miscalibration problem exists across all learning methods in low-resource setups. To achieve simultaneous gain for both in-task performance and calibration, we then study the potential of self-ensembling applied at different modeling stages (e.g., variations of in-context examples or variations in prompts or different ensembling strategies) to make the predictions more calibrated and have comparable or even better performance. We find that self-ensembling with max probability produces robust and calibrated predictions. Our work reveals the potential calibration problem of using ICL despite the improvements in task performance and sheds light on which learning paradigm to choose. We also provide practical guidelines for choosing learning paradigms depending on whether the data has been seen by the model before and a worthwhile solution via self-ensembling on how to enhance both task performance and calibration of LMs, which we hope could encourage further study.
Authors: Abdoulaye Sakho (LPSM), Emmanuel Malherbe (LPSM), Erwan Scornet (LPSM)
Abstract: Synthetic Minority Oversampling Technique (SMOTE) is a common rebalancing strategy for handling imbalanced tabular data sets. However, few works analyze SMOTE theoretically. In this paper, we derive several non-asymptotic upper bound on SMOTE density. From these results, we prove that SMOTE (with default parameter) tends to copy the original minority samples asymptotically. We confirm and illustrate empirically this first theoretical behavior on a real-world data-set.bFurthermore, we prove that SMOTE density vanishes near the boundary of the support of the minority class distribution. We then adapt SMOTE based on our theoretical findings to introduce two new variants. These strategies are compared on 13 tabular data sets with 10 state-of-the-art rebalancing procedures, including deep generative and diffusion models. One of our key findings is that, for most data sets, applying no rebalancing strategy is competitive in terms of predictive performances, would it be with LightGBM, tuned random forests or logistic regression. However, when the imbalance ratio is artificially augmented, one of our two modifications of SMOTE leads to promising predictive performances compared to SMOTE and other state-of-the-art strategies.
Authors: Nouhaila Innan, Alberto Marchisio, Mohamed Bennai, Muhammad Shafique
Abstract: This study introduces the Quantum Federated Neural Network for Financial Fraud Detection (QFNN-FFD), a cutting-edge framework merging Quantum Machine Learning (QML) and quantum computing with Federated Learning (FL) for financial fraud detection. Using quantum technologies' computational power and the robust data privacy protections offered by FL, QFNN-FFD emerges as a secure and efficient method for identifying fraudulent transactions within the financial sector. Implementing a dual-phase training model across distributed clients enhances data integrity and enables superior performance metrics, achieving precision rates consistently above 95%. Additionally, QFNN-FFD demonstrates exceptional resilience by maintaining an impressive 80% accuracy, highlighting its robustness and readiness for real-world applications. This combination of high performance, security, and robustness against noise positions QFNN-FFD as a transformative advancement in financial technology solutions and establishes it as a new benchmark for privacy-focused fraud detection systems. This framework facilitates the broader adoption of secure, quantum-enhanced financial services and inspires future innovations that could use QML to tackle complex challenges in other areas requiring high confidentiality and accuracy.
Authors: Qianren Li, Bojie Lv, Yuncong Hong, Rui Wang
Abstract: The enhanced distributed channel access (EDCA) mechanism is used in current wireless fidelity (WiFi) networks to support priority requirements of heterogeneous applications. However, the EDCA mechanism can not adapt to particular quality-of-service (QoS) objective, network topology, and interference level. In this paper, a novel reinforcement-learning-based scheduling framework is proposed and implemented to optimize the application-layer quality-of-service (QoS) of a WiFi network with commercial adapters and unknown interference. Particularly, application-layer tasks of file delivery and delay-sensitive communication are jointly scheduled by adjusting the contention window sizes and application-layer throughput limitation, such that the throughput of the former and the round trip time of the latter can be optimized. Due to the unknown interference and vendor-dependent implementation of the WiFi adapters, the relation between the scheduling policy and the system QoS is unknown. Hence, a reinforcement learning method is proposed, in which a novel Q-network is trained to map from the historical scheduling parameters and QoS observations to the current scheduling action. It is demonstrated on a testbed that the proposed framework can achieve a significantly better performance than the EDCA mechanism.
Authors: Yihong Gu, Cong Fang, Peter B\"uhlmann, Jianqing Fan
Abstract: Pursuing causality from data is a fundamental problem in scientific discovery, treatment intervention, and transfer learning. This paper introduces a novel algorithmic method for addressing nonparametric invariance and causality learning in regression models across multiple environments, where the joint distribution of response variables and covariates varies, but the conditional expectations of outcome given an unknown set of quasi-causal variables are invariant. The challenge of finding such an unknown set of quasi-causal or invariant variables is compounded by the presence of endogenous variables that have heterogeneous effects across different environments. The proposed Focused Adversarial Invariant Regularization (FAIR) framework utilizes an innovative minimax optimization approach that drives regression models toward prediction-invariant solutions through adversarial testing. Leveraging the representation power of neural networks, FAIR neural networks (FAIR-NN) are introduced for causality pursuit. It is shown that FAIR-NN can find the invariant variables and quasi-causal variables under a minimal identification condition and that the resulting procedure is adaptive to low-dimensional composition structures in a non-asymptotic analysis. Under a structural causal model, variables identified by FAIR-NN represent pragmatic causality and provably align with exact causal mechanisms under conditions of sufficient heterogeneity. Computationally, FAIR-NN employs a novel Gumbel approximation with decreased temperature and a stochastic gradient descent ascent algorithm. The procedures are demonstrated using simulated and real-data examples.
Authors: Chu Fei Luo, Ahmad Ghawanmeh, Bharat Bhimshetty, Kashyap Murali, Murli Jadhav, Xiaodan Zhu, Faiza Khan Khattak
Abstract: Ensuring the safe deployment of AI systems is critical in industry settings where biased outputs can lead to significant operational, reputational, and regulatory risks. Thorough evaluation before deployment is essential to prevent these hazards. Red-teaming addresses this need by employing adversarial attacks to develop guardrails that detect and reject biased or harmful queries, enabling models to be retrained or steered away from harmful outputs. However, most red-teaming efforts focus on harmful or unethical instructions rather than addressing social bias, leaving this critical area under-explored despite its significant real-world impact, especially in customer-facing systems. We propose two bias-specific red-teaming methods, Emotional Bias Probe (EBP) and BiasKG, to evaluate how standard safety measures for harmful content affect bias. For BiasKG, we refactor natural language stereotypes into a knowledge graph. We use these attacking strategies to induce biased responses from several open- and closed-source language models. Unlike prior work, these methods specifically target social bias. We find our method increases bias in all models, even those trained with safety guardrails. Our work emphasizes uncovering societal bias in LLMs through rigorous evaluation, and recommends measures ensure AI safety in high-stakes industry deployments.
Authors: Nanna E. Hartong, Ilias Sachpazidis, Oliver Blanck, Lucas Etzel, Jan C. Peeken, Stephanie E. Combs, Horst Urbach, Maxim Zaitsev, Dimos Baltas, Ilinca Popp, Anca-Ligia Grosu, Tobias Fechter
Abstract: Background: This study aimed to predict lesion-specific outcomes after stereotactic radiotherapy (SRT) in patients with brain metastases from malignant melanoma (MBM), using clinical, dosimetric pretherapeutic MRI data. Methods: In this multicenter retrospective study, 517 MBM from 130 patients treated with single-fraction or hypofractionated SRT across three centers were analyzed. From contrast-enhanced T1-weighted MRI, 1576 radiomic features (RF) were extracted per lesion - 788 from the gross tumor volume (GTV), 788 from a 3 mm peritumoral margin. Clinical data, radiation dose and RF from one center were used for feature selection and model development via nested cross-validation; external validation was performed using the other two centers. Results: Local failure occurred in 72 of 517 lesions (13.9%). Predictive models based on clinical data (model 1), RF (model 2), or both (model 3) achieved c-indices of 0.60 +/- 0.15, 0.65 +/- 0.11, and 0.65 +/- 0.12. RF-based models outperformed the clinical model, while dosimetric data alone were not predictive. Most predictive RF came from the peritumoral margin (92%) vs. GTV (76%). On the first external dataset, all models performed similarly (c-index: 0.60-0.63), but showed poor generalization on the second (c-index < 0.50), likely due to differences in patient characteristics and imaging protocols. Conclusions: Information extracted from pretherapeutic MRI, particularly from the peritumoral area, can support accurate prediction of lesion-specific outcomes after SRT in MBM. When combined with clinical data, these imaging-derived markers offer valuable prognostic insights. However, generalizability remains challenging by heterogeneity in patient populations and MRI protocols.
Authors: Emanuele Zappala
Abstract: We obtain a new universal approximation theorem for continuous (possibly nonlinear) operators on arbitrary Banach spaces using the Leray-Schauder mapping. Moreover, we introduce and study a method for operator learning in Banach spaces $L^p$ of functions with multiple variables, based on orthogonal projections on polynomial bases. We derive a universal approximation result for operators where we learn a linear projection and a finite dimensional mapping under some additional assumptions. For the case of $p=2$, we give some sufficient conditions for the approximation results to hold. This article serves as the theoretical framework for a deep learning methodology in operator learning.
Authors: Ziyi Huang, Henry Lam, Haofeng Zhang
Abstract: Bayesian bandit algorithms with approximate Bayesian inference have been widely used in real-world applications. Despite the superior practical performance, their theoretical justification is less investigated in the literature, especially for contextual bandit problems. To fill this gap, we propose a theoretical framework to analyze the impact of approximate inference in stochastic linear bandits and conduct frequentist regret analysis on two Bayesian bandit algorithms, Linear Thompson Sampling (LinTS) and the extension of Bayesian Upper Confidence Bound, namely Linear Bayesian Upper Confidence Bound (LinBUCB). We demonstrate that when applied in approximate inference settings, LinTS and LinBUCB can universally preserve their original rates of regret upper bound but with a sacrifice of larger constant terms. These results hold for general Bayesian inference approaches, assuming the inference error measured by two different $\alpha$-divergences is bounded. Additionally, by introducing a new definition of well-behaved distributions, we show that LinBUCB expedites the regret rate of LinTS from $\tilde{O}(d^{3/2}\sqrt{T})$ to $\tilde{O}(d\sqrt{T})$, matching the minimax optimal rate. To our knowledge, this work provides the first regret bounds in the setting of stochastic linear bandits with bounded approximate inference errors.
Authors: D. J. W. Touw, A. Alfons, P. J. F. Groenen, I. Wilms
Abstract: Graphical models serve as effective tools for visualizing conditional dependencies between variables. However, as the number of variables grows, interpretation becomes increasingly difficult, and estimation uncertainty increases due to the large number of parameters relative to the number of observations. To address these challenges, we introduce the Clusterpath estimator of the Gaussian Graphical Model (CGGM) that encourages variable clustering in the graphical model in a data-driven way. Through the use of an aggregation penalty, we group variables together, which in turn results in a block-structured precision matrix whose block structure remains preserved in the covariance matrix. The CGGM estimator is formulated as the solution to a convex optimization problem, making it easy to incorporate other popular penalization schemes which we illustrate through the combination of an aggregation and sparsity penalty. We present a computationally efficient implementation of the CGGM estimator by using a cyclic block coordinate descent algorithm. In simulations, we show that CGGM not only matches, but oftentimes outperforms other state-of-the-art methods for variable clustering in graphical models. We also demonstrate CGGM's practical advantages and versatility on a diverse collection of empirical applications.
Authors: Hong Jun Jeon, Benjamin Van Roy
Abstract: The progress of machine learning over the past decade is undeniable. In retrospect, it is both remarkable and unsettling that this progress was achievable with little to no rigorous theory to guide experimentation. Despite this fact, practitioners have been able to guide their future experimentation via observations from previous large-scale empirical investigations. In this work, we propose a theoretical framework which attempts to provide rigor to existing practices in machine learning. To the theorist, we provide a framework which is mathematically rigorous and leaves open many interesting ideas for future exploration. To the practitioner, we provide a framework whose results are simple, and provide intuition to guide future investigations across a wide range of learning paradigms. Concretely, we provide a theoretical framework rooted in Bayesian statistics and Shannon's information theory which is general enough to unify the analysis of many phenomena in machine learning. Our framework characterizes the performance of an optimal Bayesian learner as it learns from a stream of experience. Unlike existing analyses that weaken with increasing data complexity, our theoretical tools provide accurate insights across diverse machine learning settings. Throughout this work, we derive theoretical results and demonstrate their generality by apply them to derive insights specific to settings. These settings range from learning from data which is independently and identically distributed under an unknown distribution, to data which is sequential, to data which exhibits hierarchical structure amenable to meta-learning, and finally to data which is not fully explainable under the learner's beliefs (misspecification). These results are particularly relevant as we strive to understand and overcome increasingly difficult machine learning challenges in this endlessly complex world.
Authors: Patrick Kage, Jay C. Rothenberger, Pavlos Andreadis, Dimitrios I. Diochnos
Abstract: Deep neural models have achieved state of the art performance on a wide range of problems in computer science, especially in computer vision. However, deep neural networks often require large datasets of labeled samples to generalize effectively, and an important area of active research is semi-supervised learning, which attempts to instead utilize large quantities of (easily acquired) unlabeled samples. One family of methods in this space is pseudo-labeling, a class of algorithms that use model outputs to assign labels to unlabeled samples which are then used as labeled samples during training. Such assigned labels, called pseudo-labels, are most commonly associated with the field of semi-supervised learning. In this work we explore a broader interpretation of pseudo-labels within both self-supervised and unsupervised methods. By drawing the connection between these areas we identify new directions when advancements in one area would likely benefit others, such as curriculum learning and self-supervised regularization.
Authors: Aaron Mark Thomas, Harry Youel, Sharu Theresa Jose
Abstract: Recent proposals for quantum generative adversarial networks (GANs) suffer from the issue of mode collapse, analogous to classical GANs, wherein the distribution learnt by the GAN fails to capture the high mode complexities of the target distribution. Mode collapse can arise due to the use of uninformed prior distributions in the generative learning task. To alleviate the issue of mode collapse for quantum GANs, this work presents a novel \textbf{hybrid quantum-classical generative model}, the VAE-QWGAN, which combines the strengths of a classical Variational AutoEncoder (VAE) with a hybrid Quantum Wasserstein GAN (QWGAN). The VAE-QWGAN fuses the VAE decoder and QWGAN generator into a single quantum model, and utilizes the VAE encoder for data-dependant latent vector sampling during training. This in turn, enhances the diversity and quality of generated images. To generate new data from the trained model at inference, we sample from a Gaussian mixture model (GMM) prior that is learnt on the latent vectors generated during training. We conduct extensive experiments for image generation QGANs on MNIST/Fashion-MNIST datasets and compute a range of metrics that measure the diversity and quality of generated samples. We show that VAE-QWGAN demonstrates significant improvement over existing QGAN approaches.
Authors: Zhiyu Zhang, Zhou Lu, Heng Yang
Abstract: Based on the framework of Conformal Prediction (CP), we study the online construction of confidence sets given a black-box machine learning model. By converting the target confidence levels into quantile levels, the problem can be reduced to predicting the quantiles (in hindsight) of a sequentially revealed data sequence. Two very different approaches have been studied previously: (i) Assuming the data sequence is iid or exchangeable, one could maintain the empirical distribution of the observed data as an algorithmic belief, and directly predict its quantiles. (ii) Due to the fragility of statistical assumptions, a recent trend is to consider the non-distributional, adversarial setting and apply first-order online optimization algorithms to moving quantile losses. However, it requires the oracle knowledge of the target quantile level, and suffers from a previously overlooked monotonicity issue due to the associated loss linearization. This paper presents an adaptive CP algorithm that combines their strengths. Without any statistical assumption, it is able to answer multiple arbitrary confidence level queries with low regret, while also overcoming the monotonicity issue suffered by first-order optimization baselines. Furthermore, if the data sequence is actually iid, then the same algorithm is automatically equipped with the "correct" coverage probability guarantee. To achieve such strengths, our key technical innovation is to regularize the aforementioned algorithmic belief (the empirical distribution) by a Bayesian prior, which robustifies it by simulating a non-linearized Follow the Regularized Leader (FTRL) algorithm on the output. Such a belief update backbone is shared by prediction heads targeting different confidence levels, bringing practical benefits analogous to the recently proposed concept of U-calibration (Kleinberg et al., 2023).
Authors: Thibaud Gloaguen, Nikola Jovanovi\'c, Robin Staab, Martin Vechev
Abstract: LLM watermarks stand out as a promising way to attribute ownership of LLM-generated text. One threat to watermark credibility comes from spoofing attacks, where an unauthorized third party forges the watermark, enabling it to falsely attribute arbitrary texts to a particular LLM. Despite recent work demonstrating that state-of-the-art schemes are, in fact, vulnerable to spoofing, no prior work has focused on post-hoc methods to discover spoofing attempts. In this work, we for the first time propose a reliable statistical method to distinguish spoofed from genuinely watermarked text, suggesting that current spoofing attacks are less effective than previously thought. In particular, we show that regardless of their underlying approach, all current learning-based spoofing methods consistently leave observable artifacts in spoofed texts, indicative of watermark forgery. We build upon these findings to propose rigorous statistical tests that reliably reveal the presence of such artifacts and thus demonstrate that a watermark has been spoofed. Our experimental evaluation shows high test power across all learning-based spoofing methods, providing insights into their fundamental limitations and suggesting a way to mitigate this threat. We make all our code available at https://github.com/eth-sri/watermark-spoofing-detection .
URLs: https://github.com/eth-sri/watermark-spoofing-detection
Authors: Eilam Shapira, Omer Madmon, Itamar Reinman, Samuel Joseph Amouyal, Roi Reichart, Moshe Tennenholtz
Abstract: Large Language Models (LLMs) show significant potential in economic and strategic interactions, where communication via natural language is often prevalent. This raises key questions: Do LLMs behave rationally? How do they perform compared to humans? Do they tend to reach an efficient and fair outcome? What is the role of natural language in strategic interaction? How do characteristics of the economic environment influence these dynamics? These questions become crucial concerning the economic and societal implications of integrating LLM-based agents into real-world data-driven systems, such as online retail platforms and recommender systems. To answer these questions, we introduce a benchmark for standardizing research on two-player, sequential, language-based games. Inspired by the economic literature, we define three base families of games with consistent parameterization, degrees of freedom and economic measures to evaluate agents' performance (self-gain), as well as the game outcome (efficiency and fairness). We develop an open-source framework for interaction simulation and analysis, and utilize it to collect a dataset of LLM vs. LLM interactions across numerous game configurations and an additional dataset of human vs. LLM interactions. Through extensive experimentation, we demonstrate how our framework and dataset can be used to: (i) compare the behavior of LLM-based agents in various economic contexts; (ii) evaluate agents in both individual and collective performance measures; and (iii) quantify the effect of the economic characteristics of the environments on the behavior of agents. Our results suggest that the market parameters, as well as the choice of the LLMs, tend to have complex and interdependent effects on the economic outcome, which calls for careful design and analysis of the language-based economic ecosystem.
Authors: Fanmeng Wang, Minjie Cheng, Hongteng Xu
Abstract: Predicting molecular ground-state conformation (i.e., energy-minimized conformation) is crucial for many chemical applications such as molecular docking and property prediction. Classic energy-based simulation is time-consuming when solving this problem, while existing learning-based methods have advantages in computational efficiency but sacrifice accuracy and interpretability. In this work, we propose a novel and effective method to bridge the energy-based simulation and the learning-based strategy, which designs and learns a Wasserstein gradient flow-driven SE(3)-Transformer, called WGFormer, for ground-state conformation prediction. Specifically, our method tackles this task within an auto-encoding framework, which encodes low-quality conformations by the proposed WGFormer and decodes corresponding ground-state conformations by an MLP. The architecture of WGFormer corresponds to Wasserstein gradient flows -- it optimizes conformations by minimizing an energy function defined on the latent mixture models of atoms, thereby significantly improving performance and interpretability. Extensive experiments demonstrate that our method consistently outperforms state-of-the-art competitors, providing a new and insightful paradigm to predict ground-state conformation.
Authors: Ali Azizpour, Nicolas Zilberstein, Santiago Segarra
Abstract: Graphons are continuous models that represent the structure of graphs and allow the generation of graphs of varying sizes. We propose Scalable Implicit Graphon Learning (SIGL), a scalable method that combines implicit neural representations (INRs) and graph neural networks (GNNs) to estimate a graphon from observed graphs. Unlike existing methods, which face important limitations like fixed resolution and scalability issues, SIGL learns a continuous graphon at arbitrary resolutions. GNNs are used to determine the correct node ordering, improving graph alignment. Furthermore, we characterize the asymptotic consistency of our estimator, showing that more expressive INRs and GNNs lead to consistent estimators. We evaluate SIGL in synthetic and real-world graphs, showing that it outperforms existing methods and scales effectively to larger graphs, making it ideal for tasks like graph data augmentation.
Authors: Jerry Huang, Prasanna Parthasarathi, Mehdi Rezagholizadeh, Boxing Chen, Sarath Chandar
Abstract: The growth in prominence of large language models (LLMs) in everyday life can be largely attributed to their generative abilities, yet some of this is also owed to the risks and costs associated with their use. On one front is their tendency to hallucinate false or misleading information, limiting their reliability. On another is the increasing focus on the computational limitations associated with traditional self-attention based LLMs, which has brought about new alternatives, in particular recurrent models, meant to overcome them. Yet it remains uncommon to consider these two concerns simultaneously. Do changes in architecture exacerbate/alleviate existing concerns about hallucinations? Do they affect how and where they occur? Through an extensive evaluation, we study how these architecture-based inductive biases affect the propensity to hallucinate. While hallucination remains a general phenomenon not limited to specific architectures, the situations in which they occur and the ease with which specific types of hallucinations can be induced can significantly differ based on the model architecture. These findings highlight the need for better understanding both these problems in conjunction with each other, as well as consider how to design more universal techniques for handling hallucinations.
Authors: Yukun Li, Sijia Wang, Lifu Huang, Li-Ping Liu
Abstract: Reliable confidence estimation is essential for enhancing the trustworthiness of large language models (LLMs), especially in high-stakes scenarios. Despite its importance, accurately estimating confidence in LLM responses remains a significant challenge. In this work, we propose using an auxiliary learning model to assess response correctness based on the self-consistency of multiple outputs generated by the LLM. Our method builds a consistency graph to represent the agreement among multiple responses and uses a graph neural network (GNN) to estimate the likelihood that each response is correct. Experiments demonstrate that this method has strong calibration performance on various benchmark datasets and generalizes well to out-of-domain cases.
Authors: Zakariae Belmekki, Jun Li, Patrick Reuter, David Antonio G\'omez J\'auregui, Karl Jenkins
Abstract: It has been observed that Convolutional Neural Networks (CNNs) suffer from redundancy in feature maps, leading to inefficient capacity utilization. Efforts to address this issue have largely focused on kernel orthogonality method. In this work, we theoretically and empirically demonstrate that kernel orthogonality does not necessarily lead to a reduction in feature map redundancy. Based on this analysis, we propose the Convolutional Similarity method to reduce feature map similarity, independently of the CNN's input. The Convolutional Similarity can be minimized as either a regularization term or an iterative initialization method. Experimental results show that minimizing Convolutional Similarity not only improves classification accuracy but also accelerates convergence. Furthermore, our method enables the use of significantly smaller models to achieve the same level of performance, promoting a more efficient use of model capacity. Future work will focus on coupling the iterative initialization method with the optimization momentum term and examining the method's impact on generative frameworks.
Authors: Dionysis Christopoulos, Sotiris Spanos, Valsamis Ntouskos, Konstantinos Karantzalos
Abstract: We present TRACE (Transformer-based Risk Assessment for Clinical Evaluation), a novel method for clinical risk assessment based on clinical data, leveraging the self-attention mechanism for enhanced feature interaction and result interpretation. Our approach is able to handle different data modalities, including continuous, categorical and multiple-choice (checkbox) attributes. The proposed architecture features a shared representation of the clinical data obtained by integrating specialized embeddings of each data modality, enabling the detection of high-risk individuals using Transformer encoder layers. To assess the effectiveness of the proposed method, a strong baseline based on non-negative multi-layer perceptrons (MLPs) is introduced. The proposed method outperforms various baselines widely used in the domain of clinical risk assessment, while effectively handling missing values. In terms of explainability, our Transformer-based method offers easily interpretable results via attention weights, further enhancing the clinicians' decision-making process.
Authors: Qiyao Ma, Menglin Yang, Mingxuan Ju, Tong Zhao, Neil Shah, Rex Ying
Abstract: Modern recommender systems often create information cocoons, restricting users' exposure to diverse content. A key challenge lies in balancing content exploration and exploitation while allowing users to adjust their recommendation preferences. Intuitively, this balance can be modeled as a tree-structured representation, where depth search facilitates exploitation and breadth search enables exploration. However, existing approaches face two fundamental limitations: Euclidean methods struggle to capture hierarchical structures, while hyperbolic methods, despite their superior hierarchical modeling, lack semantic understanding of user and item profiles and fail to provide a principled mechanism for balancing exploration and exploitation. To address these challenges, we propose HERec, a hyperbolic graph-LLM framework that effectively balances exploration and exploitation in recommender systems. Our framework introduces two key innovations: (1) a semantic-enhanced hierarchical mechanism that aligns rich textual descriptions processed by large language models (LLMs) with collaborative information directly in hyperbolic space, allowing for more nuanced updates that respect the underlying hierarchical structure in user-item profiles; (2) an automatic hierarchical representation by optimizing Dasgupta's cost, which discovers hierarchical structures without requiring predefined hyperparameters, enabling user-adjustable exploration-exploitation trade-offs. Extensive experiments demonstrate that HERec consistently outperforms both Euclidean and hyperbolic baselines, achieving up to 5.49% improvement in utility metrics and 11.39% increase in diversity metrics, effectively mitigating information cocoons. We open-source our model implementation at https://github.com/Martin-qyma/HERec.
Authors: Haiming Zhang, Wending Zhou, Yiyao Zhu, Xu Yan, Jiantao Gao, Dongfeng Bai, Yingjie Cai, Bingbing Liu, Shuguang Cui, Zhen Li
Abstract: This paper introduces VisionPAD, a novel self-supervised pre-training paradigm designed for vision-centric algorithms in autonomous driving. In contrast to previous approaches that employ neural rendering with explicit depth supervision, VisionPAD utilizes more efficient 3D Gaussian Splatting to reconstruct multi-view representations using only images as supervision. Specifically, we introduce a self-supervised method for voxel velocity estimation. By warping voxels to adjacent frames and supervising the rendered outputs, the model effectively learns motion cues in the sequential data. Furthermore, we adopt a multi-frame photometric consistency approach to enhance geometric perception. It projects adjacent frames to the current frame based on rendered depths and relative poses, boosting the 3D geometric representation through pure image supervision. Extensive experiments on autonomous driving datasets demonstrate that VisionPAD significantly improves performance in 3D object detection, occupancy prediction and map segmentation, surpassing state-of-the-art pre-training strategies by a considerable margin.
Authors: Gianluca Capozzi, Tong Tang, Jie Wan, Ziqi Yang, Daniele Cono D'Elia, Giuseppe Antonio Di Luna, Lorenzo Cavallaro, Leonardo Querzoni
Abstract: Binary function similarity, which often relies on learning-based algorithms to identify what functions in a pool are most similar to a given query function, is a sought-after topic in different communities, including machine learning, software engineering, and security. Its importance stems from the impact it has in facilitating several crucial tasks, from reverse engineering and malware analysis to automated vulnerability detection. Whereas recent work cast light around performance on this long-studied problem, the research landscape remains largely lackluster in understanding the resiliency of the state-of-the-art machine learning models against adversarial attacks. As security requires to reason about adversaries, in this work we assess the robustness of such models through a simple yet effective black-box greedy attack, which modifies the topology and the content of the control flow of the attacked functions. We demonstrate that this attack is successful in compromising all the models, achieving average attack success rates of 57.06% and 95.81% depending on the problem settings (targeted and untargeted attacks). Our findings are insightful: top performance on clean data does not necessarily relate to top robustness properties, which explicitly highlights performance-robustness trade-offs one should consider when deploying such models, calling for further research.
Authors: Jiayu Chen, Chao Yu, Yuqing Xie, Feng Gao, Yinuo Chen, Shu'ang Yu, Wenhao Tang, Shilong Ji, Mo Mu, Yi Wu, Huazhong Yang, Yu Wang
Abstract: Executing precise and agile flight maneuvers is critical for quadrotors in various applications. Traditional quadrotor control approaches are limited by their reliance on flat trajectories or time-consuming optimization, which restricts their flexibility. Recently, RL-based policy has emerged as a promising alternative due to its ability to directly map observations to actions, reducing the need for detailed system knowledge and actuation constraints. However, a significant challenge remains in bridging the sim-to-real gap, where RL-based policies often experience instability when deployed in real world. In this paper, we investigate key factors for learning robust RL-based control policies that are capable of zero-shot deployment in real-world quadrotors. We identify five critical factors and we develop a PPO-based training framework named SimpleFlight, which integrates these five techniques. We validate the efficacy of SimpleFlight on Crazyflie quadrotor, demonstrating that it achieves more than a 50% reduction in trajectory tracking error compared to state-of-the-art RL baselines. The policy derived by SimpleFlight consistently excels across both smooth polynominal trajectories and challenging infeasible zigzag trajectories on small thrust-to-weight quadrotors. In contrast, baseline methods struggle with high-speed or infeasible trajectories. To support further research and reproducibility, we integrate SimpleFlight into a GPU-based simulator Omnidrones and provide open-source access to the code and model checkpoints. We hope SimpleFlight will offer valuable insights for advancing RL-based quadrotor control. For more details, visit our project website at https://sites.google.com/view/simpleflight/.
Authors: Inpyo Hong, Youngwan Jo, Hyojeong Lee, Sunghyun Ahn, Sanghyun Park
Abstract: We introduce AKT (Advanced Knowledge Transfer), a novel method to enhance the training ability of low-bit quantized (Q) models in the field of zero-shot quantization (ZSQ). Existing research in ZSQ has focused on generating high-quality data from full-precision (FP) models. However, these approaches struggle with reduced learning ability in low-bit quantization due to its limited information capacity. To overcome this limitation, we propose effective training strategy compared to data generation. Particularly, we analyzed that refining feature maps in the feature distillation process is an effective way to transfer knowledge to the Q model. Based on this analysis, AKT efficiently transfer core information from the FP model to the Q model. AKT is the first approach to utilize both spatial and channel attention information in feature distillation in ZSQ. Our method addresses the fundamental gradient exploding problem in low-bit Q models. Experiments on CIFAR-10 and CIFAR-100 datasets demonstrated the effectiveness of the AKT. Our method led to significant performance enhancement in existing generative models. Notably, AKT achieved significant accuracy improvements in low-bit Q models, achieving state-of-the-art in the 3,5bit scenarios on CIFAR-10. The code is available at https://github.com/Inpyo-Hong/AKT-Advanced-knowledge-Transfer.
URLs: https://github.com/Inpyo-Hong/AKT-Advanced-knowledge-Transfer.
Authors: Han Qi, Haocheng Yin, Aris Zhu, Yilun Du, Heng Yang
Abstract: We present generative predictive control (GPC), a learning control framework that (i) clones a generative diffusion-based policy from expert demonstrations, (ii) trains a predictive action-conditioned world model from both expert demonstrations and random explorations, and (iii) synthesizes an online planner that ranks and optimizes the action proposals from (i) by looking ahead into the future using the world model from (ii). Across a variety of robotic manipulation tasks, we demonstrate that GPC consistently outperforms behavior cloning in both state-based and vision-based settings, in simulation and in the real world.
Authors: Haozhe Wang, Long Li, Chao Qu, Fengming Zhu, Weidi Xu, Wei Chu, Fangzhen Lin
Abstract: Recent advances in mathematical problem-solving with language models (LMs) integrate chain-of-thought (CoT) reasoning and code execution to harness their complementary strengths. However, existing hybrid frameworks exhibit a critical limitation: they depend on externally dictated instructions or rigid code-integration templates, lacking metacognitive awareness -- the capacity to dynamically evaluate intrinsic capabilities and autonomously determine when and how to integrate tools. This rigidity motivates our study of autonomous code integration, enabling models to adapt tool-usage strategies as their reasoning abilities evolve during training. While reinforcement learning (RL) shows promise for boosting LLM reasoning at scale (e.g., DeepSeek-R1), we demonstrate its inefficiency in learning autonomous code integration due to inadequate exploration of the vast combinatorial space of CoT-code interleaving patterns. To address this challenge, we propose a novel Expectation-Maximization (EM) framework that synergizes structured exploration (E-step) with off-policy RL optimization (M-step), creating a self-reinforcing cycle between metacognitive tool-use decisions and evolving capabilities. Experiments reveal our method achieves superior results through improved exploration. Notably, our 7B model improves over 11% on MATH500 and 9.4% on AIME without o1-like CoT.
Authors: Zhizhen Zhang, Lei Zhu, Zhen Fang, Zi Huang, Yadan Luo
Abstract: Pre-training vision-language representations on human action videos has emerged as a promising approach to reduce reliance on large-scale expert demonstrations for training embodied agents. However, prior methods often employ time contrastive learning based on goal-reaching heuristics, progressively aligning language instructions from the initial to the final frame. This overemphasis on future frames can result in erroneous vision-language associations, as actions may terminate early or include irrelevant moments in the end. To address this issue, we propose Action Temporal Coherence Learning (AcTOL) to learn ordered and continuous vision-language representations without rigid goal-based constraint. AcTOL treats a video as a continuous trajectory where it (1) contrasts semantic differences between frames to reflect their natural ordering, and (2) imposes a local Brownian bridge constraint to ensure smooth transitions across intermediate frames. Extensive imitation learning experiments on both simulated and real robots show that the pretrained features significantly enhance downstream manipulation tasks with high robustness to different linguistic styles of instructions, offering a viable pathway toward generalized embodied agents.
Authors: Yimu Zhang, Dongqi Han, Yansen Wang, Zhenning Lv, Yu Gu, Dongsheng Li
Abstract: Spike sorting is an essential process in neural recording, which identifies and separates electrical signals from individual neurons recorded by electrodes in the brain, enabling researchers to study how specific neurons communicate and process information. Although there exist a number of spike sorting methods which have contributed to significant neuroscientific breakthroughs, many are heuristically designed, making it challenging to verify their correctness due to the difficulty of obtaining ground truth labels from real-world neural recordings. In this work, we explore a data-driven, deep learning-based approach. We begin by creating a large-scale dataset through electrophysiology simulations using biologically realistic computational models. We then present SimSort, a pretraining framework for spike sorting. Trained solely on simulated data, SimSort demonstrates zero-shot generalizability to real-world spike sorting tasks, yielding consistent improvements over existing methods across multiple benchmarks. These results highlight the potential of simulation-driven pretraining to enhance the robustness and scalability of spike sorting in experimental neuroscience.
Authors: Zhenqing Ling, Daoyuan Chen, Liuyi Yao, Qianli Shen, Yaliang Li, Ying Shen
Abstract: Fine-tuning large language models (LLMs) using diverse datasets is crucial for enhancing their overall performance across various domains. In practical scenarios, existing methods based on modeling the mixture proportions of data composition often struggle with data whose domain labels are missing, imprecise or non-normalized, while methods based on data selection usually encounter difficulties in balancing multi-domain performance. To address these challenges, in this work, we investigate the role of data diversity in enhancing the overall abilities of LLMs by empirically constructing contrastive data pools and theoretically deriving explanations. Building upon the insights gained, we propose a new method that gives the LLM a dual identity: an output model to cognitively probe and select data based on diversity reward, as well as an input model to be tuned with the selected data. Extensive experiments show that the proposed method notably boosts performance across domain-undetermined data and a series of foundational downstream tasks when applied to various advanced LLMs. We release our code and hope this study can shed light on the understanding of data diversity and advance feedback-driven data-model co-design for LLMs.
Authors: Guoxin Chen, Minpeng Liao, Peiying Yu, Dingmin Wang, Zile Qiao, Chao Yang, Xin Zhao, Kai Fan
Abstract: Retrieval-augmented generation (RAG) systems face a fundamental challenge in aligning independently developed retrievers and large language models (LLMs). Existing approaches typically involve modifying either component or introducing simple intermediate modules, resulting in practical limitations and sub-optimal performance. Inspired by human search behavior -- typically involving a back-and-forth process of proposing search queries and reviewing documents, we propose C-3PO, a proxy-centric framework that facilitates communication between retrievers and LLMs through a lightweight multi-agent system. Our framework implements three specialized agents that collaboratively optimize the entire RAG pipeline without altering the retriever and LLMs. These agents work together to assess the need for retrieval, generate effective queries, and select information suitable for the LLMs. To enable effective multi-agent coordination, we develop a tree-structured rollout approach for reward credit assignment in reinforcement learning. Extensive experiments in both in-domain and out-of-distribution scenarios demonstrate that C-3PO significantly enhances RAG performance while maintaining plug-and-play flexibility and superior generalization capabilities.
Authors: Ruihan Xu, Yiping Lu
Abstract: Randomized sketching accelerates large-scale numerical linear algebra by reducing computational complexity. While the traditional sketch-and-solve approach reduces the problem size directly through sketching, the sketch-and-precondition method leverages sketching to construct a computational friendly preconditioner. This preconditioner improves the convergence speed of iterative solvers applied to the original problem, maintaining accuracy in the full space. Furthermore, the convergence rate of the solver improves at least linearly with the sketch size. Despite its potential, developing a sketch-and-precondition framework for randomized algorithms in low-rank matrix approximation remains an open challenge. We introduce the Error-Powered Sketched Inverse Iteration (EPSI) Method via run sketched Newton iteration for the Lagrange form as a sketch-and-precondition variant for randomized low-rank approximation. Our method achieves theoretical guarantees, including a convergence rate that improves at least linearly with the sketch size.
Authors: Yung-Sung Chuang, Benjamin Cohen-Wang, Shannon Zejiang Shen, Zhaofeng Wu, Hu Xu, Xi Victoria Lin, James Glass, Shang-Wen Li, Wen-tau Yih
Abstract: We introduce SelfCite, a novel self-supervised approach that aligns LLMs to generate high-quality, fine-grained, sentence-level citations for the statements in their generated responses. Instead of only relying on costly and labor-intensive annotations, SelfCite leverages a reward signal provided by the LLM itself through context ablation: If a citation is necessary, removing the cited text from the context should prevent the same response; if sufficient, retaining the cited text alone should preserve the same response. This reward can guide the inference-time best-of-N sampling strategy to improve citation quality significantly, as well as be used in preference optimization to directly fine-tune the models for generating better citations. The effectiveness of SelfCite is demonstrated by increasing citation F1 up to 5.3 points on the LongBench-Cite benchmark across five long-form question answering tasks. The source code is available at https://github.com/facebookresearch/SelfCite
Authors: Kumar Manas, Adrian Paschke
Abstract: This comprehensive survey examines the integration of knowledge-based approaches in autonomous driving systems, specifically focusing on trajectory prediction and planning. We extensively analyze various methodologies for incorporating domain knowledge, traffic rules, and commonsense reasoning into autonomous driving systems. The survey categorizes and analyzes approaches based on their knowledge representation and integration methods, ranging from purely symbolic to hybrid neuro-symbolic architectures. We examine recent developments in logic programming, foundation models for knowledge representation, reinforcement learning frameworks, and other emerging technologies incorporating domain knowledge. This work systematically reviews recent approaches, identifying key challenges, opportunities, and future research directions in knowledge-enhanced autonomous driving systems. Our analysis reveals emerging trends in the field, including the increasing importance of interpretable AI, the role of formal verification in safety-critical systems, and the potential of hybrid approaches that combine traditional knowledge representation with modern machine learning techniques.
Authors: Chen-An Li, Tzu-Han Lin, Yun-Nung Chen, Hung-yi Lee
Abstract: Large vision-language models (LVLMs) perform outstandingly across various multimodal tasks. However, their ability to evaluate generated content remains limited, and training vision-language reward models (VLRMs) with preference data is computationally expensive. This paper explores a training-free alternative by merging text-based reward models (RMs) with LVLMs to create VLRMs. Our approach shows that integrating these models leads to improved performance over LVLMs' scoring and text-based RMs, offering an efficient method for incorporating textual preferences into LVLMs.
Authors: Vignesh Kothapalli, Hamed Firooz, Maziar Sanjabi
Abstract: We introduce CoT-ICL Lab, a framework and methodology to generate synthetic tokenized datasets and systematically study chain-of-thought (CoT) in-context learning (ICL) in language models. CoT-ICL Lab allows fine grained control over the complexity of in-context examples by decoupling (1) the causal structure involved in chain token generation from (2) the underlying token processing functions. We train decoder-only transformers (up to 700M parameters) on these datasets and show that CoT accelerates the accuracy transition to higher values across model sizes. In particular, we find that model depth is crucial for leveraging CoT with limited in-context examples, while more examples help shallow models match deeper model performance. Additionally, limiting the diversity of token processing functions throughout training improves causal structure learning via ICL. We also interpret these transitions by analyzing transformer embeddings and attention maps. Overall, CoT-ICL Lab serves as a simple yet powerful testbed for theoretical and empirical insights into ICL and CoT in language models.
Authors: Yang Xu, Washim Uddin Mondal, Vaneet Aggarwal
Abstract: We present the first finite-sample analysis for policy evaluation in robust average-reward Markov Decision Processes (MDPs). Prior works in this setting have established only asymptotic convergence guarantees, leaving open the question of sample complexity. In this work, we address this gap by establishing that the robust Bellman operator is a contraction under the span semi-norm, and developing a stochastic approximation framework with controlled bias. Our approach builds upon Multi-Level Monte Carlo (MLMC) techniques to estimate the robust Bellman operator efficiently. To overcome the infinite expected sample complexity inherent in standard MLMC, we introduce a truncation mechanism based on a geometric distribution, ensuring a finite constant sample complexity while maintaining a small bias that decays exponentially with the truncation level. Our method achieves the order-optimal sample complexity of $\tilde{\mathcal{O}}(\epsilon^{-2})$ for robust policy evaluation and robust average reward estimation, marking a significant advancement in robust reinforcement learning theory.
Authors: Edward Milsom, Ben Anson, Laurence Aitchison
Abstract: We consider layerwise function-space learning rates, which measure the magnitude of the change in a neural network's output function in response to an update to a parameter tensor. This contrasts with traditional learning rates, which describe the magnitude of changes in parameter space. We develop efficient methods to measure and set function-space learning rates in arbitrary neural networks, requiring only minimal computational overhead through a few additional backward passes that can be performed at the start of, or periodically during, training. We demonstrate two key applications: (1) analysing the dynamics of standard neural network optimisers in function space, rather than parameter space, and (2) introducing FLeRM (Function-space Learning Rate Matching), a novel approach to hyperparameter transfer across model scales. FLeRM records function-space learning rates while training a small, cheap base model, then automatically adjusts parameter-space layerwise learning rates when training larger models to maintain consistent function-space updates. FLeRM gives hyperparameter transfer across model width, depth, initialisation scale, and LoRA rank in various architectures including MLPs with residual connections and transformers with different layer normalisation schemes.
Authors: Huimin Yan, Xian Yang, Liang Bai, Jiye Liang
Abstract: Local alignment between medical images and text is essential for accurate diagnosis, though it remains challenging due to the absence of natural local pairings and the limitations of rigid region recognition methods. Traditional approaches rely on hard boundaries, which introduce uncertainty, whereas medical imaging demands flexible soft region recognition to handle irregular structures. To overcome these challenges, we propose the Progressive Local Alignment Network (PLAN), which designs a novel contrastive learning-based approach for local alignment to establish meaningful word-pixel relationships and introduces a progressive learning strategy to iteratively refine these relationships, enhancing alignment precision and robustness. By combining these techniques, PLAN effectively improves soft region recognition while suppressing noise interference. Extensive experiments on multiple medical datasets demonstrate that PLAN surpasses state-of-the-art methods in phrase grounding, image-text retrieval, object detection, and zero-shot classification, setting a new benchmark for medical image-text alignment.
Authors: Yuan Sui, Yufei He, Tri Cao, Simeng Han, Yulin Chen, Bryan Hooi
Abstract: Large Language Models (LLMs) increasingly rely on prolonged reasoning chains to solve complex tasks. However, this trial-and-error approach often leads to high computational overhead and error propagation, where early mistakes can derail subsequent steps. To address these issues, we introduce Meta-Reasoner, a framework that dynamically optimizes inference-time reasoning by enabling LLMs to \enquote{think about how to think.} Drawing inspiration from human meta-cognition and dual-process theory, Meta-Reasoner operates as a strategic advisor, decoupling high-level guidance from step-by-step generation. It employs contextual multi-armed bandits to iteratively evaluate reasoning progress and select optimal strategies (e.g., backtrack, clarify ambiguity, restart from scratch, or propose alternative approaches), and reallocates computational resources toward the most promising paths. Our evaluations on mathematical reasoning and puzzles highlight the potential of dynamic reasoning chains to overcome inherent challenges in the LLM reasoning process and also show promise in broader applications, offering a scalable and adaptable solution for reasoning-intensive tasks.
Authors: Zhijian Zhuo, Yutao Zeng, Ya Wang, Sijun Zhang, Jian Yang, Xiaoqing Li, Xun Zhou, Jinwen Ma
Abstract: Transformers have become the de facto architecture for a wide range of machine learning tasks, particularly in large language models (LLMs). Despite their remarkable performance, challenges remain in training deep transformer networks, especially regarding the position of layer normalization. While Pre-Norm structures facilitate more stable training owing to their stronger identity path, they often lead to suboptimal performance compared to Post-Norm. In this paper, we propose $\textbf{HybridNorm}$, a simple yet effective hybrid normalization strategy that integrates the advantages of both Pre-Norm and Post-Norm. Specifically, HybridNorm employs QKV normalization within the attention mechanism and Post-Norm in the feed-forward network (FFN) of each transformer block. We provide both theoretical insights and empirical evidence demonstrating that HybridNorm improves gradient flow and model robustness. Extensive experiments on large-scale transformer models, including both dense and sparse variants, show that HybridNorm consistently outperforms both Pre-Norm and Post-Norm approaches across multiple benchmarks. These findings highlight the potential of HybridNorm as a more stable and effective technique for improving the training and performance of deep transformer models. Code is available at https://github.com/BryceZhuo/HybridNorm.
Authors: Sophia Hager, David Mueller, Kevin Duh, Nicholas Andrews
Abstract: As large language models (LLMs) are increasingly used for factual question-answering, it becomes more important for LLMs to have the capability to communicate the likelihood that their answer is correct. For these verbalized expressions of uncertainty to be meaningful, they should reflect the error rates at the expressed level of confidence. However, when prompted to express confidence, the error rates of current LLMs are inconsistent with their communicated confidences, highlighting the need for uncertainty quantification methods. Many prior methods calculate lexical uncertainty, estimating a model's confidence in the specific string it generated. In some cases, however, it may be more useful to estimate semantic uncertainty, or the model's confidence in the answer regardless of how it is verbalized. We propose a simple procedure, uncertainty distillation, to teach an LLM to verbalize calibrated semantic confidences. Using held-out data to map initial uncertainty estimates to meaningful probabilities, we create examples annotated with verbalized probabilities for supervised fine-tuning. We compare uncertainty distillation to several strong baselines, and find that our method yields verbalized confidences that correlate well with observed error rates.
Authors: Siyuan Guo, Huiwu Liu, Xiaolong Chen, Yuming Xie, Liang Zhang, Tao Han, Hechang Chen, Yi Chang, Jun Wang
Abstract: In this work, we explore the potential of large language models (LLMs) for generating functional test scripts, which necessitates understanding the dynamically evolving code structure of the target software. To achieve this, we propose a case-based reasoning (CBR) system utilizing a 4R cycle (i.e., retrieve, reuse, revise, and retain), which maintains and leverages a case bank of test intent descriptions and corresponding test scripts to facilitate LLMs for test script generation. To improve user experience further, we introduce Re4, an optimization method for the CBR system, comprising reranking-based retrieval finetuning and reinforced reuse finetuning. Specifically, we first identify positive examples with high semantic and script similarity, providing reliable pseudo-labels for finetuning the retriever model without costly labeling. Then, we apply supervised finetuning, followed by a reinforcement learning finetuning stage, to align LLMs with our production scenarios, ensuring the faithful reuse of retrieved cases. Extensive experimental results on two product development units from Huawei Datacom demonstrate the superiority of the proposed CBR+Re4. Notably, we also show that the proposed Re4 method can help alleviate the repetitive generation issues with LLMs.
Authors: Ram Mohan Rao Kadiyala, Siddartha Pullakhandam, Kanwal Mehreen, Drishti Sharma, Siddhant Gupta, Jebish Purbey, Ashay Srivastava, Subhasya TippaReddy, Arvind Reddy Bobbili, Suraj Telugara Chandrashekhar, Modabbir Adeeb, Srinadh Vura, Hamza Farooq
Abstract: An ideal detection system for machine generated content is supposed to work well on any generator as many more advanced LLMs come into existence day by day. Existing systems often struggle with accurately identifying AI-generated content over shorter texts. Further, not all texts might be entirely authored by a human or LLM, hence we focused more over partial cases i.e human-LLM co-authored texts. Our paper introduces a set of models built for the task of token classification which are trained on an extensive collection of human-machine co-authored texts, which performed well over texts of unseen domains, unseen generators, texts by non-native speakers and those with adversarial inputs. We also introduce a new dataset of over 2.4M such texts mostly co-authored by several popular proprietary LLMs over 23 languages. We also present findings of our models' performance over each texts of each domain and generator. Additional findings include comparison of performance against each adversarial method, length of input texts and characteristics of generated texts compared to the original human authored texts.
Authors: Yilun Zhou, Austin Xu, Peifeng Wang, Caiming Xiong, Shafiq Joty
Abstract: Scaling test-time computation, or affording a generator large language model (LLM) extra compute during inference, typically employs the help of external non-generative evaluators (i.e., reward models). Concurrently, LLM-judges, models trained to generate evaluations and critiques (explanations) in natural language, are becoming increasingly popular in automatic evaluation. Despite judge empirical successes, their effectiveness as evaluators in test-time scaling settings is largely unknown. In this paper, we introduce the Judge Evaluation for Test-Time Scaling (JETTS) benchmark, which evaluates judge performance in three domains (math reasoning, code generation, and instruction following) under three task settings: response reranking, step-level beam search, and critique-based response refinement. We evaluate 10 different judge models (7B-70B parameters) for 8 different base generator models (6.7B-72B parameters). Our benchmark shows that while judges are competitive with outcome reward models in reranking, they are consistently worse than process reward models in beam search procedures. Furthermore, though unique to LLM-judges, their natural language critiques are currently ineffective in guiding the generator towards better responses.
Authors: Yuxin Zuo, Kaiyan Zhang, Li Sheng, Shang Qu, Ganqu Cui, Xuekai Zhu, Haozhan Li, Yuchen Zhang, Xinwei Long, Ermo Hua, Biqing Qi, Youbang Sun, Zhiyuan Ma, Lifan Yuan, Ning Ding, Bowen Zhou
Abstract: This paper investigates Reinforcement Learning (RL) on data without explicit labels for reasoning tasks in Large Language Models (LLMs). The core challenge of the problem is reward estimation during inference while not having access to ground-truth information. While this setting appears elusive, we find that common practices in Test-Time Scaling (TTS), such as majority voting, yield surprisingly effective rewards suitable for driving RL training. In this work, we introduce Test-Time Reinforcement Learning (TTRL), a novel method for training LLMs using RL on unlabeled data. TTRL enables self-evolution of LLMs by utilizing the priors in the pre-trained models. Our experiments demonstrate that TTRL consistently improves performance across a variety of tasks and models. Notably, TTRL boosts the pass@1 performance of Qwen-2.5-Math-7B by approximately 211% on the AIME 2024 with only unlabeled test data. Furthermore, although TTRL is only supervised by the maj@n metric, TTRL has demonstrated performance to consistently surpass the upper limit of the initial model maj@n, and approach the performance of models trained directly on test data with ground-truth labels. Our experimental findings validate the general effectiveness of TTRL across various tasks and highlight TTRL's potential for broader tasks and domains. GitHub: https://github.com/PRIME-RL/TTRL
Authors: Aishik Sanyal, Samuel Schapiro, Sumuk Shashidhar, Royce Moon, Lav R. Varshney, Dilek Hakkani-Tur
Abstract: Recently, large language models (LLMs) have shown promising abilities to generate novel research ideas in science, a direction which coincides with many foundational principles in computational creativity (CC). In light of these developments, we present an idea generation system named Spark that couples retrieval-augmented idea generation using LLMs with a reviewer model named Judge trained on 600K scientific reviews from OpenReview. Our work is both a system demonstration and intended to inspire other CC researchers to explore grounding the generation and evaluation of scientific ideas within foundational CC principles. To this end, we release the annotated dataset used to train Judge, inviting other researchers to explore the use of LLMs for idea generation and creative evaluations.
Authors: Jaouad Mourtada
Abstract: We study the problem of estimating a distribution over a finite alphabet from an i.i.d. sample, with accuracy measured in relative entropy (Kullback-Leibler divergence). While optimal expected risk bounds are known, high-probability guarantees remain less well-understood. First, we analyze the classical Laplace (add-one) estimator, obtaining matching upper and lower bounds on its performance and showing its optimality among confidence-independent estimators. We then characterize the minimax-optimal high-probability risk, which is attained via a simple confidence-dependent smoothing technique. Interestingly, the optimal non-asymptotic risk exhibits an additional logarithmic factor over the ideal asymptotic risk. Next, motivated by scenarios where the alphabet exceeds the sample size, we investigate methods that adapt to the sparsity of the distribution at hand. We introduce an estimator using data-dependent smoothing, for which we establish a high-probability risk bound depending on two effective sparsity parameters. As part of the analysis, we also derive a sharp high-probability upper bound on the missing mass.
Authors: Junchi Liu, Ying Tang, Sergei Tretiak, Wenhui Duan, Liujiang Zhou
Abstract: Recent advances in machine learning have demonstrated an enormous utility of deep learning approaches, particularly Graph Neural Networks (GNNs) for materials science. These methods have emerged as powerful tools for high-throughput prediction of material properties, offering a compelling enhancement and alternative to traditional first-principles calculations. While the community has predominantly focused on developing increasingly complex and universal models to enhance predictive accuracy, such approaches often lack physical interpretability and insights into materials behavior. Here, we introduce a novel computational paradigm, Self-Adaptable Graph Attention Networks integrated with Symbolic Regression (SA-GAT-SR), that synergistically combines the predictive capability of GNNs with the interpretative power of symbolic regression. Our framework employs a self-adaptable encoding algorithm that automatically identifies and adjust attention weights so as to screen critical features from an expansive 180-dimensional feature space while maintaining O(n) computational scaling. The integrated SR module subsequently distills these features into compact analytical expressions that explicitly reveal quantum-mechanically meaningful relationships, achieving 23 times acceleration compared to conventional SR implementations that heavily rely on first principle calculations-derived features as input. This work suggests a new framework in computational materials science, bridging the gap between predictive accuracy and physical interpretability, offering valuable physical insights into material behavior.
Authors: Minzheng Wang, Yongbin Li, Haobo Wang, Xinghua Zhang, Nan Xu, Bingli Wu, Fei Huang, Haiyang Yu, Wenji Mao
Abstract: Effective social intelligence simulation requires language agents to dynamically adjust reasoning depth, a capability notably absent in current studies. Existing methods either lack this kind of reasoning capability or enforce Long Chain-of-Thought reasoning uniformly across all scenarios, resulting in excessive token usage and inflexible social simulation. To address this, we propose an $\textbf{A}$daptive $\textbf{M}$ode $\textbf{L}$earning ($\textbf{AML}$) framework in this paper, aiming to improve the adaptive thinking ability of language agents in dynamic social interactions. To this end, we first identify hierarchical thinking modes ranging from intuitive response to deep deliberation based on the cognitive control theory. We then develop the $\textbf{A}$daptive $\textbf{M}$ode $\textbf{P}$olicy $\textbf{O}$ptimization ($\textbf{AMPO}$) algorithm to optimize the context-aware mode switching and reasoning. Our framework advances existing research in three key aspects: (1) Multi-granular thinking mode design, (2) Context-aware mode switching across social interaction, and (3) Token-efficient reasoning via depth-adaptive processing. Extensive experiments on social intelligence benchmarks verify that AML achieves 15.6% higher task performance than GPT-4o. Notably, our AMPO outperforms GRPO by 7.0% with 32.8% shorter reasoning chains, demonstrating the advantage of adaptive thinking mode selection and optimization mechanism in AMPO over GRPO's fixed-depth solution.
Authors: Hafez Ghaemi, Eilif Muller, Shahab Bakhtiari
Abstract: Current self-supervised algorithms commonly rely on transformations such as data augmentation and masking to learn visual representations. This is achieved by enforcing invariance or equivariance with respect to these transformations after encoding two views of an image. This dominant two-view paradigm often limits the flexibility of learned representations for downstream adaptation by creating performance trade-offs between high-level invariance-demanding tasks such as image classification and more fine-grained equivariance-related tasks. In this work, we proposes \emph{seq-JEPA}, a world modeling framework that introduces architectural inductive biases into joint-embedding predictive architectures to resolve this trade-off. Without relying on dual equivariance predictors or loss terms, seq-JEPA simultaneously learns two architecturally segregated representations: one equivariant to specified transformations and another invariant to them. To do so, our model processes short sequences of different views (observations) of inputs. Each encoded view is concatenated with an embedding of the relative transformation (action) that produces the next observation in the sequence. These view-action pairs are passed through a transformer encoder that outputs an aggregate representation. A predictor head then conditions this aggregate representation on the upcoming action to predict the representation of the next observation. Empirically, seq-JEPA demonstrates strong performance on both equivariant and invariant benchmarks without sacrificing one for the other. Furthermore, it excels at tasks that inherently require aggregating a sequence of observations, such as path integration across actions and predictive learning across eye movements.
Authors: Haichen Hu, David Simchi-Levi, Navid Azizan
Abstract: Contextual online decision-making problems with constraints appear in a wide range of real-world applications, such as adaptive experimental design under safety constraints, personalized recommendation with resource limits, and dynamic pricing under fairness requirements. In this paper, we investigate a general formulation of sequential decision-making with stage-wise feasibility constraints, where at each round, the learner must select an action based on observed context while ensuring that a problem-specific feasibility criterion is satisfied. We propose a unified algorithmic framework that captures many existing constrained learning problems, including constrained bandits, active learning with label budgets, online hypothesis testing with Type I error control, and model calibration. Central to our approach is the concept of upper counterfactual confidence bounds, which enables the design of practically efficient online algorithms with strong theoretical guarantees using any offline conditional density estimation oracle. To handle feasibility constraints in complex environments, we introduce a generalized notion of the eluder dimension, extending it from the classical setting based on square loss to a broader class of metric-like probability divergences. This allows us to capture the complexity of various density function classes and characterize the utility regret incurred due to feasibility constraint uncertainty. Our result offers a principled foundation for constrained sequential decision-making in both theory and practice.
Authors: Rui Melo, Claudia Mamede, Andre Catarino, Rui Abreu, Henrique Lopes Cardoso
Abstract: Software vulnerabilities such as buffer overflows and SQL injections are a major source of security breaches. Traditional methods for vulnerability detection remain essential but are limited by high false positive rates, scalability issues, and reliance on manual effort. These constraints have driven interest in AI-based approaches to automated vulnerability detection and secure code generation. While Large Language Models (LLMs) have opened new avenues for classification tasks, their complexity and opacity pose challenges for interpretability and deployment. Sparse Autoencoder offer a promising solution to this problem. We explore whether SAEs can serve as a lightweight, interpretable alternative for bug detection in Java functions. We evaluate the effectiveness of SAEs when applied to representations from GPT-2 Small and Gemma 2B, examining their capacity to highlight buggy behaviour without fine-tuning the underlying LLMs. We found that SAE-derived features enable bug detection with an F1 score of up to 89%, consistently outperforming fine-tuned transformer encoder baselines. Our work provides the first empirical evidence that SAEs can be used to detect software bugs directly from the internal representations of pretrained LLMs, without any fine-tuning or task-specific supervision.
Authors: Wasu Top Piriyakulkij, Yichao Liang, Hao Tang, Adrian Weller, Marta Kryven, Kevin Ellis
Abstract: Learning how the world works is central to building AI agents that can adapt to complex environments. Traditional world models based on deep learning demand vast amounts of training data, and do not flexibly update their knowledge from sparse observations. Recent advances in program synthesis using Large Language Models (LLMs) give an alternate approach which learns world models represented as source code, supporting strong generalization from little data. To date, application of program-structured world models remains limited to natural language and grid-world domains. We introduce a novel program synthesis method for effectively modeling complex, non-gridworld domains by representing a world model as an exponentially-weighted product of programmatic experts (PoE-World) synthesized by LLMs. We show that this approach can learn complex, stochastic world models from just a few observations. We evaluate the learned world models by embedding them in a model-based planning agent, demonstrating efficient performance and generalization to unseen levels on Atari's Pong and Montezuma's Revenge. We release our code and display the learned world models and videos of the agent's gameplay at https://topwasu.github.io/poe-world.
Authors: Oskar Allerbo, Thomas B. Sch\"on
Abstract: The success of unsupervised learning raises the question of whether also supervised models can be trained without using the information in the output $y$. In this paper, we demonstrate that this is indeed possible. The key step is to formulate the model as a smoother, i.e. on the form $\hat{f}=Sy$, and to construct the smoother matrix $S$ independently of $y$, e.g. by training on random labels. We present a simple model selection criterion based on the distribution of the out-of-sample predictions and show that, in contrast to cross-validation, this criterion can be used also without access to $y$. We demonstrate on real and synthetic data that $y$-free trained versions of linear and kernel ridge regression, smoothing splines, and neural networks perform similarly to their standard, $y$-based, versions and, most importantly, significantly better than random guessing.
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 understanding the underlying rationale. To address this limitation, we propose Critique-Guided Distillation (CGD), a novel multi-stage framework that integrates teacher model generated explanatory critiques and 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 what to imitate and why. Using entropy-based analysis, we show that CGD reduces refinement uncertainty and can be interpreted as a Bayesian posterior update. We perform extensive empirical evaluation of 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: 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: 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 \textit{function vectors} (FVs), recently proposed as a mechanism to extract few-shot ICL task representations. We generalize FVs 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 promptings forms do not induce a common task representation through FVs but elicit different, partly overlapping mechanisms. Our findings offer principled support to the practice of combining 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, Deyi Liu, Yao Luo, Xingyan Bin, Hongbin Ren, Mingji Han, Wenhao Hao, Bairen Yi, LingJun Liu, Bole Ma, Xiaoying Jia, Xun Zhou, Siyuan Qiao, 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: Shirong Xu, Hengzhi He, Guang Cheng
Abstract: In recent years, model collapse has become a critical issue in language model training, making it essential to understand the underlying mechanisms driving this phenomenon. In this paper, we investigate recursive parametric model training from a probabilistic perspective, aiming to characterize the conditions under which model collapse occurs and, crucially, how it can be mitigated. We conceptualize the recursive training process as a random walk of the model estimate, highlighting how the sample size influences the step size and how the estimation procedure determines the direction and potential bias of the random walk. Under mild conditions, we rigorously show that progressively increasing the sample size at each training step is necessary to prevent model collapse. In particular, when the estimation is unbiased, the required growth rate follows a superlinear pattern. This rate needs to be accelerated even further in the presence of substantial estimation bias. Building on this probabilistic framework, we also investigate the probability that recursive training on synthetic data yields models that outperform those trained solely on real data. Moreover, we extend these results to general parametric model family in an asymptotic regime. Finally, we validate our theoretical results through extensive simulations and a real-world dataset.
Authors: Haishi Bai, Jozo Dujmovic, Jianwu Wang
Abstract: As machine learning models and autonomous agents are increasingly deployed in high-stakes, real-world domains such as healthcare, security, finance, and robotics, the need for transparent and trustworthy explanations has become critical. To ensure end-to-end transparency of AI decisions, we need models that are not only accurate but also fully explainable and human-tunable. We introduce BACON, a novel framework for automatically training explainable AI models for decision making problems using graded logic. BACON achieves high predictive accuracy while offering full structural transparency and precise, logic-based symbolic explanations, enabling effective human-AI collaboration and expert-guided refinement. We evaluate BACON with a diverse set of scenarios: classic Boolean approximation, Iris flower classification, house purchasing decisions and breast cancer diagnosis. In each case, BACON provides high-performance models while producing compact, human-verifiable decision logic. These results demonstrate BACON's potential as a practical and principled approach for delivering crisp, trustworthy explainable AI.
Authors: Yan Wang, Ling Ding, Tien N Nguyen, Shaohua Wang, Yanan Zheng
Abstract: Large Language Models for code often entail significant computational complexity, which grows significantly with the length of the input code sequence. We propose LeanCode for code simplification to reduce training and prediction time, leveraging code contexts in utilizing attention scores to represent the tokens' importance. We advocate for the selective removal of tokens based on the average context-aware attention scores rather than average scores across all inputs. LeanCode uses the attention scores of `CLS' tokens within the encoder for classification tasks, such as code search. It also employs the encoder-decoder attention scores to determine token significance for sequence-to-sequence tasks like code summarization. Our evaluation shows LeanCode's superiority over the SOTAs DietCode and Slimcode, with improvements of 60% and 16% for code search, and 29% and 27% for code summarization, respectively.
Authors: Minglu Zhao, Dehong Xu, Deqian Kong, Wen-Hao Zhang, Ying Nian Wu
Abstract: The hippocampus orchestrates spatial navigation through collective place cell encodings that form cognitive maps. We reconceptualize the population of place cells as position embeddings approximating multi-scale symmetric random walk transition kernels: the inner product $\langle h(x, t), h(y, t) \rangle = q(y|x, t)$ represents normalized transition probabilities, where $h(x, t)$ is the embedding at location $ x $, and $q(y|x, t)$ is the normalized symmetric transition probability over time $t$. The time parameter $\sqrt{t}$ defines a spatial scale hierarchy, mirroring the hippocampal dorsoventral axis. $q(y|x, t)$ defines spatial adjacency between $x$ and $y$ at scale or resolution $\sqrt{t}$, and the pairwise adjacency relationships $(q(y|x, t), \forall x, y)$ are reduced into individual embeddings $(h(x, t), \forall x)$ that collectively form a map of the environment at sale $\sqrt{t}$. Our framework employs gradient ascent on $q(y|x, t) = \langle h(x, t), h(y, t)\rangle$ with adaptive scale selection, choosing the time scale with maximal gradient at each step for trap-free, smooth trajectories. Efficient matrix squaring $P_{2t} = P_t^2$ builds global representations from local transitions $P_1$ without memorizing past trajectories, enabling hippocampal preplay-like path planning. This produces robust navigation through complex environments, aligning with hippocampal navigation. Experimental results show that our model captures place cell properties -- field size distribution, adaptability, and remapping -- while achieving computational efficiency. By modeling collective transition probabilities rather than individual place fields, we offer a biologically plausible, scalable framework for spatial navigation.
Authors: David Nordstr\"om, Johan Edstedt, Fredrik Kahl, Georg B\"okman
Abstract: Recent efforts at scaling computer vision models have established Vision Transformers (ViTs) as the leading architecture. ViTs incorporate weight sharing over image patches as an important inductive bias. In this work, we show that ViTs benefit from incorporating equivariance under the octic group, i.e., reflections and 90-degree rotations, as a further inductive bias. We develop new architectures, octic ViTs, that use octic-equivariant layers and put them to the test on both supervised and self-supervised learning. Through extensive experiments on DeiT-III and DINOv2 training on ImageNet-1K, we show that octic ViTs yield more computationally efficient networks while also improving performance. In particular, we achieve approximately 40% reduction in FLOPs for ViT-H while simultaneously improving both classification and segmentation results.
Authors: Zhiwen Chen, Bo Leng, Zhuoren Li, Hanming Deng, Guizhe Jin, Ran Yu, Huanxi Wen
Abstract: Integrating Large Language Models (LLMs) with Reinforcement Learning (RL) can enhance autonomous driving (AD) performance in complex scenarios. However, current LLM-Dominated RL methods over-rely on LLM outputs, which are prone to hallucinations. Evaluations show that state-of-the-art LLM indicates a non-hallucination rate of only approximately 57.95% when assessed on essential driving-related tasks. Thus, in these methods, hallucinations from the LLM can directly jeopardize the performance of driving policies. This paper argues that maintaining relative independence between the LLM and the RL is vital for solving the hallucinations problem. Consequently, this paper is devoted to propose a novel LLM-Hinted RL paradigm. The LLM is used to generate semantic hints for state augmentation and policy optimization to assist RL agent in motion planning, while the RL agent counteracts potential erroneous semantic indications through policy learning to achieve excellent driving performance. Based on this paradigm, we propose the HCRMP (LLM-Hinted Contextual Reinforcement Learning Motion Planner) architecture, which is designed that includes Augmented Semantic Representation Module to extend state space. Contextual Stability Anchor Module enhances the reliability of multi-critic weight hints by utilizing information from the knowledge base. Semantic Cache Module is employed to seamlessly integrate LLM low-frequency guidance with RL high-frequency control. Extensive experiments in CARLA validate HCRMP's strong overall driving performance. HCRMP achieves a task success rate of up to 80.3% under diverse driving conditions with different traffic densities. Under safety-critical driving conditions, HCRMP significantly reduces the collision rate by 11.4%, which effectively improves the driving performance in complex scenarios.