Authors: Amir Noorizadegan, Sifan Wang, Leevan Ling
Abstract: Kolmogorov-Arnold Networks (KANs) have recently emerged as a promising alternative to traditional Multilayer Perceptrons (MLPs), inspired by the Kolmogorov-Arnold representation theorem. Unlike MLPs, which use fixed activation functions on nodes, KANs employ learnable univariate basis functions on edges, offering enhanced expressivity and interpretability. This review provides a systematic and comprehensive overview of the rapidly expanding KAN landscape, moving beyond simple performance comparisons to offer a structured synthesis of theoretical foundations, architectural variants, and practical implementation strategies. By collecting and categorizing a vast array of open-source implementations, we map the vibrant ecosystem supporting KAN development. We begin by bridging the conceptual gap between KANs and MLPs, establishing their formal equivalence and highlighting the superior parameter efficiency of the KAN formulation. A central theme of our review is the critical role of the basis function; we survey a wide array of choices, including B-splines, Chebyshev and Jacobi polynomials, ReLU compositions, Gaussian RBFs, and Fourier series, and analyze their respective trade-offs in terms of smoothness, locality, and computational cost. We then categorize recent advancements into a clear roadmap, covering techniques for improving accuracy, efficiency, and regularization. Key topics include physics-informed loss design, adaptive sampling, domain decomposition, hybrid architectures, and specialized methods for handling discontinuities. Finally, we provide a practical "Choose-Your-KAN" guide to help practitioners select appropriate architectures, and we conclude by identifying current research gaps. The associated GitHub repository https://github.com/AmirNoori68/kan-review complements this paper and serves as a structured reference for ongoing KAN research.
Authors: Simon A. Lee, Cyrus Tanade, Hao Zhou, Juhyeon Lee, Megha Thukral, Minji Han, Rachel Choi, Md Sazzad Hissain Khan, Baiying Lu, Migyeong Gwak, Mehrab Bin Morshed, Viswam Nathan, Md Mahbubur Rahman, Li Zhu, Subramaniam Venkatraman, Sharanya Arcot Desai
Abstract: Wearable sensors provide abundant physiological time series, yet the principles governing their predictive utility remain unclear. We hypothesize that temporal resolution is a fundamental axis of representation learning, with different clinical and behavioral outcomes relying on structure at distinct scales. To test this resolution hypothesis, we introduce HiMAE (Hierarchical Masked Autoencoder), a self supervised framework that combines masked autoencoding with a hierarchical convolutional encoder decoder. HiMAE produces multi resolution embeddings that enable systematic evaluation of which temporal scales carry predictive signal, transforming resolution from a hyperparameter into a probe for interpretability. Across classification, regression, and generative benchmarks, HiMAE consistently outperforms state of the art foundation models that collapse scale, while being orders of magnitude smaller. HiMAE is an efficient representation learner compact enough to run entirely on watch, achieving sub millisecond inference on smartwatch class CPUs for true edge inference. Together, these contributions position HiMAE as both an efficient self supervised learning method and a discovery tool for scale sensitive structure in wearable health.
Authors: Siddharth Verma, Alankar Alankar
Abstract: High-energy materials (HEMs) are critical for propulsion and defense domains, yet their discovery remains constrained by experimental data and restricted access to testing facilities. This work presents a novel approach toward high-energy molecules by combining Long Short-Term Memory (LSTM) networks for molecular generation and Attentive Graph Neural Networks (GNN) for property predictions. We propose a transformative embedding space construction strategy that integrates fixed SHA-256 embeddings with partially trainable representations. Unlike conventional regularization techniques, this changes the representational basis itself, reshaping the molecular input space before learning begins. Without recourse to pretraining, the generator achieves 67.5% validity and 37.5% novelty. The generated library exhibits a mean Tanimoto coefficient of 0.214 relative to training set signifying the ability of framework to generate a diverse chemical space. We identified 37 new super explosives higher than 9 km/s predicted detonation velocity.
Authors: Zihan Pengmei, Costas Mavromatis, Zhengyuan Shen, Yunyi Zhang, Vassilis N. Ioannidis, Huzefa Rangwala
Abstract: Chain-of-thought (CoT) supervision can substantially improve transformer performance, yet the mechanisms by which models learn to follow and benefit from CoT remain poorly understood. We investigate these learning dynamics through the lens of grokking by pretraining transformers on symbolic reasoning tasks with tunable algorithmic complexity and controllable data composition to study their generalization. Models were trained under two settings: (i) producing only final answers, and (ii) emitting explicit CoT traces before answering. Our results show that while CoT generally improves task performance, its benefits depend on task complexity. To quantify these effects, we model the accuracy of the logarithmic training steps with a three-parameter logistic curve, revealing how the learning speed and shape vary with task complexity, data distribution, and the presence of CoT supervision. We also uncover a transient trace unfaithfulness phase: early in training, models often produce correct answers while skipping or contradicting CoT steps, before later aligning their reasoning traces with answers. Empirically, we (1) demonstrate that CoT accelerates generalization but does not overcome tasks with higher algorithmic complexity, such as finding list intersections; (2) introduce a kinetic modeling framework for understanding transformer learning; (3) characterize trace faithfulness as a dynamic property that emerges over training; and (4) show CoT alters internal transformer computation mechanistically.
Authors: Siavash M. Alamouti, Fay Arjomandi
Abstract: Modern multi-agent systems ranging from sensor networks monitoring critical infrastructure to crowdsourcing platforms aggregating human intelligence can suffer significant performance degradation due to systematic biases that vary with environmental conditions. Current approaches either ignore these biases, leading to suboptimal decisions, or require expensive calibration procedures that are often infeasible in practice. This performance gap has real consequences: inaccurate environmental monitoring, unreliable financial predictions, and flawed aggregation of human judgments. This paper addresses the fundamental question: when can we learn and correct for these unknown biases to recover near-optimal performance, and when is such learning futile? We develop a theoretical framework that decomposes biases into learnable systematic components and irreducible stochastic components, introducing the concept of learnability ratio as the fraction of bias variance predictable from observable covariates. This ratio determines whether bias learning is worthwhile for a given system. We prove that the achievable performance improvement is fundamentally bounded by this learnability ratio, providing system designers with quantitative guidance on when to invest in bias learning versus simpler approaches. We present the Adaptive Bias Learning and Optimal Combining (ABLOC) algorithm, which iteratively learns bias-correcting transformations while optimizing combination weights through closedform solutions, guaranteeing convergence to these theoretical bounds. Experimental validation demonstrates that systems with high learnability ratios can recover significant performance (we achieved 40%-70% of theoretical maximum improvement in our examples), while those with low learnability show minimal benefit, validating our diagnostic criteria for practical deployment decisions.
Authors: Farnoosh Namdarpour, Joseph Y. J. Chow
Abstract: Ride-pooling, also known as ride-sharing, shared ride-hailing, or microtransit, is a service wherein passengers share rides. This service can reduce costs for both passengers and operators and reduce congestion and environmental impacts. A key limitation, however, is its myopic decision-making, which overlooks long-term effects of dispatch decisions. To address this, we propose a simulation-informed reinforcement learning (RL) approach. While RL has been widely studied in the context of ride-hailing systems, its application in ride-pooling systems has been less explored. In this study, we extend the learning and planning framework of Xu et al. (2018) from ride-hailing to ride-pooling by embedding a ride-pooling simulation within the learning mechanism to enable non-myopic decision-making. In addition, we propose a complementary policy for rebalancing idle vehicles. By employing n-step temporal difference learning on simulated experiences, we derive spatiotemporal state values and subsequently evaluate the effectiveness of the non-myopic policy using NYC taxi request data. Results demonstrate that the non-myopic policy for matching can increase the service rate by up to 8.4% versus a myopic policy while reducing both in-vehicle and wait times for passengers. Furthermore, the proposed non-myopic policy can decrease fleet size by over 25% compared to a myopic policy, while maintaining the same level of performance, thereby offering significant cost savings for operators. Incorporating rebalancing operations into the proposed framework cuts wait time by up to 27.3%, in-vehicle time by 12.5%, and raises service rate by 15.1% compared to using the framework for matching decisions alone at the cost of increased vehicle minutes traveled per passenger.
Authors: Jin Seong, Jiyun Park, Wencke Liermann, Hongseok Choi, Yoonji Nam, Hyun Kim, Soojong Lim, Namhoon Lee
Abstract: The dynamic nature of information necessitates continuously updating large vision-language models (LVLMs). While recent knowledge editing techniques hint at promising directions, they often focus on editing a single modality (vision or language) in isolation. This prevalent practice neglects the inherent multimodality of LVLMs and the continuous nature of knowledge updates, potentially leading to suboptimal editing outcomes when considering the interplay between modalities and the need for ongoing knowledge refinement. To address these limitations, we propose MemEIC, a novel method for Continual and Compositional Knowledge Editing (CCKE) in LVLMs. MemEIC enables compositional editing of both visual and textual knowledge sequentially. Our approach employs a hybrid external-internal editor featuring a dual external memory for cross-modal evidence retrieval and dual LoRA adapters that facilitate disentangled parameter updates for each modality. A key component is a brain-inspired knowledge connector, activated selectively for compositional reasoning, that integrates information across different modalities. Experiments demonstrate that MemEIC significantly improves performance on complex multimodal questions and effectively preserves prior edits, setting a new benchmark for CCKE in LVLMs.
Authors: Jialong Sun, Xinpeng Ling, Jiaxuan Zou, Jiawen Kang, Kejia Zhang
Abstract: The inherent autocorrelation of time series data presents an ongoing challenge to multivariate time series prediction. Recently, a widely adopted approach has been the incorporation of frequency domain information to assist in long-term prediction tasks. Many researchers have independently observed the spectral bias phenomenon in neural networks, where models tend to fit low-frequency signals before high-frequency ones. However, these observations have often been attributed to the specific architectures designed by the researchers, rather than recognizing the phenomenon as a universal characteristic across models. To unify the understanding of the spectral bias phenomenon in long-term time series prediction, we conducted extensive empirical experiments to measure spectral bias in existing mainstream models. Our findings reveal that virtually all models exhibit this phenomenon. To mitigate the impact of spectral bias, we propose the FreLE (Frequency Loss Enhancement) algorithm, which enhances model generalization through both explicit and implicit frequency regularization. This is a plug-and-play model loss function unit. A large number of experiments have proven the superior performance of FreLE. Code is available at https://github.com/Chenxing-Xuan/FreLE.
Authors: Kun Chen, Peng Shi, Haibo Qiu, Zhixiong Zeng, Siqi Yang, Wenji Mao, Lin Ma
Abstract: Reinforcement learning (RL) with verifiable rewards has recently catalyzed a wave of "MLLM-r1" approaches that bring RL to vision language models. Most representative paradigms begin with a cold start, typically employing supervised fine-tuning (SFT), to initialize the policy before RL. However, SFT-based cold start adopts the reasoning paradigm intertwined with task solution and output format, which may induce instruction-style overfitting, weakens out-of-distribution generalization, and ultimately affects downstream RL. We revisit the cold start along two views, its training method and data construction, and introduce the Generalization Factor (GF) coefficient to quantify the generalization capability under different methods. Our empirical study finds that preference-based training methods (e.g. DPO) generalizes better than SFT-based methods in cold start. Motivated by this, we propose SPECS-a Self-distilled, Preference-based Cold Start framework that decouples multimodal learning: (1) generates introspective preference data pairs via self-distillation, avoiding reliance on larger teachers or manual annotation; (2) performs preference-based training to learn, focusing on shallow, transferable surface-form criteria (format, structure, style) rather than memorizing content; and (3) hands off to RL with verifiable rewards for deep reasoning results. Experimental results across multiple multimodal benchmarks show that our decoupling learning framework yields consistent performance gains over strong baselines, improving MEGA-Bench by 4.1% and MathVista by 12.2%. Additional experiments indicate that SPECS contributes to reducing in-distribution "stuckness," improving exploration, stabilizing training, and raising the performance ceiling.
Authors: Hong Wang, Haiyang Xin, Jie Wang, Xuanze Yang, Fei Zha, Huanshuo Dong, Yan Jiang
Abstract: Pre-training has proven effective in addressing data scarcity and performance limitations in solving PDE problems with neural operators. However, challenges remain due to the heterogeneity of PDE datasets in equation types, which leads to high errors in mixed training. Additionally, dense pre-training models that scale parameters by increasing network width or depth incur significant inference costs. To tackle these challenges, we propose a novel Mixture-of-Experts Pre-training Operator Transformer (MoE-POT), a sparse-activated architecture that scales parameters efficiently while controlling inference costs. Specifically, our model adopts a layer-wise router-gating network to dynamically select 4 routed experts from 16 expert networks during inference, enabling the model to focus on equation-specific features. Meanwhile, we also integrate 2 shared experts, aiming to capture common properties of PDE and reduce redundancy among routed experts. The final output is computed as the weighted average of the results from all activated experts. We pre-train models with parameters from 30M to 0.5B on 6 public PDE datasets. Our model with 90M activated parameters achieves up to a 40% reduction in zero-shot error compared with existing models with 120M activated parameters. Additionally, we conduct interpretability analysis, showing that dataset types can be inferred from router-gating network decisions, which validates the rationality and effectiveness of the MoE architecture.
Authors: Jaewon Chu, Seunghun Lee, Hyunwoo J. Kim
Abstract: Large language models (LLMs) have achieved remarkable success across diverse domains, due to their strong instruction-following capabilities. This has led to increasing interest in optimizing instructions for black-box LLMs, whose internal parameters are inaccessible but widely used due to their strong performance. To optimize instructions for black-box LLMs, recent methods employ white-box LLMs to generate candidate instructions from optimized soft prompts. However, white-box LLMs often map different soft prompts to the same instruction, leading to redundant queries. While previous studies regarded this many-to-one mapping as a structure that hinders optimization efficiency, we reinterpret it as a useful prior knowledge that can accelerate the optimization. To this end, we introduce PREimage-informed inSTruction Optimization (PRESTO), a novel framework that leverages the preimage structure of soft prompts for efficient optimization. PRESTO consists of three key components: (1) score sharing, which shares the evaluation score with all soft prompts in a preimage; (2) preimage-based initialization, which selects initial data points that maximize search space coverage using preimage information; and (3) score consistency regularization, which enforces prediction consistency within each preimage. By leveraging preimages, PRESTO achieves the effect of effectively obtaining 14 times more scored data under the same query budget, resulting in more efficient optimization. Experimental results on 33 instruction optimization tasks demonstrate the superior performance of PRESTO. Code is available at https://github.com/mlvlab/PRESTO
Authors: Sungho Koh, SeungJu Cha, Hyunwoo Oh, Kwanyoung Lee, Dong-Jin Kim
Abstract: Text-to-image diffusion models often exhibit degraded performance when generating images beyond their training resolution. Recent training-free methods can mitigate this limitation, but they often require substantial computation or are incompatible with recent Diffusion Transformer models. In this paper, we propose ScaleDiff, a model-agnostic and highly efficient framework for extending the resolution of pretrained diffusion models without any additional training. A core component of our framework is Neighborhood Patch Attention (NPA), an efficient mechanism that reduces computational redundancy in the self-attention layer with non-overlapping patches. We integrate NPA into an SDEdit pipeline and introduce Latent Frequency Mixing (LFM) to better generate fine details. Furthermore, we apply Structure Guidance to enhance global structure during the denoising process. Experimental results demonstrate that ScaleDiff achieves state-of-the-art performance among training-free methods in terms of both image quality and inference speed on both U-Net and Diffusion Transformer architectures.
Authors: Xiaoke Huang, Ningsen Wang, Hui Liu, Xianfeng Tang, Yuyin Zhou
Abstract: Large Multimodal Models (LMMs) are increasingly capable of answering medical questions that require joint reasoning over images and text, yet training general medical VQA systems is impeded by the lack of large, openly usable, high-quality corpora. We present MedVLSynther, a rubric-guided generator-verifier framework that synthesizes high-quality multiple-choice VQA items directly from open biomedical literature by conditioning on figures, captions, and in-text references. The generator produces self-contained stems and parallel, mutually exclusive options under a machine-checkable JSON schema; a multi-stage verifier enforces essential gates (self-containment, single correct answer, clinical validity, image-text consistency), awards fine-grained positive points, and penalizes common failure modes before acceptance. Applying this pipeline to PubMed Central yields MedSynVQA: 13,087 audited questions over 14,803 images spanning 13 imaging modalities and 28 anatomical regions. Training open-weight LMMs with reinforcement learning using verifiable rewards improves accuracy across six medical VQA benchmarks, achieving averages of 55.85 (3B) and 58.15 (7B), with up to 77.57 on VQA-RAD and 67.76 on PathVQA, outperforming strong medical LMMs. A Ablations verify that both generation and verification are necessary and that more verified data consistently helps, and a targeted contamination analysis detects no leakage from evaluation suites. By operating entirely on open literature and open-weight models, MedVLSynther offers an auditable, reproducible, and privacy-preserving path to scalable medical VQA training data.
Authors: Kang Chen, Zhihao Liu, Tonghe Zhang, Zhen Guo, Si Xu, Hao Lin, Hongzhi Zang, Quanlu Zhang, Zhaofei Yu, Guoliang Fan, Tiejun Huang, Yu Wang, Chao Yu
Abstract: Vision-Language-Action (VLA) models enable robots to understand and perform complex tasks from multimodal input. Although recent work explores using reinforcement learning (RL) to automate the laborious data collection process in scaling supervised fine-tuning (SFT), applying large-scale RL to flow-based VLAs (e.g., $\pi_0$, $\pi_{0.5}$) remains challenging due to intractable action log-likelihoods from iterative denoising. We address this challenge with $\pi_{\text{RL}}$, an open-source framework for training flow-based VLAs in parallel simulation. $\pi_{\text{RL}}$ implements two RL algorithms: (1) {Flow-Noise} models the denoising process as a discrete-time MDP with a learnable noise network for exact log-likelihood computation. (2) {Flow-SDE} integrates denoising with agent-environment interaction, formulating a two-layer MDP that employs ODE-to-SDE conversion for efficient RL exploration. We evaluate $\pi_{\text{RL}}$ on LIBERO and ManiSkill benchmarks. On LIBERO, $\pi_{\text{RL}}$ boosts few-shot SFT models $\pi_0$ and $\pi_{0.5}$ from 57.6% to 97.6% and from 77.1% to 98.3%, respectively. In ManiSkill, we train $\pi_{\text{RL}}$ in 320 parallel environments, improving $\pi_0$ from 41.6% to 85.7% and $\pi_{0.5}$ from 40.0% to 84.8% across 4352 pick-and-place tasks, demonstrating scalable multitask RL under heterogeneous simulation. Overall, $\pi_{\text{RL}}$ achieves significant performance gains and stronger generalization over SFT-models, validating the effectiveness of online RL for flow-based VLAs.
Authors: Harris Hardiman-Mostow, Jack Mauro, Adrien Weihs, Andrea L. Bertozzi
Abstract: We propose a graph-topological approach to active learning that directly targets the core challenge of exploration versus exploitation under scarce label budgets. To guide exploration, we introduce a coreset construction algorithm based on Balanced Forman Curvature (BFC), which selects representative initial labels that reflect the graph's cluster structure. This method includes a data-driven stopping criterion that signals when the graph has been sufficiently explored. We further use BFC to dynamically trigger the shift from exploration to exploitation within active learning routines, replacing hand-tuned heuristics. To improve exploitation, we introduce a localized graph rewiring strategy that efficiently incorporates multiscale information around labeled nodes, enhancing label propagation while preserving sparsity. Experiments on benchmark classification tasks show that our methods consistently outperform existing graph-based semi-supervised baselines at low label rates.
Authors: Manuel Iglesias-Alonso, Felix Schur, Julius von K\"ugelgen, Jonas Peters
Abstract: We consider the problem of estimating a causal effect in a multi-domain setting. The causal effect of interest is confounded by an unobserved confounder and can change between the different domains. We assume that we have access to a proxy of the hidden confounder and that all variables are discrete or categorical. We propose methodology to estimate the causal effect in the target domain, where we assume to observe only the proxy variable. Under these conditions, we prove identifiability (even when treatment and response variables are continuous). We introduce two estimation techniques, prove consistency, and derive confidence intervals. The theoretical results are supported by simulation studies and a real-world example studying the causal effect of website rankings on consumer choices.
Authors: Kianoosh Ashouritaklimi, Tom Rainforth
Abstract: Active learning has the potential to be especially useful for messy, uncurated pools where datapoints vary in relevance to the target task. However, state-of-the-art approaches to this problem currently rely on using fixed, unsupervised representations of the pool, focusing on modifying the acquisition function instead. We show that this model setup can undermine their effectiveness at dealing with messy pools, as such representations can fail to capture important information relevant to the task. To address this, we propose using task-driven representations that are periodically updated during the active learning process using the previously collected labels. We introduce two specific strategies for learning these representations, one based on directly learning semi-supervised representations and the other based on supervised fine-tuning of an initial unsupervised representation. We find that both significantly improve empirical performance over using unsupervised or pretrained representations.
Authors: Jipeng Li, Yannning Shen
Abstract: Graph Neural Networks (GNNs) are valuable intellectual property, yet many watermarks rely on backdoor triggers that break under common model edits and create ownership ambiguity. We present InvGNN-WM, which ties ownership to a model's implicit perception of a graph invariant, enabling trigger-free, black-box verification with negligible task impact. A lightweight head predicts normalized algebraic connectivity on an owner-private carrier set; a sign-sensitive decoder outputs bits, and a calibrated threshold controls the false-positive rate. Across diverse node and graph classification datasets and backbones, InvGNN-WM matches clean accuracy while yielding higher watermark accuracy than trigger- and compression-based baselines. It remains strong under unstructured pruning, fine-tuning, and post-training quantization; plain knowledge distillation (KD) weakens the mark, while KD with a watermark loss (KD+WM) restores it. We provide guarantees for imperceptibility and robustness, and we prove that exact removal is NP-complete.
Authors: Tcharlies Schmitz
Abstract: This paper introduces Modular Linear Tokenization (MLT), a reversible and deterministic technique for encoding high-cardinality categorical identifiers into compact numerical vectors. Unlike traditional hashing or one-hot encodings, MLT preserves bijective mappings by leveraging modular arithmetic over finite fields and invertible linear transformations. The method offers explicit control of dimensionality and computational scalability while maintaining full reversibility, even for millions of identifiers. Experimental results on the MovieLens 20M dataset show that MLT achieves comparable predictive performance to supervised embeddings while requiring significantly fewer parameters and lower training cost. An open-source implementation of MLT is available on PyPI (https://pypi.org/project/light-mlt/) and GitHub (https://github.com/tcharliesschmitz/light-mlt).
URLs: https://pypi.org/project/light-mlt/), https://github.com/tcharliesschmitz/light-mlt).
Authors: Lynn Metz, Rachel Haggard, Michael Moszczynski, Samer Asbah, Chris Mwase, Patricia Khomani, Tyler Smith, Hannah Cooper, Annie Mwale, Arbaaz Muslim, Gautam Prasad, Mimi Sun, Tomer Shekel, Joydeep Paul, Anna Carter, Shravya Shetty, Dylan Green
Abstract: The reliability of routine health data in low and middle-income countries (LMICs) is often constrained by reporting delays and incomplete coverage, necessitating the exploration of novel data sources and analytics. Geospatial Foundation Models (GeoFMs) offer a promising avenue by synthesizing diverse spatial, temporal, and behavioral data into mathematical embeddings that can be efficiently used for downstream prediction tasks. This study evaluated the predictive performance of three GeoFM embedding sources - Google Population Dynamics Foundation Model (PDFM), Google AlphaEarth (derived from satellite imagery), and mobile phone call detail records (CDR) - for modeling 15 routine health programmatic outputs in Malawi, and compared their utility to traditional geospatial interpolation methods. We used XGBoost models on data from 552 health catchment areas (January 2021-May 2023), assessing performance with R2, and using an 80/20 training and test data split with 5-fold cross-validation used in training. While predictive performance was mixed, the embedding-based approaches improved upon baseline geostatistical methods in 13 of 15 (87%) indicators tested. A Multi-GeoFM model integrating all three embedding sources produced the most robust predictions, achieving average 5-fold cross validated R2 values for indicators like population density (0.63), new HIV cases (0.57), and child vaccinations (0.47) and test set R2 of 0.64, 0.68, and 0.55, respectively. Prediction was poor for prediction targets with low primary data availability, such as TB and malnutrition cases. These results demonstrate that GeoFM embeddings imbue a modest predictive improvement for select health and demographic outcomes in an LMIC context. We conclude that the integration of multiple GeoFM sources is an efficient and valuable tool for supplementing and strengthening constrained routine health information systems.
Authors: Alvaro Velasquez, Susmit Jha, Ismail R. Alkhouri
Abstract: This paper surveys studies on the use of neural networks for optimization in the training-data-free setting. Specifically, we examine the dataless application of neural network architectures in optimization by re-parameterizing problems using fully connected (or MLP), convolutional, graph, and quadratic neural networks. Although MLPs have been used to solve linear programs a few decades ago, this approach has recently gained increasing attention due to its promising results across diverse applications, including those based on combinatorial optimization, inverse problems, and partial differential equations. The motivation for this setting stems from two key (possibly over-lapping) factors: (i) data-driven learning approaches are still underdeveloped and have yet to demonstrate strong results, as seen in combinatorial optimization, and (ii) the availability of training data is inherently limited, such as in medical image reconstruction and other scientific applications. In this paper, we define the dataless setting and categorize it into two variants based on how a problem instance -- defined by a single datum -- is encoded onto the neural network: (i) architecture-agnostic methods and (ii) architecture-specific methods. Additionally, we discuss similarities and clarify distinctions between the dataless neural network (dNN) settings and related concepts such as zero-shot learning, one-shot learning, lifting in optimization, and over-parameterization.
Authors: J. Jon Ryu, Pavan Yeddanapudi, Xiangxiang Xu, Gregory W. Wornell
Abstract: The InfoNCE objective, originally introduced for contrastive representation learning, has become a popular choice for mutual information (MI) estimation, despite its indirect connection to MI. In this paper, we demonstrate why InfoNCE should not be regarded as a valid MI estimator, and we introduce a simple modification, which we refer to as InfoNCE-anchor, for accurate MI estimation. Our modification introduces an auxiliary anchor class, enabling consistent density ratio estimation and yielding a plug-in MI estimator with significantly reduced bias. Beyond this, we generalize our framework using proper scoring rules, which recover InfoNCE-anchor as a special case when the log score is employed. This formulation unifies a broad spectrum of contrastive objectives, including NCE, InfoNCE, and $f$-divergence variants, under a single principled framework. Empirically, we find that InfoNCE-anchor with the log score achieves the most accurate MI estimates; however, in self-supervised representation learning experiments, we find that the anchor does not improve the downstream task performance. These findings corroborate that contrastive representation learning benefits not from accurate MI estimation per se, but from the learning of structured density ratios.
Authors: Andrew W. Rosemberg, Joaquim Dias Garcia, Fran\c{c}ois Pacaud, Robert B. Parker, Beno\^it Legat, Kaarthik Sundar, Russell Bent, Pascal Van Hentenryck
Abstract: Differentiating through constrained optimization problems is increasingly central to learning, control, and large-scale decision-making systems, yet practical integration remains challenging due to solver specialization and interface mismatches. This paper presents a general and streamlined framework-an updated DiffOpt.jl-that unifies modeling and differentiation within the Julia optimization stack. The framework computes forward - and reverse-mode solution and objective sensitivities for smooth, potentially nonconvex programs by differentiating the KKT system under standard regularity assumptions. A first-class, JuMP-native parameter-centric API allows users to declare named parameters and obtain derivatives directly with respect to them - even when a parameter appears in multiple constraints and objectives - eliminating brittle bookkeeping from coefficient-level interfaces. We illustrate these capabilities on convex and nonconvex models, including economic dispatch, mean-variance portfolio selection with conic risk constraints, and nonlinear robot inverse kinematics. Two companion studies further demonstrate impact at scale: gradient-based iterative methods for strategic bidding in energy markets and Sobolev-style training of end-to-end optimization proxies using solver-accurate sensitivities. Together, these results demonstrate that differentiable optimization can be deployed as a routine tool for experimentation, learning, calibration, and design-without deviating from standard JuMP modeling practices and while retaining access to a broad ecosystem of solvers.
Authors: Darius Masoum Zadeh-Jousdani, Elvin Hajizada, Eyke H\"ullermeier
Abstract: Robotic systems operating at the edge require efficient online learning algorithms that can continuously adapt to changing environments while processing streaming sensory data. Traditional backpropagation, while effective, conflicts with biological plausibility principles and may be suboptimal for continuous adaptation scenarios. The Predictive Coding (PC) framework offers a biologically plausible alternative with local, Hebbian-like update rules, making it suitable for neuromorphic hardware implementation. However, PC's main limitation is its computational overhead due to multiple inference iterations during training. We present Predictive Coding Network with Temporal Amortization (PCN-TA), which preserves latent states across temporal frames. By leveraging temporal correlations, PCN-TA significantly reduces computational demands while maintaining learning performance. Our experiments on the COIL-20 robotic perception dataset demonstrate that PCN-TA achieves 10% fewer weight updates compared to backpropagation and requires 50% fewer inference steps than baseline PC networks. These efficiency gains directly translate to reduced computational overhead for moving another step toward edge deployment and real-time adaptation support in resource-constrained robotic systems. The biologically-inspired nature of our approach also makes it a promising candidate for future neuromorphic hardware implementations, enabling efficient online learning at the edge.
Authors: Harin Lee, Min-hwan Oh
Abstract: We study the problem of infrequent exploration in linear bandits, addressing a significant yet overlooked gap between fully adaptive exploratory methods (e.g., UCB and Thompson Sampling), which explore potentially at every time step, and purely greedy approaches, which require stringent diversity assumptions to succeed. Continuous exploration can be impractical or unethical in safety-critical or costly domains, while purely greedy strategies typically fail without adequate contextual diversity. To bridge these extremes, we introduce a simple and practical framework, INFEX, explicitly designed for infrequent exploration. INFEX executes a base exploratory policy according to a given schedule while predominantly choosing greedy actions in between. Despite its simplicity, our theoretical analysis demonstrates that INFEX achieves instance-dependent regret matching standard provably efficient algorithms, provided the exploration frequency exceeds a logarithmic threshold. Additionally, INFEX is a general, modular framework that allows seamless integration of any fully adaptive exploration method, enabling wide applicability and ease of adoption. By restricting intensive exploratory computations to infrequent intervals, our approach can also enhance computational efficiency. Empirical evaluations confirm our theoretical findings, showing state-of-the-art regret performance and runtime improvements over existing methods.
Authors: Hyeonjun Lee, Hyungseob Shin, Gunhee Nam, Hyeonsoo Lee
Abstract: Survival analysis is a task to model the time until an event of interest occurs, widely used in clinical and biomedical research. A key challenge is to model patient heterogeneity while also adapting risk predictions to both individual characteristics and temporal dynamics. We propose a dual mixture-of-experts (MoE) framework for discrete-time survival analysis. Our approach combines a feature-encoder MoE for subgroup-aware representation learning with a hazard MoE that leverages patient features and time embeddings to capture temporal dynamics. This dual-MoE design flexibly integrates with existing deep learning based survival pipelines. On METABRIC and GBSG breast cancer datasets, our method consistently improves performance, boosting the time-dependent C-index up to 0.04 on the test sets, and yields further gains when incorporated into the Consurv framework.
Authors: Semyon Lomaso, Judah Goldfeder, Mehmet Hamza Erol, Matthew So, Yao Yan, Addison Howard, Nathan Kutz, Ravid Shwartz Ziv
Abstract: Do AI systems truly understand human concepts or merely mimic surface patterns? We investigate this through chess, where human creativity meets precise strategic concepts. Analyzing a 270M-parameter transformer that achieves grandmaster-level play, we uncover a striking paradox: while early layers encode human concepts like center control and knight outposts with up to 85\% accuracy, deeper layers, despite driving superior performance, drift toward alien representations, dropping to 50-65\% accuracy. To test conceptual robustness beyond memorization, we introduce the first Chess960 dataset: 240 expert-annotated positions across 6 strategic concepts. When opening theory is eliminated through randomized starting positions, concept recognition drops 10-20\% across all methods, revealing the model's reliance on memorized patterns rather than abstract understanding. Our layer-wise analysis exposes a fundamental tension in current architectures: the representations that win games diverge from those that align with human thinking. These findings suggest that as AI systems optimize for performance, they develop increasingly alien intelligence, a critical challenge for creative AI applications requiring genuine human-AI collaboration. Dataset and code are available at: https://github.com/slomasov/ChessConceptsLLM.
Authors: Jiali Cheng, Chirag Agarwal, Hadi Amiri
Abstract: Knowledge distillation (KD) is an effective method for model compression and transferring knowledge between models. However, its effect on model's robustness against spurious correlations that degrade performance on out-of-distribution data remains underexplored. This study investigates the effect of knowledge distillation on the transferability of ``debiasing'' capabilities from teacher models to student models on natural language inference (NLI) and image classification tasks. Through extensive experiments, we illustrate several key findings: (i) overall the debiasing capability of a model is undermined post-KD; (ii) training a debiased model does not benefit from injecting teacher knowledge; (iii) although the overall robustness of a model may remain stable post-distillation, significant variations can occur across different types of biases; and (iv) we pin-point the internal attention pattern and circuit that causes the distinct behavior post-KD. Given the above findings, we propose three effective solutions to improve the distillability of debiasing methods: developing high quality data for augmentation, implementing iterative knowledge distillation, and initializing student models with weights obtained from teacher models. To the best of our knowledge, this is the first study on the effect of KD on debiasing and its interenal mechanism at scale. Our findings provide understandings on how KD works and how to design better debiasing methods.
Authors: David Otte, J\"org K. H. Franke, Frank Hutter
Abstract: Symbolic regression (SR) aims to discover the underlying mathematical expressions that explain observed data. This holds promise for both gaining scientific insight and for producing inherently interpretable and generalizable models for tabular data. In this work we focus on the basics of SR. Deep learning-based SR has recently become competitive with genetic programming approaches, but the role of scale has remained largely unexplored. Inspired by scaling laws in language modeling, we present the first systematic investigation of scaling in SR, using a scalable end-to-end transformer pipeline and carefully generated training data. Across five different model sizes and spanning three orders of magnitude in compute, we find that both validation loss and solved rate follow clear power-law trends with compute. We further identify compute-optimal hyperparameter scaling: optimal batch size and learning rate grow with model size, and a token-to-parameter ratio of $\approx$15 is optimal in our regime, with a slight upward trend as compute increases. These results demonstrate that SR performance is largely predictable from compute and offer important insights for training the next generation of SR models.
Authors: Di Zhang
Abstract: This paper proposes a novel paradigm for machine learning that moves beyond traditional parameter optimization. Unlike conventional approaches that search for optimal parameters within a fixed geometric space, our core idea is to treat the model itself as a malleable geometric entity. Specifically, we optimize the metric tensor field on a manifold with a predefined topology, thereby dynamically shaping the geometric structure of the model space. To achieve this, we construct a variational framework whose loss function carefully balances data fidelity against the intrinsic geometric complexity of the manifold. The former ensures the model effectively explains observed data, while the latter acts as a regularizer, penalizing overly curved or irregular geometries to encourage simpler models and prevent overfitting. To address the computational challenges of this infinite-dimensional optimization problem, we introduce a practical method based on discrete differential geometry: the continuous manifold is discretized into a triangular mesh, and the metric tensor is parameterized by edge lengths, enabling efficient optimization using automatic differentiation tools. Theoretical analysis reveals a profound analogy between our framework and the Einstein-Hilbert action in general relativity, providing an elegant physical interpretation for the concept of "data-driven geometry". We further argue that even with fixed topology, metric optimization offers significantly greater expressive power than models with fixed geometry. This work lays a solid foundation for constructing fully dynamic "meta-learners" capable of autonomously evolving their geometry and topology, and it points to broad application prospects in areas such as scientific model discovery and robust representation learning.
Authors: James Meldrum, Basem Suleiman, Fethi Rabhi, Muhammad Johan Alibasa
Abstract: Synthetic data generation has emerged as a promising approach to address the challenges of using sensitive financial data in machine learning applications. By leveraging generative models, such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), it is possible to create artificial datasets that preserve the statistical properties of real financial records while mitigating privacy risks and regulatory constraints. Despite the rapid growth of this field, a comprehensive synthesis of the current research landscape has been lacking. This systematic review consolidates and analyses 72 studies published since 2018 that focus on synthetic financial data generation. We categorise the types of financial information synthesised, the generative methods employed, and the evaluation strategies used to assess data utility and privacy. The findings indicate that GAN-based approaches dominate the literature, particularly for generating time-series market data and tabular credit data. While several innovative techniques demonstrate potential for improved realism and privacy preservation, there remains a notable lack of rigorous evaluation of privacy safeguards across studies. By providing an integrated overview of generative techniques, applications, and evaluation methods, this review highlights critical research gaps and offers guidance for future work aimed at developing robust, privacy-preserving synthetic data solutions for the financial domain.
Authors: Yuhua Jiang, Shuang Cheng, Yihao Liu, Ermo Hua, Che Jiang, Weigao Sun, Yu Cheng, Feifei Gao, Biqing Qi, Bowen Zhou
Abstract: Specialized Generalist Models (SGMs) aim to preserve broad capabilities while achieving expert-level performance in target domains. However, traditional LLM structures including Transformer, Linear Attention, and hybrid models do not employ specialized memory mechanism guided by task information. In this paper, we present Nirvana, an SGM with specialized memory mechanism, linear time complexity, and test-time task information extraction. Besides, we propose the Task-Aware Memory Trigger ($\textit{Trigger}$) that flexibly adjusts memory mechanism based on the current task's requirements. In Trigger, each incoming sample is treated as a self-supervised fine-tuning task, enabling Nirvana to adapt its task-related parameters on the fly to domain shifts. We also design the Specialized Memory Updater ($\textit{Updater}$) that dynamically memorizes the context guided by Trigger. We conduct experiments on both general language tasks and specialized medical tasks. On a variety of natural language modeling benchmarks, Nirvana achieves competitive or superior results compared to the existing LLM structures. To prove the effectiveness of Trigger on specialized tasks, we test Nirvana's performance on a challenging medical task, i.e., Magnetic Resonance Imaging (MRI). We post-train frozen Nirvana backbone with lightweight codecs on paired electromagnetic signals and MRI images. Despite the frozen Nirvana backbone, Trigger guides the model to adapt to the MRI domain with the change of task-related parameters. Nirvana achieves higher-quality MRI reconstruction compared to conventional MRI models as well as the models with traditional LLMs' backbone, and can also generate accurate preliminary clinical reports accordingly.
Authors: Zheng Zhang, Haonan Li, Xingyu Li, Hang Zhang, Zhiyun Qian
Abstract: Bug bisection has been an important security task that aims to understand the range of software versions impacted by a bug, i.e., identifying the commit that introduced the bug. However, traditional patch-based bisection methods are faced with several significant barriers: For example, they assume that the bug-inducing commit (BIC) and the patch commit modify the same functions, which is not always true. They often rely solely on code changes, while the commit message frequently contains a wealth of vulnerability-related information. They are also based on simple heuristics (e.g., assuming the BIC initializes lines deleted in the patch) and lack any logical analysis of the vulnerability. In this paper, we make the observation that Large Language Models (LLMs) are well-positioned to break the barriers of existing solutions, e.g., comprehend both textual data and code in patches and commits. Unlike previous BIC identification approaches, which yield poor results, we propose a comprehensive multi-stage pipeline that leverages LLMs to: (1) fully utilize patch information, (2) compare multiple candidate commits in context, and (3) progressively narrow down the candidates through a series of down-selection steps. In our evaluation, we demonstrate that our approach achieves significantly better accuracy than the state-of-the-art solution by more than 38\%. Our results further confirm that the comprehensive multi-stage pipeline is essential, as it improves accuracy by 60\% over a baseline LLM-based bisection method.
Authors: Fazel Arasteh, Arian Haghparast, Manos Papagelis
Abstract: Traffic congestion in urban road networks leads to longer trip times and higher emissions, especially during peak periods. While the Shortest Path First (SPF) algorithm is optimal for a single vehicle in a static network, it performs poorly in dynamic, multi-vehicle settings, often worsening congestion by routing all vehicles along identical paths. We address dynamic vehicle routing through a multi-agent reinforcement learning (MARL) framework for coordinated, network-aware fleet navigation. We first propose Adaptive Navigation (AN), a decentralized MARL model where each intersection agent provides routing guidance based on (i) local traffic and (ii) neighborhood state modeled using Graph Attention Networks (GAT). To improve scalability in large networks, we further propose Hierarchical Hub-based Adaptive Navigation (HHAN), an extension of AN that assigns agents only to key intersections (hubs). Vehicles are routed hub-to-hub under agent control, while SPF handles micro-routing within each hub region. For hub coordination, HHAN adopts centralized training with decentralized execution (CTDE) under the Attentive Q-Mixing (A-QMIX) framework, which aggregates asynchronous vehicle decisions via attention. Hub agents use flow-aware state features that combine local congestion and predictive dynamics for proactive routing. Experiments on synthetic grids and real urban maps (Toronto, Manhattan) show that AN reduces average travel time versus SPF and learning baselines, maintaining 100% routing success. HHAN scales to networks with hundreds of intersections, achieving up to 15.9% improvement under heavy traffic. These findings highlight the potential of network-constrained MARL for scalable, coordinated, and congestion-aware routing in intelligent transportation systems.
Authors: Nick Masi, Randall Balestriero
Abstract: The dominant paradigm in machine learning is to assess model performance based on average loss across all samples in some test set. This amounts to averaging performance geospatially across the Earth in weather and climate settings, failing to account for the non-uniform distribution of human development and geography. We introduce Stratified Assessments of Forecasts over Earth (SAFE), a package for elucidating the stratified performance of a set of predictions made over Earth. SAFE integrates various data domains to stratify by different attributes associated with geospatial gridpoints: territory (usually country), global subregion, income, and landcover (land or water). This allows us to examine the performance of models for each individual stratum of the different attributes (e.g., the accuracy in every individual country). To demonstrate its importance, we utilize SAFE to benchmark a zoo of state-of-the-art AI-based weather prediction models, finding that they all exhibit disparities in forecasting skill across every attribute. We use this to seed a benchmark of model forecast fairness through stratification at different lead times for various climatic variables. By moving beyond globally-averaged metrics, we for the first time ask: where do models perform best or worst, and which models are most fair? To support further work in this direction, the SAFE package is open source and available at https://github.com/N-Masi/safe
Authors: Chenming Tang, Hsiu-Yuan Huang, Weijie Liu, Saiyong Yang, Yunfang Wu
Abstract: Reinforcement learning with verifiable rewards (RLVR) has significantly boosted the reasoning capability of large language models (LLMs) recently. However, existing RLVR approaches merely train LLMs based on their own generated responses and are constrained by the initial capability of LLMs, thus prone to exploration stagnation, in which LLMs fail to solve more training problems and cannot further learn from the training data. Some work tries to address this by leveraging off-policy solutions to training problems but requires external guidance from experts which suffers from limited availability. In this work, we propose LTE (Learning to reason from Trial and Error), an approach hinting LLMs with their previously self-generated incorrect answers and problem of overlong responses, which does not require any external expert guidance. Experiments validate the effectiveness of LTE, which outperforms the normal group relative policy optimization (GRPO) by 6.38 in Pass@1 and 9.00 in Pass@k on average across six mathematics benchmarks for Qwen3-4B-Base. Further analysis confirms that LTE successfully mitigates the problem of exploration stagnation and enhances both exploitation and exploration during training.
Authors: Kexing Liu
Abstract: WiFi Channel State Information (CSI)-based human activity recognition (HAR) provides a privacy-preserving, device-free sensing solution for smart environments. However, its deployment on edge devices is severely constrained by domain shift, where recognition performance deteriorates under varying environmental and hardware conditions. This study presents maxVSTAR (maximally adaptive Vision-guided Sensing Technology for Activity Recognition), a closed-loop, vision-guided model adaptation framework that autonomously mitigates domain shift for edge-deployed CSI sensing systems. The proposed system integrates a cross-modal teacher-student architecture, where a high-accuracy YOLO-based vision model serves as a dynamic supervisory signal, delivering real-time activity labels for the CSI data stream. These labels enable autonomous, online fine-tuning of a lightweight CSI-based HAR model, termed Sensing Technology for Activity Recognition (STAR), directly at the edge. This closed-loop retraining mechanism allows STAR to continuously adapt to environmental changes without manual intervention. Extensive experiments demonstrate the effectiveness of maxVSTAR. When deployed on uncalibrated hardware, the baseline STAR model's recognition accuracy declined from 93.52% to 49.14%. Following a single vision-guided adaptation cycle, maxVSTAR restored the accuracy to 81.51%. These results confirm the system's capacity for dynamic, self-supervised model adaptation in privacy-conscious IoT environments, establishing a scalable and practical paradigm for long-term autonomous HAR using CSI sensing at the network edge.
Authors: Kexing Liu
Abstract: Human Activity Recognition (HAR) via Wi-Fi Channel State Information (CSI) presents a privacy-preserving, contactless sensing approach suitable for smart homes, healthcare monitoring, and mobile IoT systems. However, existing methods often encounter computational inefficiency, high latency, and limited feasibility within resource-constrained, embedded mobile edge environments. This paper proposes STAR (Sensing Technology for Activity Recognition), an edge-AI-optimized framework that integrates a lightweight neural architecture, adaptive signal processing, and hardware-aware co-optimization to enable real-time, energy-efficient HAR on low-power embedded devices. STAR incorporates a streamlined Gated Recurrent Unit (GRU)-based recurrent neural network, reducing model parameters by 33% compared to conventional LSTM models while maintaining effective temporal modeling capability. A multi-stage pre-processing pipeline combining median filtering, 8th-order Butterworth low-pass filtering, and Empirical Mode Decomposition (EMD) is employed to denoise CSI amplitude data and extract spatial-temporal features. For on-device deployment, STAR is implemented on a Rockchip RV1126 processor equipped with an embedded Neural Processing Unit (NPU), interfaced with an ESP32-S3-based CSI acquisition module. Experimental results demonstrate a mean recognition accuracy of 93.52% across seven activity classes and 99.11% for human presence detection, utilizing a compact 97.6k-parameter model. INT8 quantized inference achieves a processing speed of 33 MHz with just 8% CPU utilization, delivering sixfold speed improvements over CPU-based execution. With sub-second response latency and low power consumption, the system ensures real-time, privacy-preserving HAR, offering a practical, scalable solution for mobile and pervasive computing environments.
Authors: Hyuntae Park, Yeachan Kim, SangKeun Lee
Abstract: Molecule and text representation learning has gained increasing interest due to its potential for enhancing the understanding of chemical information. However, existing models often struggle to capture subtle differences between molecules and their descriptions, as they lack the ability to learn fine-grained alignments between molecular substructures and chemical phrases. To address this limitation, we introduce MolBridge, a novel molecule-text learning framework based on substructure-aware alignments. Specifically, we augment the original molecule-description pairs with additional alignment signals derived from molecular substructures and chemical phrases. To effectively learn from these enriched alignments, MolBridge employs substructure-aware contrastive learning, coupled with a self-refinement mechanism that filters out noisy alignment signals. Experimental results show that MolBridge effectively captures fine-grained correspondences and outperforms state-of-the-art baselines on a wide range of molecular benchmarks, highlighting the significance of substructure-aware alignment in molecule-text learning.
Authors: Emilio Mastriani, Alessandro Costa, Federico Incardona, Kevin Munari, Sebastiano Spinello
Abstract: In this study, we investigate the effectiveness of advanced feature engineering and hybrid model architectures for anomaly detection in a multivariate industrial time series, focusing on a steam turbine system. We evaluate the impact of change point-derived statistical features, clustering-based substructure representations, and hybrid learning strategies on detection performance. Despite their theoretical appeal, these complex approaches consistently underperformed compared to a simple Random Forest + XGBoost ensemble trained on segmented data. The ensemble achieved an AUC-ROC of 0.976, F1-score of 0.41, and 100% early detection within the defined time window. Our findings highlight that, in scenarios with highly imbalanced and temporally uncertain data, model simplicity combined with optimized segmentation can outperform more sophisticated architectures, offering greater robustness, interpretability, and operational utility.
Authors: Songxin Lei, Qiongyan Wang, Yanchen Zhu, Hanyu Yao, Sijie Ruan, Weilin Ruan, Yuyu Luo, Huaming Wu, Yuxuan Liang
Abstract: Public resource allocation involves the efficient distribution of resources, including urban infrastructure, energy, and transportation, to effectively meet societal demands. However, existing methods focus on optimizing the movement of individual resources independently, without considering their capacity constraints. To address this limitation, we propose a novel and more practical problem: Collaborative Public Resource Allocation (CPRA), which explicitly incorporates capacity constraints and spatio-temporal dynamics in real-world scenarios. We propose a new framework called Game-Theoretic Spatio-Temporal Reinforcement Learning (GSTRL) for solving CPRA. Our contributions are twofold: 1) We formulate the CPRA problem as a potential game and demonstrate that there is no gap between the potential function and the optimal target, laying a solid theoretical foundation for approximating the Nash equilibrium of this NP-hard problem; and 2) Our designed GSTRL framework effectively captures the spatio-temporal dynamics of the overall system. We evaluate GSTRL on two real-world datasets, where experiments show its superior performance. Our source codes are available in the supplementary materials.
Authors: Yunxiao Shi, Shuo Yang, Yixin Su, Rui Zhang, Min Xu
Abstract: Modern data-centric AI needs precise per-sample influence. Standard SGD-IE approximates leave-one-out effects by summing per-epoch surrogates and ignores cross-epoch compounding, which misranks critical examples. We propose ACC-SGD-IE, a trajectory-aware estimator that propagates the leave-one-out perturbation across training and updates an accumulative influence state at each step. In smooth strongly convex settings it achieves geometric error contraction and, in smooth non-convex regimes, it tightens error bounds; larger mini-batches further reduce constants. Empirically, on Adult, 20 Newsgroups, and MNIST under clean and corrupted data and both convex and non-convex training, ACC-SGD-IE yields more accurate influence estimates, especially over long epochs. For downstream data cleansing it more reliably flags noisy samples, producing models trained on ACC-SGD-IE cleaned data that outperform those cleaned with SGD-IE.
Authors: Avinash Kadimisetty, Arun Rajagopalan, Vijendra SK
Abstract: Reducing preventable hospital readmissions is a national priority for payers, providers, and policymakers seeking to improve health care and lower costs. The rate of readmission is being used as a benchmark to determine the quality of healthcare provided by the hospitals. In thisproject, we have used machine learning techniques like Logistic Regression, Random Forest and Support Vector Machines to analyze the health claims data and identify demographic and medical factors that play a crucial role in predicting all-cause readmissions. As the health claims data is high dimensional, we have used Principal Component Analysis as a dimension reduction technique and used the results for building regression models. We compared and evaluated these models based on the Area Under Curve (AUC) metric. Random Forest model gave the highest performance followed by Logistic Regression and Support Vector Machine models. These models can be used to identify the crucial factors causing readmissions and help identify patients to focus on to reduce the chances of readmission, ultimately bringing down the cost and increasing the quality of healthcare provided to the patients.
Authors: Sekitoshi Kanai, Tsukasa Yoshida, Hiroshi Takahashi, Haru Kuroki, Kazumune Hashimoto
Abstract: Test-time alignment of large language models (LLMs) attracts attention because fine-tuning LLMs requires high computational costs. In this paper, we propose a new test-time alignment method called adaptive importance sampling on pre-logits (AISP) on the basis of the sampling-based model predictive control with the stochastic control input. AISP applies the Gaussian perturbation into pre-logits, which are outputs of the penultimate layer, so as to maximize expected rewards with respect to the mean of the perturbation. We demonstrate that the optimal mean is obtained by importance sampling with sampled rewards. AISP outperforms best-of-n sampling in terms of rewards over the number of used samples and achieves higher rewards than other reward-based test-time alignment methods.
Authors: Minyi Peng, Darian Gunamardi, Ivan Tjuawinata, Kwok-Yan Lam
Abstract: As a new and promising approach, existing machine unlearning (MU) works typically emphasize theoretical formulations or optimization objectives to achieve knowledge removal. However, when deployed in real-world scenarios, such solutions typically face scalability issues and have to address practical requirements such as full access to original datasets and model. In contrast to the existing approaches, we regard classification training as a sequential process where classes are learned sequentially, which we call \emph{inductive approach}. Unlearning can then be done by reversing the last training sequence. This is implemented by appending a projection-redistribution layer in the end of the model. Such an approach does not require full access to the original dataset or the model, addressing the challenges of existing methods. This enables modular and model-agnostic deployment as an output filter into existing classification pipelines with minimal alterations. We conducted multiple experiments across multiple datasets including image (CIFAR-10/100 using CNN-based model) and tabular datasets (Covertype using tree-based model). Experiment results show consistently similar output to a fully retrained model with a high computational cost reduction. This demonstrates the applicability, scalability, and system compatibility of our solution while maintaining the performance of the output in a more practical setting.
Authors: Hieu M. Vu, Tan M. Nguyen
Abstract: Controlling specific behaviors in large language models while preserving their general capabilities is a central challenge for safe and reliable artificial intelligence deployment. Current steering methods, such as vector addition and directional ablation, are constrained within a two-dimensional subspace defined by the activation and feature direction, making them sensitive to chosen parameters and potentially affecting unrelated features due to unintended interactions in activation space. We introduce Angular Steering, a novel and flexible method for behavior modulation that operates by rotating activations within a fixed two-dimensional subspace. By formulating steering as a geometric rotation toward or away from a target behavior direction, Angular Steering provides continuous, fine-grained control over behaviors such as refusal and compliance. We demonstrate this method using refusal steering emotion steering as use cases. Additionally, we propose Adaptive Angular Steering, a selective variant that rotates only activations aligned with the target feature, further enhancing stability and coherence. Angular Steering generalizes existing addition and orthogonalization techniques under a unified geometric rotation framework, simplifying parameter selection and maintaining model stability across a broader range of adjustments. Experiments across multiple model families and sizes show that Angular Steering achieves robust behavioral control while maintaining general language modeling performance, underscoring its flexibility, generalization, and robustness compared to prior approaches. Code and artifacts are available at https://github.com/lone17/angular-steering/.
Authors: Frederik M\"obius Rygaard, Shen Zhu, Yinzhu Jin, S{\o}ren Hauberg, Tom Fletcher
Abstract: Interpolation in generative models allows for controlled generation, model inspection, and more. Unfortunately, most generative models lack a principal notion of interpolants without restrictive assumptions on either the model or data dimension. In this paper, we develop a general interpolation scheme that targets likely transition paths compatible with different metrics and probability distributions. We consider interpolants analogous to a geodesic constrained to a suitable data distribution and derive a novel algorithm for computing these curves, which requires no additional training. Theoretically, we show that our method locally can be considered as a geodesic under a suitable Riemannian metric. We quantitatively show that our interpolation scheme traverses higher density regions than baselines across a range of models and datasets.
Authors: Kim Yong Tan, Yueming Lyu, Ivor Tsang, Yew-Soon Ong
Abstract: Diffusion models have been successful in learning complex data distributions. This capability has driven their application to high-dimensional multi-objective black-box optimization problem. Existing approaches often employ an external optimization loop, such as an evolutionary algorithm, to the diffusion model. However, these approaches treat the diffusion model as a black-box refiner, which overlooks the internal distribution transition of the diffusion generation process, limiting their efficiency. To address these challenges, we propose the Inference-time Multi-target Generation (IMG) algorithm, which optimizes the diffusion process at inference-time to generate samples that simultaneously satisfy multiple objectives. Specifically, our IMG performs weighted resampling during the diffusion generation process according to the expected aggregated multi-objective values. This weighted resampling strategy ensures the diffusion-generated samples are distributed according to our desired multi-target Boltzmann distribution. We further derive that the multi-target Boltzmann distribution has an interesting log-likelihood interpretation, where it is the optimal solution to the distributional multi-objective optimization problem. We implemented IMG for a multi-objective molecule generation task. Experiments show that IMG, requiring only a single generation pass, achieves a significantly higher hypervolume than baseline optimization algorithms that often require hundreds of diffusion generations. Notably, our algorithm can be viewed as an optimized diffusion process and can be integrated into existing methods to further improve their performance.
Authors: Xia Jiang, Rong J. B. Zhu
Abstract: Multi-task learning in contextual bandits has attracted significant research interest due to its potential to enhance decision-making across multiple related tasks by leveraging shared structures and task-specific heterogeneity. In this article, we propose a novel hierarchical Bayesian framework for learning in various bandit instances. This framework captures both the heterogeneity and the correlations among different bandit instances through a hierarchical Bayesian model, enabling effective information sharing while accommodating instance-specific variations. Unlike previous methods that overlook the learning of the covariance structure across bandits, we introduce an empirical Bayesian approach to estimate the covariance matrix of the prior distribution.This enhances both the practicality and flexibility of learning across multi-bandits. Building on this approach, we develop two efficient algorithms: ebmTS (Empirical Bayesian Multi-Bandit Thompson Sampling) and ebmUCB (Empirical Bayesian Multi-Bandit Upper Confidence Bound), both of which incorporate the estimated prior into the decision-making process. We provide the frequentist regret upper bounds for the proposed algorithms, thereby filling a research gap in the field of multi-bandit problems. Extensive experiments on both synthetic and real-world datasets demonstrate the superior performance of our algorithms, particularly in complex environments. Our methods achieve lower cumulative regret compared to existing techniques, highlighting their effectiveness in balancing exploration and exploitation across multi-bandits.
Authors: Jingyuan Liu, Fatemeh Ghaffari, Xuchuang Wang, Mohammad Hajiesmaili, Carlee Joe-Wong
Abstract: Preference learning from pairwise feedback is a widely adopted framework in applications such as reinforcement learning with human feedback and recommendations. In many practical settings, however, user interactions are limited or costly, making offline preference learning necessary. Moreover, real-world preference learning often involves users with different preferences. For example, annotators from different backgrounds may rank the same responses differently. This setting presents two central challenges: (1) identifying similarity across users to effectively aggregate data, especially under scenarios where offline data is imbalanced across dimensions, and (2) handling the imbalanced offline data where some preference dimensions are underrepresented. To address these challenges, we study the Offline Clustering of Preference Learning problem, where the learner has access to fixed datasets from multiple users with potentially different preferences and aims to maximize utility for a test user. To tackle the first challenge, we first propose Off-C$^2$PL for the pure offline setting, where the learner relies solely on offline data. Our theoretical analysis provides a suboptimality bound that explicitly captures the tradeoff between sample noise and bias. To address the second challenge of inbalanced data, we extend our framework to the setting with active-data augmentation where the learner is allowed to select a limited number of additional active-data for the test user based on the cluster structure learned by Off-C$^2$PL. In this setting, our second algorithm, A$^2$-Off-C$^2$PL, actively selects samples that target the least-informative dimensions of the test user's preference. We prove that these actively collected samples contribute more effectively than offline ones. Finally, we validate our theoretical results through simulations on synthetic and real-world datasets.
Authors: Ziliang Chen, Tianang Xiao, Jusheng Zhang, Yongsen Zheng, Xipeng Chen
Abstract: Contrastive Language-Image Pre-training (CLIP) delivers strong cross modal generalization by aligning images and texts in a shared embedding space, yet it persistently fails at compositional reasoning over objects, attributes, and relations often behaving like a bag-of-words matcher. Prior causal accounts typically model text as a single vector, obscuring token-level structure and leaving core phenomena-such as prompt sensitivity and failures on hard negatives unexplained. We address this gap with a token-aware causal representation learning (CRL) framework grounded in a sequential, language-token SCM. Our theory extends block identifiability to tokenized text, proving that CLIP's contrastive objective can recover the modal-invariant latent variable under both sentence-level and token-level SCMs. Crucially, token granularity yields the first principled explanation of CLIP's compositional brittleness: composition nonidentifiability. We show the existence of pseudo-optimal text encoders that achieve perfect modal-invariant alignment yet are provably insensitive to SWAP, REPLACE, and ADD operations over atomic concepts, thereby failing to distinguish correct captions from hard negatives despite optimizing the same training objective as true-optimal encoders. The analysis further links language-side nonidentifiability to visual-side failures via the modality gap and shows how iterated composition operators compound hardness, motivating improved negative mining strategies.
Authors: Beomhan Baek, Minhak Song, Chulhee Yun
Abstract: Adam [Kingma and Ba, 2015] is the de facto optimizer in deep learning, yet its theoretical understanding remains limited. Prior analyses show that Adam favors solutions aligned with $\ell_\infty$-geometry, but these results are restricted to the full-batch regime. In this work, we study the implicit bias of incremental Adam (using one sample per step) for logistic regression on linearly separable data, and we show that its bias can deviate from the full-batch behavior. To illustrate this, we construct a class of structured datasets where incremental Adam provably converges to the $\ell_2$-max-margin classifier, in contrast to the $\ell_\infty$-max-margin bias of full-batch Adam. For general datasets, we develop a proxy algorithm that captures the limiting behavior of incremental Adam as $\beta_2 \to 1$ and we characterize its convergence direction via a data-dependent dual fixed-point formulation. Finally, we prove that, unlike Adam, Signum [Bernstein et al., 2018] converges to the $\ell_\infty$-max-margin classifier for any batch size by taking $\beta$ close enough to 1. Overall, our results highlight that the implicit bias of Adam crucially depends on both the batching scheme and the dataset, while Signum remains invariant.
Authors: Ruilin Tong, Haodong Lu, Yuhang Liu, Dong Gong
Abstract: Continual learning (CL) aims to incrementally train a model on a sequence of tasks while retaining performance on prior ones. However, storing and replaying data is often infeasible due to privacy or security constraints and impractical for arbitrary pre-trained models. Data-free CL seeks to update models without access to previous data. Beyond regularization, we employ model inversion to synthesize data from the trained model, enabling replay without storing samples. Yet, model inversion in predictive models faces two challenges: (1) generating inputs solely from compressed output labels causes drift between synthetic and real data, and replaying such data can erode prior knowledge; (2) inversion is computationally expensive since each step backpropagates through the full model. These issues are amplified in large pre-trained models such as CLIP. To improve efficiency, we propose Per-layer Model Inversion (PMI), inspired by faster convergence in single-layer optimization. PMI provides strong initialization for full-model inversion, substantially reducing iterations. To mitigate feature shift, we model class-wise features via Gaussian distributions and contrastive model, ensuring alignment between synthetic and real features. Combining PMI and feature modeling, our approach enables continual learning of new classes by generating pseudo-images from semantic-aware projected features, achieving strong effectiveness and compatibility across multiple CL settings.
Authors: Sascha M\"ucke
Abstract: Training Support Vector Machines (SVMs) can be formulated as a QUBO problem, enabling the use of quantum annealing for model optimization. In this work, we study how the number of qubits - linked to the discretization level of dual weights - affects predictive performance across datasets. We compare QUBO-based SVM training to the classical LIBSVM solver and find that even low-precision QUBO encodings (e.g., 1 bit per parameter) yield competitive, and sometimes superior, accuracy. While increased bit-depth enables larger regularization parameters, it does not always improve classification. Our findings suggest that selecting the right support vectors may matter more than their precise weighting. Although current hardware limits the size of solvable QUBOs, our results highlight the potential of quantum annealing for efficient SVM training as quantum devices scale.
Authors: Zhiyang Xun, Shivam Gupta, Eric Price
Abstract: Given a noisy linear measurement $y = Ax + \xi$ of a distribution $p(x)$, and a good approximation to the prior $p(x)$, when can we sample from the posterior $p(x \mid y)$? Posterior sampling provides an accurate and fair framework for tasks such as inpainting, deblurring, and MRI reconstruction, and several heuristics attempt to approximate it. Unfortunately, approximate posterior sampling is computationally intractable in general. To sidestep this hardness, we focus on (local or global) log-concave distributions $p(x)$. In this regime, Langevin dynamics yields posterior samples when the exact scores of $p(x)$ are available, but it is brittle to score--estimation error, requiring an MGF bound (sub-exponential error). By contrast, in the unconditional setting, diffusion models succeed with only an $L^2$ bound on the score error. We prove that combining diffusion models with an annealed variant of Langevin dynamics achieves conditional sampling in polynomial time using merely an $L^4$ bound on the score error.
Authors: David Schmotz, Sahar Abdelnabi, Maksym Andriushchenko
Abstract: Enabling continual learning in LLMs remains a key unresolved research challenge. In a recent announcement, a frontier LLM company made a step towards this by introducing Agent Skills, a framework that equips agents with new knowledge based on instructions stored in simple markdown files. Although Agent Skills can be a very useful tool, we show that they are fundamentally insecure, since they enable trivially simple prompt injections. We demonstrate how to hide malicious instructions in long Agent Skill files and referenced scripts to exfiltrate sensitive data, such as internal files or passwords. Importantly, we show how to bypass system-level guardrails of a popular coding agent: a benign, task-specific approval with the "Don't ask again" option can carry over to closely related but harmful actions. Overall, we conclude that despite ongoing research efforts and scaling model capabilities, frontier LLMs remain vulnerable to very simple prompt injections in realistic scenarios. Our code is available at https://github.com/aisa-group/promptinject-agent-skills.
URLs: https://github.com/aisa-group/promptinject-agent-skills.
Authors: Zhigao Guo, Feng Dong
Abstract: Incorporating causal knowledge and mechanisms is essential for refining causal models and improving downstream tasks such as designing new treatments. In this paper, we introduce a novel concept in causal discovery, termed interventional constraints, which differs fundamentally from interventional data. While interventional data require direct perturbations of variables, interventional constraints encode high-level causal knowledge in the form of inequality constraints on causal effects. For instance, in the Sachs dataset (Sachs et al.\ 2005), Akt has been shown to be activated by PIP3, meaning PIP3 exerts a positive causal effect on Akt. Existing causal discovery methods allow enforcing structural constraints (for example, requiring a causal path from PIP3 to Akt), but they may still produce incorrect causal conclusions such as learning that "PIP3 inhibits Akt". Interventional constraints bridge this gap by explicitly constraining the total causal effect between variable pairs, ensuring learned models respect known causal influences. To formalize interventional constraints, we propose a metric to quantify total causal effects for linear causal models and formulate the problem as a constrained optimization task, solved using a two-stage constrained optimization method. We evaluate our approach on real-world datasets and demonstrate that integrating interventional constraints not only improves model accuracy and ensures consistency with established findings, making models more explainable, but also facilitates the discovery of new causal relationships that would otherwise be costly to identify.
Authors: Sebastian Zieglmeier, Niklas Erdmann, Narada D. Warakagoda
Abstract: Reinforcement learning (RL) algorithms are designed to optimize problem-solving by learning actions that maximize rewards, a task that becomes particularly challenging in random and nonstationary environments. Even advanced RL algorithms are often limited in their ability to solve problems in these conditions. In applications such as searching for underwater pollution clouds with autonomous underwater vehicles (AUVs), RL algorithms must navigate reward-sparse environments, where actions frequently result in a zero reward. This paper aims to address these challenges by revisiting and modifying classical RL approaches to efficiently operate in sparse, randomized, and nonstationary environments. We systematically study a large number of modifications, including hierarchical algorithm changes, multigoal learning, and the integration of a location memory as an external output filter to prevent state revisits. Our results demonstrate that a modified Monte Carlo-based approach significantly outperforms traditional Q-learning and two exhaustive search patterns, illustrating its potential in adapting RL to complex environments. These findings suggest that reinforcement learning approaches can be effectively adapted for use in random, nonstationary, and reward-sparse environments.
Authors: Furkan Pala, Islem Rekik
Abstract: Federated learning (FL) has emerged as a key paradigm for collaborative model training across multiple clients without sharing raw data, enabling privacy-preserving applications in areas such as radiology and pathology. However, works on collaborative training across clients with fundamentally different neural architectures and non-identically distributed datasets remain scarce. Existing FL frameworks face several limitations. Despite claiming to support architectural heterogeneity, most recent FL methods only tolerate variants within a single model family (e.g., shallower, deeper, or wider CNNs), still presuming a shared global architecture and failing to accommodate federations where clients deploy fundamentally different network types (e.g., CNNs, GNNs, MLPs). Moreover, existing approaches often address only statistical heterogeneity while overlooking the domain-fracture problem, where each client's data distribution differs markedly from that faced at testing time, undermining model generalizability. When clients use different architectures, have non-identically distributed data, and encounter distinct test domains, current methods perform poorly. To address these challenges, we propose UnifiedFL, a dynamic federated learning framework that represents heterogeneous local networks as nodes and edges in a directed model graph optimized by a shared graph neural network (GNN). UnifiedFL introduces (i) a common GNN to parameterize all architectures, (ii) distance-driven clustering via Euclidean distances between clients' parameters, and (iii) a two-tier aggregation policy balancing convergence and diversity. Experiments on MedMNIST classification and hippocampus segmentation benchmarks demonstrate UnifiedFL's superior performance. Code and data: https://github.com/basiralab/UnifiedFL
Authors: Albi Isufaj, Pablo Moll\'a, Helmut Prendinger
Abstract: Financial forecasting increasingly uses large neural network models, but their opacity raises challenges for trust and regulatory compliance. We present several approaches to explainable and reliable AI in finance. \emph{First}, we describe how Time-LLM, a time series foundation model, uses a prompt to avoid a wrong directional forecast. \emph{Second}, we show that combining foundation models for time series forecasting with a reliability estimator can filter our unreliable predictions. \emph{Third}, we argue for symbolic reasoning encoding domain rules for transparent justification. These approaches shift emphasize executing only forecasts that are both reliable and explainable. Experiments on equity and cryptocurrency data show that the architecture reduces false positives and supports selective execution. By integrating predictive performance with reliability estimation and rule-based reasoning, our framework advances transparent and auditable financial AI systems.
Authors: Kazuma Kano, Yuki Mori, Shin Katayama, Kenta Urano, Takuro Yonezawa, Nobuo Kawaguchi
Abstract: Worker location data is key to higher productivity in industrial sites. Cameras are a promising tool for localization in logistics warehouses since they also offer valuable environmental contexts such as package status. However, identifying individuals with only visual data is often impractical. Accordingly, several prior studies identified people in videos by comparing their trajectories and wearable sensor measurements. While this approach has advantages such as independence from appearance, the existing methods may break down under real-world conditions. To overcome this challenge, we propose CorVS, a novel data-driven person identification method based on correspondence between visual tracking trajectories and sensor measurements. Firstly, our deep learning model predicts correspondence probabilities and reliabilities for every pair of a trajectory and sensor measurements. Secondly, our algorithm matches the trajectories and sensor measurements over time using the predicted probabilities and reliabilities. We developed a dataset with actual warehouse operations and demonstrated the method's effectiveness for real-world applications.
Authors: Ningning Tao, Fei Xie, Baoxiang Pan, Hongyu Wang, Han Huang, Zhongpu Qiu, Ke Gui, Jiali Luo, Xiaosong Chen
Abstract: Sudden Stratospheric Warmings (SSWs) are key sources of subseasonal predictability and major drivers of extreme winter weather. Yet, their accurate and efficient forecast remains a persistent challenge for numerical weather prediction (NWP) systems due to limitations in physical representation, initialization, and the immense computational demands of ensemble forecasts. While data-driven forecasting is rapidly evolving, its application to the complex, three-dimensional dynamics of SSWs, particularly for probabilistic forecast, remains underexplored. Here, we bridge this gap by developing a Flow Matching-based generative AI model (FM-Cast) for efficient and skillful probabilistic forecasting of the spatiotemporal evolution of stratospheric circulation. Evaluated across 18 major SSW events (1998-2024), FM-Cast skillfully forecasts the onset, intensity, and morphology of 10 events up to 20 days in advance, achieving ensemble accuracies above 50%. Its performance is comparable to or exceeds leading NWP systems while requiring only two minutes for a 50-member, 30-day forecast on a consumer GPU. Furthermore, leveraging FM-Cast as a scientific tool, we demonstrate through idealized experiments that SSW predictability is fundamentally linked to its underlying physical drivers, distinguishing between events forced from the troposphere and those driven by internal stratospheric dynamics. Our work thus establishes a computationally efficient paradigm for probabilistic forecasting stratospheric anomalies and showcases generative AI's potential to deepen the physical understanding of atmosphere-climate dynamics.
Authors: Wenchang Duan, Yaoliang Yu, Jiwan He, Yi Shi
Abstract: Recently, deep multi-agent reinforcement learning (MARL) has demonstrated promising performance for solving challenging tasks, such as long-term dependencies and non-Markovian environments. Its success is partly attributed to conditioning policies on large fixed context length. However, such large fixed context lengths may lead to limited exploration efficiency and redundant information. In this paper, we propose a novel MARL framework to obtain adaptive and effective contextual information. Specifically, we design a central agent that dynamically optimizes context length via temporal gradient analysis, enhancing exploration to facilitate convergence to global optima in MARL. Furthermore, to enhance the adaptive optimization capability of the context length, we present an efficient input representation for the central agent, which effectively filters redundant information. By leveraging a Fourier-based low-frequency truncation method, we extract global temporal trends across decentralized agents, providing an effective and efficient representation of the MARL environment. Extensive experiments demonstrate that the proposed method achieves state-of-the-art (SOTA) performance on long-term dependency tasks, including PettingZoo, MiniGrid, Google Research Football (GRF), and StarCraft Multi-Agent Challenge v2 (SMACv2).
Authors: Fatemeh Bazikar, Hossein Moosaei, Atefeh Hemmati, Panos M. Pardalos
Abstract: Multi-task learning (MTL) enables simultaneous training across related tasks, leveraging shared information to improve generalization, efficiency, and robustness, especially in data-scarce or high-dimensional scenarios. While deep learning dominates recent MTL research, Support Vector Machines (SVMs) and Twin SVMs (TWSVMs) remain relevant due to their interpretability, theoretical rigor, and effectiveness with small datasets. This chapter surveys MTL approaches based on SVM and TWSVM, highlighting shared representations, task regularization, and structural coupling strategies. Special attention is given to emerging TWSVM extensions for multi-task settings, which show promise but remain underexplored. We compare these models in terms of theoretical properties, optimization strategies, and empirical performance, and discuss applications in fields such as computer vision, natural language processing, and bioinformatics. Finally, we identify research gaps and outline future directions for building scalable, interpretable, and reliable margin-based MTL frameworks. This work provides a comprehensive resource for researchers and practitioners interested in SVM- and TWSVM-based multi-task learning.
Authors: Yucen Wang, Fengming Zhang, De-Chuan Zhan, Li Zhao, Kaixin Wang, Jiang Bian
Abstract: Adapting pre-trained video generation models into controllable world models via latent actions is a promising step towards creating generalist world models. The dominant paradigm adopts a two-stage approach that trains latent action model (LAM) and the world model separately, resulting in redundant training and limiting their potential for co-adaptation. A conceptually simple and appealing idea is to directly replace the forward dynamic model in LAM with a powerful world model and training them jointly, but it is non-trivial and prone to representational collapse. In this work, we propose CoLA-World, which for the first time successfully realizes this synergistic paradigm, resolving the core challenge in joint learning through a critical warm-up phase that effectively aligns the representations of the from-scratch LAM with the pre-trained world model. This unlocks a co-evolution cycle: the world model acts as a knowledgeable tutor, providing gradients to shape a high-quality LAM, while the LAM offers a more precise and adaptable control interface to the world model. Empirically, CoLA-World matches or outperforms prior two-stage methods in both video simulation quality and downstream visual planning, establishing a robust and efficient new paradigm for the field.
Authors: Wenjie Chen, Li Zhuang, Ziying Luo, Yu Liu, Jiahao Wu, Shengcai Liu
Abstract: Personalized treatment outcome prediction based on trial data for small-sample and rare patient groups is critical in precision medicine. However, the costly trial data limit the prediction performance. To address this issue, we propose a cross-fidelity knowledge distillation and adaptive fusion network (CFKD-AFN), which leverages abundant but low-fidelity simulation data to enhance predictions on scarce but high-fidelity trial data. CFKD-AFN incorporates a dual-channel knowledge distillation module to extract complementary knowledge from the low-fidelity model, along with an attention-guided fusion module to dynamically integrate multi-source information. Experiments on treatment outcome prediction for the chronic obstructive pulmonary disease demonstrates significant improvements of CFKD-AFN over state-of-the-art methods in prediction accuracy, ranging from 6.67\% to 74.55\%, and strong robustness to varying high-fidelity dataset sizes. Furthermore, we extend CFKD-AFN to an interpretable variant, enabling the exploration of latent medical semantics to support clinical decision-making.
Authors: Jiayi Luo, Qingyun Sun, Beining Yang, Haonan Yuan, Xingcheng Fu, Yanbiao Ma, Jianxin Li, Philip S. Yu
Abstract: Graph condensation (GC) has gained significant attention for its ability to synthesize smaller yet informative graphs. However, existing studies often overlook the robustness of GC in scenarios where the original graph is corrupted. In such cases, we observe that the performance of GC deteriorates significantly, while existing robust graph learning technologies offer only limited effectiveness. Through both empirical investigation and theoretical analysis, we reveal that GC is inherently an intrinsic-dimension-reducing process, synthesizing a condensed graph with lower classification complexity. Although this property is critical for effective GC performance, it remains highly vulnerable to adversarial perturbations. To tackle this vulnerability and improve GC robustness, we adopt the geometry perspective of graph data manifold and propose a novel Manifold-constrained Robust Graph Condensation framework named MRGC. Specifically, we introduce three graph data manifold learning modules that guide the condensed graph to lie within a smooth, low-dimensional manifold with minimal class ambiguity, thereby preserving the classification complexity reduction capability of GC and ensuring robust performance under universal adversarial attacks. Extensive experiments demonstrate the robustness of \ModelName\ across diverse attack scenarios.
Authors: Qiaoling Chen, Zijun Liu, Peng Sun, Shenggui Li, Guoteng Wang, Ziming Liu, Yonggang Wen, Siyuan Feng, Tianwei Zhang
Abstract: Adapting large language models (LLMs) via reinforcement learning (RL) is often bottlenecked by the generation stage, which can consume over 75\% of the training time. Speculative decoding (SD) accelerates autoregressive generation in serving systems, but its behavior under RL training remains largely unexplored. We identify three critical gaps that hinder the naive integration of SD into RL systems: diminishing speedups at large batch sizes, drafter staleness under continual actor updates, and drafter-induced policy degradation. To address these gaps, we present ReSpec, a system that adapts SD to RL through three complementary mechanisms: dynamically tuning SD configurations, evolving the drafter via knowledge distillation, and weighting updates by rollout rewards. On Qwen models (3B--14B), ReSpec achieves up to 4.5x speedup while preserving reward convergence and training stability, providing a practical solution for efficient RL-based LLM adaptation.
Authors: Wajdi Hammami, Soumaya Cherkaoui, Jean-Frederic Laprade, Ola Ahmad, Shengrui Wang
Abstract: Anomaly detection in time-series data is a critical challenge with significant implications for network security. Recent quantum machine learning approaches, such as quantum kernel methods and variational quantum circuits, have shown promise in capturing complex data distributions for anomaly detection but remain constrained by limited qubit counts. We introduce in this work a novel Quantum Gated Recurrent Unit (QGRU)-based Generative Adversarial Network (GAN) employing Successive Data Injection (SuDaI) and a multi-metric gating strategy for robust network anomaly detection. Our model uniquely utilizes a quantum-enhanced generator that outputs parameters (mean and log-variance) of a Gaussian distribution via reparameterization, combined with a Wasserstein critic to stabilize adversarial training. Anomalies are identified through a novel gating mechanism that initially flags potential anomalies based on Gaussian uncertainty estimates and subsequently verifies them using a composite of critic scores and reconstruction errors. Evaluated on benchmark datasets, our method achieves a high time-series aware F1 score (TaF1) of 89.43% demonstrating superior capability in detecting anomalies accurately and promptly as compared to existing classical and quantum models. Furthermore, the trained QGRU-WGAN was deployed on real IBM Quantum hardware, where it retained high anomaly detection performance, confirming its robustness and practical feasibility on current noisy intermediate-scale quantum (NISQ) devices.
Authors: Erle Zhu, Dazhi Jiang, Yuan Wang, Xujun Li, Jiale Cheng, Yuxian Gu, Yilin Niu, Aohan Zeng, Jie Tang, Minlie Huang, Hongning Wang
Abstract: Data selection is a critical aspect of Reinforcement Learning with Verifiable Rewards (RLVR) for enhancing the reasoning capabilities of large language models (LLMs). Current data selection methods are largely heuristic-based, lacking theoretical guarantees and generalizability. This work proposes a theoretically-grounded approach using influence functions to estimate the contribution of each data point to the learning objective. To overcome the prohibitive computational cost of policy rollouts required for online influence estimation, we introduce an off-policy influence estimation method that efficiently approximates data influence using pre-collected offline trajectories. Furthermore, to manage the high-dimensional gradients of LLMs, we employ sparse random projection to reduce dimensionality and improve storage and computation efficiency. Leveraging these techniques, we develop \textbf{C}urriculum \textbf{R}L with \textbf{O}ff-\textbf{P}olicy \text{I}nfluence guidance (\textbf{CROPI}), a multi-stage RL framework that iteratively selects the most influential data for the current policy. Experiments on models up to 7B parameters demonstrate that CROPI significantly accelerates training. On a 1.5B model, it achieves a 2.66x step-level acceleration while using only 10\% of the data per stage compared to full-dataset training. Our results highlight the substantial potential of influence-based data selection for efficient RLVR.
Authors: Mustafa Fuad Rifet Ibrahim, Maurice Meijer, Alexander Schlaefer, Peer Stelldinger
Abstract: Continuous electrocardiogram (ECG) monitoring via wearables offers significant potential for early cardiovascular disease (CVD) detection. However, deploying deep learning models for automated analysis in resource-constrained environments faces reliability challenges due to inevitable Out-of-Distribution (OOD) data. OOD inputs, such as unseen pathologies or noisecorrupted signals, often cause erroneous, high-confidence predictions by standard classifiers, compromising patient safety. Existing OOD detection methods either neglect computational constraints or address noise and unseen classes separately. This paper explores Unsupervised Anomaly Detection (UAD) as an independent, upstream filtering mechanism to improve robustness. We benchmark six UAD approaches, including Deep SVDD, reconstruction-based models, Masked Anomaly Detection, normalizing flows, and diffusion models, optimized via Neural Architecture Search (NAS) under strict resource constraints (at most 512k parameters). Evaluation on PTB-XL and BUT QDB datasets assessed detection of OOD CVD classes and signals unsuitable for analysis due to noise. Results show Deep SVDD consistently achieves the best trade-off between detection and efficiency. In a realistic deployment simulation, integrating the optimized Deep SVDD filter with a diagnostic classifier improved accuracy by up to 21 percentage points over a classifier-only baseline. This study demonstrates that optimized UAD filters can safeguard automated ECG analysis, enabling safer, more reliable continuous cardiovascular monitoring on wearables.
Authors: Youssef Attia El Hili, Albert Thomas, Malik Tiomoko, Abdelhakim Benechehab, Corentin L\'eger, Corinne Ancourt, Bal\'azs K\'egl
Abstract: Model and hyperparameter selection are critical but challenging in machine learning, typically requiring expert intuition or expensive automated search. We investigate whether large language models (LLMs) can act as in-context meta-learners for this task. By converting each dataset into interpretable metadata, we prompt an LLM to recommend both model families and hyperparameters. We study two prompting strategies: (1) a zero-shot mode relying solely on pretrained knowledge, and (2) a meta-informed mode augmented with examples of models and their performance on past tasks. Across synthetic and real-world benchmarks, we show that LLMs can exploit dataset metadata to recommend competitive models and hyperparameters without search, and that improvements from meta-informed prompting demonstrate their capacity for in-context meta-learning. These results highlight a promising new role for LLMs as lightweight, general-purpose assistants for model selection and hyperparameter optimization.
Authors: Hsiu-Yuan Huang, Chenming Tang, Weijie Liu, Saiyong Yang, Yunfang Wu
Abstract: Existing Reinforcement Learning from Verifiable Rewards (RLVR) methods, such as Group Relative Policy Optimization (GRPO), have achieved remarkable progress in improving the reasoning capabilities of Large Reasoning Models (LRMs). However, they exhibit limited exploration due to reliance on on-policy rollouts where confined to the current policy's distribution, resulting in narrow trajectory diversity. Recent approaches attempt to expand policy coverage by incorporating trajectories generated from stronger expert models, yet this reliance increases computational cost and such advaned models are often inaccessible. To address these issues, we propose In-Context Steered Policy Optimization (ICPO), a unified framework that leverages the inherent in-context learning capability of LRMs to provide expert guidance using existing datasets. ICPO introduces Mixed-Policy GRPO with Implicit Expert Forcing, which expands exploration beyond the current policy distribution without requiring advanced LRM trajectories. To further stabilize optimization, ICPO integrates Expert Region Reject Sampling to filter unreliable off-policy trajectories and Annealed Expert-Bonus Reward Shaping to balance early expert guidance with later autonomous improvement. Results demonstrate that ICPO consistently enhances reinforcement learning performance and training stability on mathematical reasoning benchmarks, revealing a scalable and effective RLVR paradigm for LRMs.
Authors: Ruilin Wang, Huixia Li, Yuexiao Ma, Xiawu Zheng, Fei Chao, Xuefeng Xiao, Rongrong Ji
Abstract: Inference latency stands as a critical bottleneck in the large-scale deployment of Large Language Models (LLMs). Speculative decoding methods have recently shown promise in accelerating inference without compromising the output distribution. However, existing work typically relies on a dualistic draft-verify framework and lacks rigorous theoretical grounding. In this paper, we introduce a novel \emph{polybasic} speculative decoding framework, underpinned by a comprehensive theoretical analysis. Specifically, we prove a fundamental theorem that characterizes the optimal inference time for multi-model speculative decoding systems, shedding light on how to extend beyond the dualistic approach to a more general polybasic paradigm. Through our theoretical investigation of multi-model token generation, we expose and optimize the interplay between model capabilities, acceptance lengths, and overall computational cost. Our framework supports both standalone implementation and integration with existing speculative techniques, leading to accelerated performance in practice. Experimental results across multiple model families demonstrate that our approach yields speedup ratios ranging from $3.31\times$ to $4.01\times$ for LLaMA2-Chat 7B, up to $3.87 \times$ for LLaMA3-8B, up to $4.43 \times$ for Vicuna-7B and up to $3.85 \times$ for Qwen2-7B -- all while preserving the original output distribution. We release our theoretical proofs and implementation code to facilitate further investigation into polybasic speculative decoding.
Authors: Adrien Weihs, Andrea Bertozzi, Matthew Thorpe
Abstract: Higher-Order Hypergraph Learning (HOHL) was recently introduced as a principled alternative to classical hypergraph regularization, enforcing higher-order smoothness via powers of multiscale Laplacians induced by the hypergraph structure. Prior work established the well- and ill-posedness of HOHL through an asymptotic consistency analysis in geometric settings. We extend this theoretical foundation by proving the consistency of a truncated version of HOHL and deriving explicit convergence rates when HOHL is used as a regularizer in fully supervised learning. We further demonstrate its strong empirical performance in active learning and in datasets lacking an underlying geometric structure, highlighting HOHL's versatility and robustness across diverse learning settings.
Authors: Aidan Furlong, Robert Salko, Xingang Zhao, Xu Wu
Abstract: The use of ML in engineering has grown steadily to support a wide array of applications. Among these methods, deep neural networks have been widely adopted due to their performance and accessibility, but they require large, high-quality datasets. Experimental data are often sparse, noisy, or insufficient to build resilient data-driven models. Transfer learning, which leverages relevant data-abundant source domains to assist learning in data-scarce target domains, has shown efficacy. Parameter transfer, where pretrained weights are reused, is common but degrades under large domain shifts. Domain-adversarial neural networks (DANNs) help address this issue by learning domain-invariant representations, thereby improving transfer under greater domain shifts in a semi-supervised setting. However, DANNs can be unstable during training and lack a native means for uncertainty quantification. This study introduces a fully-supervised three-stage framework, the staged Bayesian domain-adversarial neural network (staged B-DANN), that combines parameter transfer and shared latent space adaptation. In Stage 1, a deterministic feature extractor is trained on the source domain. This feature extractor is then adversarially refined using a DANN in Stage 2. In Stage 3, a Bayesian neural network is built on the adapted feature extractor for fine-tuning on the target domain to handle conditional shifts and yield calibrated uncertainty estimates. This staged B-DANN approach was first validated on a synthetic benchmark, where it was shown to significantly outperform standard transfer techniques. It was then applied to the task of predicting critical heat flux in rectangular channels, leveraging data from tube experiments as the source domain. The results of this study show that the staged B-DANN method can improve predictive accuracy and generalization, potentially assisting other domains in nuclear engineering.
Authors: Jan Stenkamp, Nina Herrmann, Benjamin Karic, Stefan Oehmcke, Fabian Gieseke
Abstract: Deploying machine learning models on compute-constrained devices has become a key building block of modern IoT applications. In this work, we present a compression scheme for boosted decision trees, addressing the growing need for lightweight machine learning models. Specifically, we provide techniques for training compact boosted decision tree ensembles that exhibit a reduced memory footprint by rewarding, among other things, the reuse of features and thresholds during training. Our experimental evaluation shows that models achieved the same performance with a compression ratio of 4-16x compared to LightGBM models using an adapted training process and an alternative memory layout. Once deployed, the corresponding IoT devices can operate independently of constant communication or external energy supply, and, thus, autonomously, requiring only minimal computing power and energy. This capability opens the door to a wide range of IoT applications, including remote monitoring, edge analytics, and real-time decision making in isolated or power-limited environments.
Authors: Nikita Tsoy, Nikola Konstantinov
Abstract: Shortcuts, spurious rules that perform well during training but fail to generalize, present a major challenge to the reliability of deep networks (Geirhos et al., 2020). However, the impact of shortcuts on feature representations remains understudied, obstructing the design of principled shortcut-mitigation methods. To overcome this limitation, we investigate the layer-wise localization of shortcuts in deep models. Our novel experiment design quantifies the layer-wise contribution to accuracy degradation caused by a shortcut-inducing skew by counterfactual training on clean and skewed datasets. We employ our design to study shortcuts on CIFAR-10, Waterbirds, and CelebA datasets across VGG, ResNet, DeiT, and ConvNeXt architectures. We find that shortcut learning is not localized in specific layers but distributed throughout the network. Different network parts play different roles in this process: shallow layers predominantly encode spurious features, while deeper layers predominantly forget core features that are predictive on clean data. We also analyze the differences in localization and describe its principal axes of variation. Finally, our analysis of layer-wise shortcut-mitigation strategies suggests the hardness of designing general methods, supporting dataset- and architecture-specific approaches instead.
Authors: Cesare Barbera, Lorenzo Perini, Giovanni De Toni, Andrea Passerini, Andrea Pugnana
Abstract: Developing trustworthy Machine Learning (ML) models requires their predicted probabilities to be well-calibrated, meaning they should reflect true-class frequencies. Among calibration notions in multiclass classification, strong calibration is the most stringent, as it requires all predicted probabilities to be simultaneously calibrated across all classes. However, existing approaches to multiclass calibration lack a notion of distance among inputs, which makes them vulnerable to proximity bias: predictions in sparse regions of the feature space are systematically miscalibrated. This is especially relevant in high-stakes settings, such as healthcare, where the sparse instances are exactly those most at risk of biased treatment. In this work, we address this main shortcoming by introducing a local perspective on multiclass calibration. First, we formally define multiclass local calibration and establish its relationship with strong calibration. Second, we theoretically analyze the pitfalls of existing evaluation metrics when applied to multiclass local calibration. Third, we propose a practical method for enhancing local calibration in Neural Networks, which enforces alignment between predicted probabilities and local estimates of class frequencies using the Jensen-Shannon distance. Finally, we empirically validate our approach against existing multiclass calibration techniques.
Authors: Maksim Maslov, Alexander Kugaevskikh, Matthew Ivanov
Abstract: This paper considers the problem of regression over distributions, which is becoming increasingly important in machine learning. Existing approaches often ignore the geometry of the probability space or are computationally expensive. To overcome these limitations, a new method is proposed that combines the parameterization of probability trajectories using a Bernstein basis and the minimization of the Wasserstein distance between distributions. The key idea is to model a conditional distribution as a smooth probability trajectory defined by a weighted sum of Gaussian components whose parameters -- the mean and covariance -- are functions of the input variable constructed using Bernstein polynomials. The loss function is the averaged squared Wasserstein distance between the predicted Gaussian distributions and the empirical data, which takes into account the geometry of the distributions. An autodiff-based optimization method is used to train the model. Experiments on synthetic datasets that include complex trajectories demonstrated that the proposed method provides competitive approximation quality in terms of the Wasserstein distance, Energy Distance, and RMSE metrics, especially in cases of pronounced nonlinearity. The model demonstrates trajectory smoothness that is better than or comparable to alternatives and robustness to changes in data structure, while maintaining high interpretability due to explicit parameterization via control points. The developed approach represents a balanced solution that combines geometric accuracy, computational practicality, and interpretability. Prospects for further research include extending the method to non-Gaussian distributions, applying entropy regularization to speed up computations, and adapting the approach to working with high-dimensional data for approximating surfaces and more complex structures.
Authors: Lin Xu, Xinyun Yuan, Yuxuan Liang, Suwan Yin, Yuankai Wu
Abstract: We introduce Aeolus, a large-scale Multi-modal Flight Delay Dataset designed to advance research on flight delay prediction and support the development of foundation models for tabular data. Existing datasets in this domain are typically limited to flat tabular structures and fail to capture the spatiotemporal dynamics inherent in delay propagation. Aeolus addresses this limitation by providing three aligned modalities: (i) a tabular dataset with rich operational, meteorological, and airportlevel features for over 50 million flights; (ii) a flight chain module that models delay propagation along sequential flight legs, capturing upstream and downstream dependencies; and (iii) a flight network graph that encodes shared aircraft, crew, and airport resource connections, enabling cross-flight relational reasoning. The dataset is carefully constructed with temporal splits, comprehensive features, and strict leakage prevention to support realistic and reproducible machine learning evaluation. Aeolus supports a broad range of tasks, including regression, classification, temporal structure modeling, and graph learning, serving as a unified benchmark across tabular, sequential, and graph modalities. We release baseline experiments and preprocessing tools to facilitate adoption. Aeolus fills a key gap for both domain-specific modeling and general-purpose structured data research.Our source code and data can be accessed at https://github.com/Flnny/Delay-data
Authors: Colin Doumont, Victor Picheny, Viacheslav Borovitskiy, Henry Moss
Abstract: Bayesian Optimization (BO) has the potential to solve various combinatorial tasks, ranging from materials science to neural architecture search. However, BO requires specialized kernels to effectively model combinatorial domains. Recent efforts have introduced several combinatorial kernels, but the relationships among them are not well understood. To bridge this gap, we develop a unifying framework based on heat kernels, which we derive in a systematic way and express as simple closed-form expressions. Using this framework, we prove that many successful combinatorial kernels are either related or equivalent to heat kernels, and validate this theoretical claim in our experiments. Moreover, our analysis confirms and extends the results presented in Bounce: certain algorithms' performance decreases substantially when the unknown optima of the function do not have a certain structure. In contrast, heat kernels are not sensitive to the location of the optima. Lastly, we show that a fast and simple pipeline, relying on heat kernels, is able to achieve state-of-the-art results, matching or even outperforming certain slow or complex algorithms.
Authors: Emmanouil Sylligardos, John Paparrizos, Themis Palpanas, Pierre Senellart, Paul Boniol
Abstract: Anomaly detection is a fundamental task for time series analytics with important implications for the downstream performance of many applications. Despite increasing academic interest and the large number of methods proposed in the literature, recent benchmarks and evaluation studies demonstrated that no overall best anomaly detection methods exist when applied to very heterogeneous time series datasets. Therefore, the only scalable and viable solution to solve anomaly detection over very different time series collected from diverse domains is to propose a model selection method that will select, based on time series characteristics, the best anomaly detection methods to run. Existing AutoML solutions are, unfortunately, not directly applicable to time series anomaly detection, and no evaluation of time series-based approaches for model selection exists. Towards that direction, this paper studies the performance of time series classification methods used as model selection for anomaly detection. In total, we evaluate 234 model configurations derived from 16 base classifiers across more than 1980 time series, and we propose the first extensive experimental evaluation of time series classification as model selection for anomaly detection. Our results demonstrate that model selection methods outperform every single anomaly detection method while being in the same order of magnitude regarding execution time. This evaluation is the first step to demonstrate the accuracy and efficiency of time series classification algorithms for anomaly detection, and represents a strong baseline that can then be used to guide the model selection step in general AutoML pipelines. Preprint version of an article accepted at the VLDB Journal.
Authors: Katarina Petrovi\'c, Lazar Atanackovic, Viggo Moro, Kacper Kapu\'sniak, \.Ismail \.Ilkan Ceylan, Michael Bronstein, Avishek Joey Bose, Alexander Tong
Abstract: Modeling the transport dynamics of natural processes from population-level observations is a ubiquitous problem in the natural sciences. Such models rely on key assumptions about the underlying process in order to enable faithful learning of governing dynamics that mimic the actual system behavior. The de facto assumption in current approaches relies on the principle of least action that results in gradient field dynamics and leads to trajectories minimizing an energy functional between two probability measures. However, many real-world systems, such as cell cycles in single-cell RNA, are known to exhibit non-gradient, periodic behavior, which fundamentally cannot be captured by current state-of-the-art methods such as flow and bridge matching. In this paper, we introduce Curly Flow Matching (Curly-FM), a novel approach that is capable of learning non-gradient field dynamics by designing and solving a Schr\"odinger bridge problem with a non-zero drift reference process -- in stark contrast to typical zero-drift reference processes -- which is constructed using inferred velocities in addition to population snapshot data. We showcase Curly-FM by solving the trajectory inference problems for single cells, computational fluid dynamics, and ocean currents with approximate velocities. We demonstrate that Curly-FM can learn trajectories that better match both the reference process and population marginals. Curly-FM expands flow matching models beyond the modeling of populations and towards the modeling of known periodic behavior in physical systems. Our code repository is accessible at: https://github.com/kpetrovicc/curly-flow-matching.git
Authors: Tommaso d'Orsi, Gleb Novikov
Abstract: We revisit the task of computing the span of the top $r$ singular vectors $u_1, \ldots, u_r$ of a matrix under differential privacy. We show that a simple and efficient algorithm -- based on singular value decomposition and standard perturbation mechanisms -- returns a private rank-$r$ approximation whose error depends only on the \emph{rank-$r$ coherence} of $u_1, \ldots, u_r$ and the spectral gap $\sigma_r - \sigma_{r+1}$. This resolves a question posed by Hardt and Roth~\cite{hardt2013beyond}. Our estimator outperforms the state of the art -- significantly so in some regimes. In particular, we show that in the dense setting, it achieves the same guarantees for single-spike PCA in the Wishart model as those attained by optimal non-private algorithms, whereas prior private algorithms failed to do so. In addition, we prove that (rank-$r$) coherence does not increase under Gaussian perturbations. This implies that any estimator based on the Gaussian mechanism -- including ours -- preserves the coherence of the input. We conjecture that similar behavior holds for other structured models, including planted problems in graphs. We also explore applications of coherence to graph problems. In particular, we present a differentially private algorithm for Max-Cut and other constraint satisfaction problems under low coherence assumptions.
Authors: Amir Reza Mirzaei, Yuqiao Wen, Yanshuai Cao, Lili Mou
Abstract: Low-Rank Adaptation (LoRA) has become a popular technique for parameter-efficient fine-tuning of large language models (LLMs). In many real-world scenarios, multiple adapters are loaded simultaneously to enable LLM customization for personalized user experiences or to support a diverse range of tasks. Although each adapter is lightweight in isolation, their aggregate cost becomes substantial at scale. To address this, we propose LoRAQuant, a mixed-precision post-training quantization method tailored to LoRA. Specifically, LoRAQuant reparameterizes each adapter by singular value decomposition (SVD) to concentrate the most important information into specific rows and columns. This makes it possible to quantize the important components to higher precision, while quantizing the rest to ultra-low bitwidth. We conduct comprehensive experiments with LLaMA 2-7B, LLaMA 2-13B, and Mistral 7B models on mathematical reasoning, coding, and summarization tasks. Results show that our LoRAQuant uses significantly lower bits than other quantization methods, but achieves comparable or even higher performance.
Authors: Nick Heilenk\"otter
Abstract: Can regularization terms in the training of invertible neural networks lead to known Bayesian point estimators in reconstruction? Invertible networks are attractive for inverse problems due to their inherent stability and interpretability. Recently, optimization strategies for invertible neural networks that approximate either a reconstruction map or the forward operator have been studied from a Bayesian perspective, but each has limitations. To address this, we introduce and analyze two regularization terms for the network training that, upon inversion of the network, recover properties of classical Bayesian point estimators: while the first can be connected to the posterior mean, the second resembles the MAP estimator. Our theoretical analysis characterizes how each loss shapes both the learned forward operator and its inverse reconstruction map. Numerical experiments support our findings and demonstrate how these loss-term regularizers introduce data-dependence in a stable and interpretable way.
Authors: Giulia DeSalvo, Clara Mohri, Mehryar Mohri, Yutao Zhong
Abstract: Learning to defer uncertain predictions to costly experts offers a powerful strategy for improving the accuracy and efficiency of machine learning systems. However, standard training procedures for deferral algorithms typically require querying all experts for every training instance, an approach that becomes prohibitively expensive when expert queries incur significant computational or resource costs. This undermines the core goal of deferral: to limit unnecessary expert usage. To overcome this challenge, we introduce the budgeted deferral framework, which aims to train effective deferral algorithms while minimizing expert query costs during training. We propose new algorithms for both two-stage and single-stage multiple-expert deferral settings that selectively query only a subset of experts per training example. While inspired by active learning, our setting is fundamentally different: labels are already known, and the core challenge is to decide which experts to query in order to balance cost and predictive performance. We establish theoretical guarantees for both of our algorithms, including generalization bounds and label complexity analyses. Empirical results across several domains show that our algorithms substantially reduce training costs without sacrificing prediction accuracy, demonstrating the practical value of our budget-aware deferral algorithms.
Authors: Chuyan Chen, Chenyang Ma, Zhangxin Li, Yutong He, Yanjie Dong, Kun Yuan
Abstract: Communication remains a central bottleneck in large-scale distributed machine learning, and gradient sparsification has emerged as a promising strategy to alleviate this challenge. However, existing gradient compressors face notable limitations: Rand-$K$\ discards structural information and performs poorly in practice, while Top-$K$\ preserves informative entries but loses the contraction property and requires costly All-Gather operations. In this paper, we propose ARC-Top-$K$, an {All-Reduce}-Compatible Top-$K$ compressor that aligns sparsity patterns across nodes using a lightweight sketch of the gradient, enabling index-free All-Reduce while preserving globally significant information. ARC-Top-$K$\ is provably contractive and, when combined with momentum error feedback (EF21M), achieves linear speedup and sharper convergence rates than the original EF21M under standard assumptions. Empirically, ARC-Top-$K$\ matches the accuracy of Top-$K$\ while reducing wall-clock training time by up to 60.7\%, offering an efficient and scalable solution that combines the robustness of Rand-$K$\ with the strong performance of Top-$K$.
Authors: Jamie Lanyon, Axel Finke, Petros Andreou, Georgina Cosma
Abstract: Machine unlearning (MU) aims to remove the influence of certain data points from a trained model without costly retraining. Most practical MU algorithms are only approximate and their performance can only be assessed empirically. Care must therefore be taken to make empirical comparisons as representative as possible. A common practice is to run the MU algorithm multiple times independently starting from the same trained model. In this work, we demonstrate that this practice can give highly non-representative results because -- even for the same architecture and same dataset -- some MU methods can be highly sensitive to the choice of random number seed used for model training. We therefore recommend that empirical comphttps://info.arxiv.org/help/prep#commentsarisons of MU algorithms should also reflect the variability across different model training seeds.
Authors: Gabriel Asher, Devesh Shah, Amy A. Caudy, Luke Ferro, Lea Amar, Ana S. H. Costa, Thomas Patton, Niall O'Connor, Jennifer M. Campbell, Jack Geremia
Abstract: A vast majority of mass spectrometry data remains uncharacterized, leaving much of its biological and chemical information untapped. Recent advances in machine learning have begun to address this gap, particularly for tasks such as spectral identification in tandem mass spectrometry data. Here, we present the latest generation of LSM-MS2, a large-scale deep learning foundation model trained on millions of spectra to learn a semantic chemical space. LSM-MS2 achieves state-of-the-art performance in spectral identification, improving on existing methods by 30% in accuracy of identifying challenging isomeric compounds, yielding 42% more correct identifications in complex biological samples, and maintaining robustness under low-concentration conditions. Furthermore, LSM-MS2 produces rich spectral embeddings that enable direct biological interpretation from minimal downstream data, successfully differentiating disease states and predicting clinical outcomes across diverse translational applications.
Authors: Tommaso d'Orsi, Gleb Novikov
Abstract: We present a simple perturbation mechanism for the release of $d$-dimensional covariance matrices $\Sigma$ under pure differential privacy. For large datasets with at least $n\geq d^2/\varepsilon$ elements, our mechanism recovers the provably optimal Frobenius norm error guarantees of \cite{nikolov2023private}, while simultaneously achieving best known error for all other $p$-Schatten norms, with $p\in [1,\infty]$. Our error is information-theoretically optimal for all $p\ge 2$, in particular, our mechanism is the first purely private covariance estimator that achieves optimal error in spectral norm. For small datasets $n< d^2/\varepsilon$, we further show that by projecting the output onto the nuclear norm ball of appropriate radius, our algorithm achieves the optimal Frobenius norm error $O(\sqrt{d\;\text{Tr}(\Sigma) /n})$, improving over the known bounds of $O(\sqrt{d/n})$ of \cite{nikolov2023private} and ${O}\big(d^{3/4}\sqrt{\text{Tr}(\Sigma)/n}\big)$ of \cite{dong2022differentially}.
Authors: Muhammad Faraz Ul Abrar, Nicol\`o Michelusi
Abstract: Over-the-air (OTA) federated learning (FL) has been well recognized as a scalable paradigm that exploits the waveform superposition of the wireless multiple-access channel to aggregate model updates in a single use. Existing OTA-FL designs largely enforce zero-bias model updates by either assuming \emph{homogeneous} wireless conditions (equal path loss across devices) or forcing zero-bias updates to guarantee convergence. Under \emph{heterogeneous} wireless scenarios, however, such designs are constrained by the weakest device and inflate the update variance. Moreover, prior analyses of biased OTA-FL largely address convex objectives, while most modern AI models are highly non-convex. Motivated by these gaps, we study OTA-FL with stochastic gradient descent (SGD) for general smooth non-convex objectives under wireless heterogeneity. We develop novel OTA-FL SGD updates that allow a structured, time-invariant model bias while facilitating reduced variance updates. We derive a finite-time stationarity bound (expected time average squared gradient norm) that explicitly reveals a bias-variance trade-off. To optimize this trade-off, we pose a non-convex joint OTA power-control design and develop an efficient successive convex approximation (SCA) algorithm that requires only statistical CSI at the base station. Experiments on a non-convex image classification task validate the approach: the SCA-based design accelerates convergence via an optimized bias and improves generalization over prior OTA-FL baselines.
Authors: Shahriar Noroozizadeh, Vaishnavh Nagarajan, Elan Rosenfeld, Sanjiv Kumar
Abstract: In sequence modeling, the parametric memory of atomic facts has been predominantly abstracted as a brute-force lookup of co-occurrences between entities. We contrast this associative view against a geometric view of how memory is stored. We begin by isolating a clean and analyzable instance of Transformer reasoning that is incompatible with memory as strictly a storage of the local co-occurrences specified during training. Instead, the model must have somehow synthesized its own geometry of atomic facts, encoding global relationships between all entities, including non-co-occurring ones. This in turn has simplified a hard reasoning task involving an $\ell$-fold composition into an easy-to-learn 1-step geometric task. From this phenomenon, we extract fundamental aspects of neural embedding geometries that are hard to explain. We argue that the rise of such a geometry, despite optimizing over mere local associations, cannot be straightforwardly attributed to typical architectural or optimizational pressures. Counterintuitively, an elegant geometry is learned even when it is not more succinct than a brute-force lookup of associations. Then, by analyzing a connection to Node2Vec, we demonstrate how the geometry stems from a spectral bias that -- in contrast to prevailing theories -- indeed arises naturally despite the lack of various pressures. This analysis also points to practitioners a visible headroom to make Transformer memory more strongly geometric. We hope the geometric view of parametric memory encourages revisiting the default intuitions that guide researchers in areas like knowledge acquisition, capacity, discovery and unlearning.
Authors: Marco Federici, Riccardo Del Chiaro, Boris van Breugel, Paul Whatmough, Markus Nagel
Abstract: Quantization is the key method for reducing inference latency, power and memory footprint of generative AI models. However, accuracy often degrades sharply when activations are quantized below eight bits. Recent work suggests that invertible linear transformations (e.g. rotations) can aid quantization, by reparameterizing feature channels and weights. In this paper, we propose \textit{Sequence Transformation and Mixed Precision} (STaMP) quantization, a novel strategy that applies linear transformations along the \textit{sequence} dimension to exploit the strong local correlation in language and visual data. By keeping a small number of tokens in each intermediate activation at higher precision, we can maintain model accuracy at lower (average) activations bit-widths. We evaluate STaMP on recent LVM and LLM architectures, demonstrating that it significantly improves low bit width activation quantization and complements established activation and weight quantization methods including recent feature transformations.
Authors: Jungyeon Koh, Hyeonsu Lyu, Jonggyu Jang, Hyun Jong Yang
Abstract: How can we explain the influence of training data on black-box models? Influence functions (IFs) offer a post-hoc solution by utilizing gradients and Hessians. However, computing the Hessian for an entire dataset is resource-intensive, necessitating a feasible alternative. A common approach involves randomly sampling a small subset of the training data, but this method often results in highly inconsistent IF estimates due to the high variance in sample configurations. To address this, we propose two advanced sampling techniques based on features and logits. These samplers select a small yet representative subset of the entire dataset by considering the stochastic distribution of features or logits, thereby enhancing the accuracy of IF estimations. We validate our approach through class removal experiments, a typical application of IFs, using the F1-score to measure how effectively the model forgets the removed class while maintaining inference consistency on the remaining classes. Our method reduces computation time by 30.1% and memory usage by 42.2%, or improves the F1-score by 2.5% compared to the baseline.
Authors: Andreas Auer, Daniel Klotz, Sebastinan B\"ock, Sepp Hochreiter
Abstract: Recent research on time series foundation models has primarily focused on forecasting, leaving it unclear how generalizable their learned representations are. In this study, we examine whether frozen pre-trained forecasting models can provide effective representations for classification. To this end, we compare different representation extraction strategies and introduce two model-agnostic embedding augmentations. Our experiments show that the best forecasting models achieve classification accuracy that matches or even surpasses that of state-of-the-art models pre-trained specifically for classification. Moreover, we observe a positive correlation between forecasting and classification performance. These findings challenge the assumption that task-specific pre-training is necessary, and suggest that learning to forecast may provide a powerful route toward constructing general-purpose time series foundation models.
Authors: Zaishuo Xia, Yukuan Lu, Xinyi Li, Yifan Xu, Yubei Chen
Abstract: A world model is an internal model that simulates how the world evolves. Given past observations and actions, it predicts the future of both the embodied agent and its environment. Accurate world models are essential for enabling agents to think, plan, and reason effectively in complex, dynamic settings. Despite rapid progress, current world models remain brittle and degrade over long horizons. We argue that a central cause is representation quality: exteroceptive inputs (e.g., images) are high-dimensional, and lossy or entangled latents make dynamics learning unnecessarily hard. We therefore ask whether improving representation learning alone can substantially improve world-model performance. In this work, we take a step toward building a truly accurate world model by addressing a fundamental yet open problem: constructing a model that can fully clone and overfit to a deterministic 3D world. We propose Geometrically-Regularized World Models (GRWM), which enforces that consecutive points along a natural sensory trajectory remain close in latent representation space. This approach yields significantly improved latent representations that align closely with the true topology of the environment. GRWM is plug-and-play, requires only minimal architectural modification, scales with trajectory length, and is compatible with diverse latent generative backbones. Across deterministic 3D settings and long-horizon prediction tasks, GRWM significantly increases rollout fidelity and stability. Analyses show that its benefits stem from learning a latent manifold with superior geometric structure. These findings support a clear takeaway: improving representation learning is a direct and useful path to robust world models, delivering reliable long-horizon predictions without enlarging the dynamics module.
Authors: Mantas Mazeika, Alice Gatti, Cristina Menghini, Udari Madhushani Sehwag, Shivam Singhal, Yury Orlovskiy, Steven Basart, Manasi Sharma, Denis Peskoff, Elaine Lau, Jaehyuk Lim, Lachlan Carroll, Alice Blair, Vinaya Sivakumar, Sumana Basu, Brad Kenstler, Yuntao Ma, Julian Michael, Xiaoke Li, Oliver Ingebretsen, Aditya Mehta, Jean Mottola, John Teichmann, Kevin Yu, Zaina Shaik, Adam Khoja, Richard Ren, Jason Hausenloy, Long Phan, Ye Htet, Ankit Aich, Tahseen Rabbani, Vivswan Shah, Andriy Novykov, Felix Binder, Kirill Chugunov, Luis Ramirez, Matias Geralnik, Hern\'an Mesura, Dean Lee, Ed-Yeremai Hernandez Cardona, Annette Diamond, Summer Yue, Alexandr Wang, Bing Liu, Ernesto Hernandez, Dan Hendrycks
Abstract: AIs have made rapid progress on research-oriented benchmarks of knowledge and reasoning, but it remains unclear how these gains translate into economic value and automation. To measure this, we introduce the Remote Labor Index (RLI), a broadly multi-sector benchmark comprising real-world, economically valuable projects designed to evaluate end-to-end agent performance in practical settings. AI agents perform near the floor on RLI, with the highest-performing agent achieving an automation rate of 2.5%. These results help ground discussions of AI automation in empirical evidence, setting a common basis for tracking AI impacts and enabling stakeholders to proactively navigate AI-driven labor automation.
Authors: Penghui Qi, Zichen Liu, Xiangxin Zhou, Tianyu Pang, Chao Du, Wee Sun Lee, Min Lin
Abstract: Reinforcement learning (RL) fine-tuning of large language models (LLMs) often suffers from instability due to the numerical mismatch between the training and inference policies. While prior work has attempted to mitigate this issue through algorithmic corrections or engineering alignments, we show that its root cause lies in the floating point precision itself. The widely adopted BF16, despite its large dynamic range, introduces large rounding errors that breaks the consistency between training and inference. In this work, we demonstrate that simply reverting to \textbf{FP16} effectively eliminates this mismatch. The change is simple, fully supported by modern frameworks with only a few lines of code change, and requires no modification to the model architecture or learning algorithm. Our results suggest that using FP16 uniformly yields more stable optimization, faster convergence, and stronger performance across diverse tasks, algorithms and frameworks. We hope these findings motivate a broader reconsideration of precision trade-offs in RL fine-tuning.
Authors: Tao Tao, Maissam Barkeshli
Abstract: We study the ability of Transformer models to learn sequences generated by Permuted Congruential Generators (PCGs), a widely used family of pseudo-random number generators (PRNGs). PCGs introduce substantial additional difficulty over linear congruential generators (LCGs) by applying a series of bit-wise shifts, XORs, rotations and truncations to the hidden state. We show that Transformers can nevertheless successfully perform in-context prediction on unseen sequences from diverse PCG variants, in tasks that are beyond published classical attacks. In our experiments we scale moduli up to $2^{22}$ using up to $50$ million model parameters and datasets with up to $5$ billion tokens. Surprisingly, we find even when the output is truncated to a single bit, it can be reliably predicted by the model. When multiple distinct PRNGs are presented together during training, the model can jointly learn them, identifying structures from different permutations. We demonstrate a scaling law with modulus $m$: the number of in-context sequence elements required for near-perfect prediction grows as $\sqrt{m}$. For larger moduli, optimization enters extended stagnation phases; in our experiments, learning moduli $m \geq 2^{20}$ requires incorporating training data from smaller moduli, demonstrating a critical necessity for curriculum learning. Finally, we analyze embedding layers and uncover a novel clustering phenomenon: the model spontaneously groups the integer inputs into bitwise rotationally-invariant clusters, revealing how representations can transfer from smaller to larger moduli.
Authors: Danqi Liao, Chen Liu, Xingzhi Sun, Di\'e Tang, Haochen Wang, Scott Youlten, Srikar Krishna Gopinath, Haejeong Lee, Ethan C. Strayer, Antonio J. Giraldez, Smita Krishnaswamy
Abstract: mRNA design and optimization are important in synthetic biology and therapeutic development, but remain understudied in machine learning. Systematic optimization of mRNAs is hindered by the scarce and imbalanced data as well as complex sequence-function relationships. We present RNAGenScape, a property-guided manifold Langevin dynamics framework that iteratively updates mRNA sequences within a learned latent manifold. RNAGenScape combines an organized autoencoder, which structures the latent space by target properties for efficient and biologically plausible exploration, with a manifold projector that contracts each step of update back to the manifold. RNAGenScape supports property-guided optimization and smooth interpolation between sequences, while remaining robust under scarce and undersampled data, and ensuring that intermediate products are close to the viable mRNA manifold. Across three real mRNA datasets, RNAGenScape improves the target properties with high success rates and efficiency, outperforming various generative or optimization methods developed for proteins or non-biological data. By providing continuous, data-aligned trajectories that reveal how edits influence function, RNAGenScape establishes a scalable paradigm for controllable mRNA design and latent space exploration in mRNA sequence modeling.
Authors: Manideep Pendyala
Abstract: Pulsar surveys generate millions of candidates per run, overwhelming manual inspection. This thesis builds a deep learning pipeline for radio pulsar candidate selection that fuses array-derived features with image diagnostics. From approximately 500 GB of Giant Metrewave Radio Telescope (GMRT) data, raw voltages are converted to filterbanks (SIGPROC), then de-dispersed and folded across trial dispersion measures (PRESTO) to produce approximately 32,000 candidates. Each candidate yields four diagnostics--summed profile, time vs. phase, subbands vs. phase, and DM curve--represented as arrays and images. A baseline stacked model (ANNs for arrays + CNNs for images with logistic-regression fusion) reaches 68% accuracy. We then refine the CNN architecture and training (regularization, learning-rate scheduling, max-norm constraints) and mitigate class imbalance via targeted augmentation, including a GAN-based generator for the minority class. The enhanced CNN attains 87% accuracy; the final GAN+CNN system achieves 94% accuracy with balanced precision and recall on a held-out test set, while remaining lightweight enough for near--real-time triage. The results show that combining array and image channels improves separability over image-only approaches, and that modest generative augmentation substantially boosts minority (pulsar) recall. The methods are survey-agnostic and extensible to forthcoming high-throughput facilities.
Authors: Francesco Spinnato
Abstract: Contemporary chess engines offer precise yet opaque evaluations, typically expressed as centipawn scores. While effective for decision-making, these outputs obscure the underlying contributions of individual pieces or patterns. In this paper, we explore adapting SHAP (SHapley Additive exPlanations) to the domain of chess analysis, aiming to attribute a chess engines evaluation to specific pieces on the board. By treating pieces as features and systematically ablating them, we compute additive, per-piece contributions that explain the engines output in a locally faithful and human-interpretable manner. This method draws inspiration from classical chess pedagogy, where players assess positions by mentally removing pieces, and grounds it in modern explainable AI techniques. Our approach opens new possibilities for visualization, human training, and engine comparison. We release accompanying code and data to foster future research in interpretable chess AI.
Authors: Chiung-Yi Tseng, Somshubhra Roy, Maisha Thasin, Danyang Zhang, Blessing Effiong
Abstract: There is a substantial body of literature examining the mathematical reasoning capabilities of large language models (LLMs), particularly their performance on precise arithmetic operations in autoregressive architectures. However, their ability to perform approximate reasoning in informal, fast-paced mathematical operations has received far less attention, especially among non-autoregressive decoder models. Our work addresses this gap by introducing StreetMath, a benchmark designed to evaluate models' approximation abilities under real-world approximation scenarios. We conduct extensive evaluations across different LLM architectures: Qwen3-4B-Instruct-2507, Qwen3-4B-Thinking-2507, Dream-v0-Instruct-7B, Falcon-Mamba-7B-Instruct, and Mamba-GPT-3B. Furthermore, we apply mechanistic interpretability techniques to probe their internal computational states. Our analysis reveals that LLMs generally attempt to compute exact values or invoke external tools even in tasks that call for approximation. Moreover, while models sometimes reach the correct answer in early layers or steps, they still consume more tokens when solving approximation tasks. Additional experiments indicate that exact and approximate arithmetic operations rely on largely separate neural components. Drawing upon research on cognitive psychology, we argue that LLMs do not exhibit cognitive miserliness in the same way humans do in street math settings. We open source our work https://github.com/ctseng777/StreetMath
Authors: Pratik N. Kalamkar, Anupama G. Phakatkar
Abstract: Opinion mining, also called sentiment analysis, is the field of study that analyzes people opinions, sentiments, evaluations, appraisals, attitudes, and emotions towards entities such as products, services, organizations, individuals, issues, events, topics, and their attributes. Holistic lexicon-based approach does not consider the strength of each opinion, i.e., whether the opinion is very strongly negative (or positive), strongly negative (or positive), moderate negative (or positive), very weakly negative (or positive) and weakly negative (or positive). In this paper, we propose approach to rank entities based on orientation and strength of the entity reviews and user's queries by classifying them in granularity levels (i.e. very weak, weak, moderate, very strong and strong) by combining opinion words (i.e. adverb, adjective, noun and verb) that are related to aspect of interest of certain product. We shall use fuzzy logic algorithmic approach in order to classify opinion words into different category and syntactic dependency resolution to find relations for desired aspect words. Opinion words related to certain aspects of interest are considered to find the entity score for that aspect in the review.
Authors: Dhananjaya Gowda, Seoha Song, Harshith Goka, Junhyun Lee
Abstract: Large language models (LLMs) are increasingly deployed with task-specific adapters catering to multiple downstream applications. In such a scenario, the additional compute associated with these apparently insignificant number of adapter parameters (typically less than 1% of the base model) turns out to be disproportionately significant during inference time (upto 2.5x times that of the base model). In this paper, we propose a new zero-latency fused low-rank adapter (zFLoRA) that introduces zero or negligible latency overhead on top of the base model. Experimental results on LLMs of size 1B, 3B and 7B show that zFLoRA compares favorably against the popular supervised fine-tuning benchmarks including low-rank adapters (LoRA) as well as full fine-tuning (FFT). Experiments are conducted on 18 different tasks across three different categories namely commonsense reasoning, math reasoning and summary-dialogue. Latency measurements made on NPU (Samsung Galaxy S25+) as well as GPU (NVIDIA H100) platforms show that the proposed zFLoRA adapters introduce zero to negligible latency overhead.
Authors: Nikhil Garg, Ismael Balafrej, Joao Henrique Quintino Palhares, Laura B\'egon-Lours, Davide Florini, Donato Francesco Falcone, Tommaso Stecconi, Valeria Bragaglia, Bert Jan Offrein, Jean-Michel Portal, Damien Querlioz, Yann Beilliard, Dominique Drouin, Fabien Alibart
Abstract: The deployment of AI on edge computing devices faces significant challenges related to energy consumption and functionality. These devices could greatly benefit from brain-inspired learning mechanisms, allowing for real-time adaptation while using low-power. In-memory computing with nanoscale resistive memories may play a crucial role in enabling the execution of AI workloads on these edge devices. In this study, we introduce voltage-dependent synaptic plasticity (VDSP) as an efficient approach for unsupervised and local learning in memristive synapses based on Hebbian principles. This method enables online learning without requiring complex pulse-shaping circuits typically necessary for spike-timing-dependent plasticity (STDP). We show how VDSP can be advantageously adapted to three types of memristive devices (TiO$_2$, HfO$_2$-based metal-oxide filamentary synapses, and HfZrO$_4$-based ferroelectric tunnel junctions (FTJ)) with disctinctive switching characteristics. System-level simulations of spiking neural networks incorporating these devices were conducted to validate unsupervised learning on MNIST-based pattern recognition tasks, achieving state-of-the-art performance. The results demonstrated over 83% accuracy across all devices using 200 neurons. Additionally, we assessed the impact of device variability, such as switching thresholds and ratios between high and low resistance state levels, and proposed mitigation strategies to enhance robustness.
Authors: Jayant Biradar, Smit Shah, Tanmay Naik
Abstract: In this paper, we propose a novel hybrid deep learning architecture that synergistically combines Graph Neural Networks (GNNs), Recurrent Neural Networks (RNNs), and multi-head attention mechanisms to significantly enhance cy- bersecurity intrusion detection capabilities. By leveraging the comprehensive UNSW-NB15 dataset containing diverse network traffic patterns, our approach effectively captures both spatial dependencies through graph structural relationships and tem- poral dynamics through sequential analysis of network events. The integrated attention mechanism provides dual benefits of improved model interpretability and enhanced feature selection, enabling cybersecurity analysts to focus computational resources on high-impact security events - a critical requirement in modern real-time intrusion detection systems. Our extensive experimental evaluation demonstrates that the proposed hybrid model achieves superior performance compared to traditional machine learning approaches and standalone deep learning models across multiple evaluation metrics, including accuracy, precision, recall, and F1-score. The model achieves particularly strong performance in detecting sophisticated attack patterns such as Advanced Persistent Threats (APTs), Distributed Denial of Service (DDoS) attacks, and zero-day exploits, making it a promising solution for next-generation cybersecurity applications in complex network environments.
Authors: Charlotte Claye (MICS), Pierre Marschall (MICS), Wassila Ouerdane (MICS), C\'eline Hudelot (MICS), Julien Duquesne
Abstract: Single-cell RNA-seq foundation models achieve strong performance on downstream tasks but remain black boxes, limiting their utility for biological discovery. Recent work has shown that sparse dictionary learning can extract concepts from deep learning models, with promising applications in biomedical imaging and protein models. However, interpreting biological concepts remains challenging, as biological sequences are not inherently human-interpretable. We introduce a novel concept-based interpretability framework for single-cell RNA-seq models with a focus on concept interpretation and evaluation. We propose an attribution method with counterfactual perturbations that identifies genes that influence concept activation, moving beyond correlational approaches like differential expression analysis. We then provide two complementary interpretation approaches: an expert-driven analysis facilitated by an interactive interface and an ontology-driven method with attribution-based biological pathway enrichment. Applying our framework to two well-known single-cell RNA-seq models from the literature, we interpret concepts extracted by Top-K Sparse Auto-Encoders trained on two immune cell datasets. With a domain expert in immunology, we show that concepts improve interpretability compared to individual neurons while preserving the richness and informativeness of the latent representations. This work provides a principled framework for interpreting what biological knowledge foundation models have encoded, paving the way for their use for hypothesis generation and discovery.
Authors: Apu Chakraborty, Anshul Kumar, Gagan Raj Gupta
Abstract: Detecting anomalous nodes in attributed networks, where each node is associated with both structural connections and descriptive attributes, is essential for identifying fraud, misinformation, and suspicious behavior in domains such as social networks, academic citation graphs, and e-commerce platforms. We propose Flex-GAD, a novel unsupervised framework for graph anomaly detection at the node level. Flex-GAD integrates two encoders to capture complementary aspects of graph data. The framework incorporates a novel community-based GCN encoder to model intra-community and inter-community information into node embeddings, thereby ensuring structural consistency, along with a standard attribute encoder. These diverse representations are fused using a self-attention-based representation fusion module, which enables adaptive weighting and effective integration of the encoded information. This fusion mechanism allows automatic emphasis of the most relevant node representation across different encoders. We evaluate Flex-GAD on seven real-world attributed graphs with varying sizes, node degrees, and attribute homogeneity. Flex-GAD achieves an average AUC improvement of 7.98% over the previously best-performing method, GAD-NR, demonstrating its effectiveness and flexibility across diverse graph structures. Moreover, it significantly reduces training time, running 102x faster per epoch than Anomaly DAE and 3x faster per epoch than GAD-NR on average across seven benchmark datasets.
Authors: William R\'eveillard, Richard Combes
Abstract: We consider a stochastic multi-armed bandit problem with i.i.d. rewards where the expected reward function is multimodal with at most m modes. We propose the first known computationally tractable algorithm for computing the solution to the Graves-Lai optimization problem, which in turn enables the implementation of asymptotically optimal algorithms for this bandit problem. The code for the proposed algorithms is publicly available at https://github.com/wilrev/MultimodalBandits
Authors: Yilong Lu, Si Chen, Songyan Gao, Han Liu, Xin Dong, Wenfeng Shen, Guangtai Ding
Abstract: Traditional non-biological storage media, such as hard drives, face limitations in both storage density and lifespan due to the rapid growth of data in the big data era. Mirror-image peptides composed of D-amino acids have emerged as a promising biological storage medium due to their high storage density, structural stability, and long lifespan. The sequencing of mirror-image peptides relies on \textit{de-novo} technology. However, its accuracy is limited by the scarcity of tandem mass spectrometry datasets and the challenges that current algorithms encounter when processing these peptides directly. This study is the first to propose improving sequencing accuracy indirectly by optimizing the design of mirror-image peptide sequences. In this work, we introduce DBond, a deep neural network based model that integrates sequence features, precursor ion properties, and mass spectrometry environmental factors for the prediction of mirror-image peptide bond cleavage. In this process, sequences with a high peptide bond cleavage ratio, which are easy to sequence, are selected. The main contributions of this study are as follows. First, we constructed MiPD513, a tandem mass spectrometry dataset containing 513 mirror-image peptides. Second, we developed the peptide bond cleavage labeling algorithm (PBCLA), which generated approximately 12.5 million labeled data based on MiPD513. Third, we proposed a dual prediction strategy that combines multi-label and single-label classification. On an independent test set, the single-label classification strategy outperformed other methods in both single and multiple peptide bond cleavage prediction tasks, offering a strong foundation for sequence optimization.
Authors: Tarun Kumar Chawdhury, Jon D. Duke
Abstract: Electronic Health Records (EHR) store clinical documentation as base64 encoded attachments in FHIR DocumentReference resources, which makes semantic question answering difficult. Traditional vector database methods often miss nuanced clinical relationships. The Clinical Entity Augmented Retrieval (CLEAR) method, introduced by Lopez et al. 2025, uses entity aware retrieval and achieved improved performance with an F1 score of 0.90 versus 0.86 for embedding based retrieval, while using over 70 percent fewer tokens. We developed a Clinical Notes QA Evaluation Platform to validate CLEAR against zero shot large context inference and traditional chunk based retrieval augmented generation. The platform was tested on 12 clinical notes ranging from 10,000 to 65,000 tokens representing realistic EHR content. CLEAR achieved a 58.3 percent win rate, an average semantic similarity of 0.878, and used 78 percent fewer tokens than wide context processing. The largest performance gains occurred on long notes, with a 75 percent win rate for documents exceeding 65,000 tokens. These findings confirm that entity aware retrieval improves both efficiency and accuracy in clinical natural language processing. The evaluation framework provides a reusable and transparent benchmark for assessing clinical question answering systems where semantic precision and computational efficiency are critical.
Authors: Kevin Qiu, Marek Cygan
Abstract: Automating the co-design of a robot's morphology and control is a long-standing challenge due to the vast design space and the tight coupling between body and behavior. We introduce Debate2Create (D2C), a framework in which large language model (LLM) agents engage in a structured dialectical debate to jointly optimize a robot's design and its reward function. In each round, a design agent proposes targeted morphological modifications, and a control agent devises a reward function tailored to exploit the new design. A panel of pluralistic judges then evaluates the design-control pair in simulation and provides feedback that guides the next round of debate. Through iterative debates, the agents progressively refine their proposals, producing increasingly effective robot designs. Notably, D2C yields diverse and specialized morphologies despite no explicit diversity objective. On a quadruped locomotion benchmark, D2C discovers designs that travel 73% farther than the default, demonstrating that structured LLM-based debate can serve as a powerful mechanism for emergent robot co-design. Our results suggest that multi-agent debate, when coupled with physics-grounded feedback, is a promising new paradigm for automated robot design.
Authors: Eit\'an Sprejer, Fernando Avalos, Augusto Bernardi, Jose Pedro Brito de Azevedo Faustino, Jacob Haimes, Narmeen Fatimah Oozeer
Abstract: Aligning LLM-based judges with human preferences is a significant challenge, as they are difficult to calibrate and often suffer from rubric sensitivity, bias, and instability. Overcoming this challenge advances key applications, such as creating reliable reward models for Reinforcement Learning from Human Feedback (RLHF) and building effective routing systems that select the best-suited model for a given user query. In this work, we propose a framework for modeling diverse, persona-based preferences by learning to aggregate outputs from multiple rubric-conditioned judges. We investigate the performance of this approach against naive baselines and assess its robustness through case studies on both human and LLM-judges biases. Our primary contributions include a persona-based method for synthesizing preference labels at scale and two distinct implementations of our aggregator: Generalized Additive Model (GAM) and a Multi-Layer Perceptron (MLP).
Authors: Nicolas Dufour, Lucas Degeorge, Arijit Ghosh, Vicky Kalogeiton, David Picard
Abstract: Current text-to-image generative models are trained on large uncurated datasets to enable diverse generation capabilities. However, this does not align well with user preferences. Recently, reward models have been specifically designed to perform post-hoc selection of generated images and align them to a reward, typically user preference. This discarding of informative data together with the optimizing for a single reward tend to harm diversity, semantic fidelity and efficiency. Instead of this post-processing, we propose to condition the model on multiple reward models during training to let the model learn user preferences directly. We show that this not only dramatically improves the visual quality of the generated images but it also significantly speeds up the training. Our proposed method, called MIRO, achieves state-of-the-art performances on the GenEval compositional benchmark and user-preference scores (PickAScore, ImageReward, HPSv2).
Authors: Nissan Yaron, Dan Bystritsky, Ben-Etzion Yaron
Abstract: We introduce Humans-Junior, a 3.8B model that matches GPT-4o on the FACTS Grounding public subset within a $\pm 5$ pp equivalence margin. Results. On Q1--Q500 under identical judges, GPT-4o scores 73.5% (95% CI 69.5--77.2) and Humans-Junior 72.7% (95% CI 68.7--76.5); the paired difference is 0.8 pp (bootstrap 95% CI $-3.1$ to $+4.7$; permutation $p = 0.72$; Cohen's $d = 0.023$). TOST establishes equivalence at $\pm 5$ pp (not at $\pm 3$ pp). When purchased as managed APIs, Humans-Junior's base model (Phi-3.5-mini-instruct) is $\approx 19\times$ less expensive than GPT-4o on Microsoft AI Foundry pricing; self-hosted or edge deployments can drive incremental inference cost toward zero. Measured vs estimated pricing sources are tabulated in Appendix E. Method. Our approach combines minimal directed "Exoskeleton Reasoning" scaffolds with behavioral fine-tuning that teaches protocol compliance (epistemic discipline) rather than domain answers. Fine-tuning alone adds little; combined, they synergize (+17.7 pp, $p < 0.001$) and reduce variance ($\approx 25\%$). In prompt-only settings on frontier models (Q1--Q100; non-comparable), directed reasoning improved GPT-4o by +11.8 pp to 85.3% and Gemini-2.5-Pro by +5.0 pp to 93.3% (baseline 88.3%, $n = 100$); see Section~5. TL;DR. A 3.8B model achieves GPT-4o-level FACTS accuracy (equivalent within $\pm 5$ pp on Q1--Q500). Cloud pricing shows $\approx 19\times$ lower cost versus GPT-4o, and self-hosted/edge deployments can approach zero marginal cost. Pricing sources are listed in Appendix E. Frontier prompt-only gains (Q1--Q100; non-comparable) and optimized-prompt exploratory results under earlier judges are summarized in Appendix F. Keywords: Small Language Models, Factual Grounding, Directed Reasoning, Fine-Tuning, Model Alignment, Cost-Efficient AI
Authors: Ann Huang, Mitchell Ostrow, Satpreet H. Singh, Leo Kozachkov, Ila Fiete, Kanaka Rajan
Abstract: In control problems and basic scientific modeling, it is important to compare observations with dynamical simulations. For example, comparing two neural systems can shed light on the nature of emergent computations in the brain and deep neural networks. Recently, Ostrow et al. (2023) introduced Dynamical Similarity Analysis (DSA), a method to measure the similarity of two systems based on their recurrent dynamics rather than geometry or topology. However, DSA does not consider how inputs affect the dynamics, meaning that two similar systems, if driven differently, may be classified as different. Because real-world dynamical systems are rarely autonomous, it is important to account for the effects of input drive. To this end, we introduce a novel metric for comparing both intrinsic (recurrent) and input-driven dynamics, called InputDSA (iDSA). InputDSA extends the DSA framework by estimating and comparing both input and intrinsic dynamic operators using a variant of Dynamic Mode Decomposition with control (DMDc) based on subspace identification. We demonstrate that InputDSA can successfully compare partially observed, input-driven systems from noisy data. We show that when the true inputs are unknown, surrogate inputs can be substituted without a major deterioration in similarity estimates. We apply InputDSA on Recurrent Neural Networks (RNNs) trained with Deep Reinforcement Learning, identifying that high-performing networks are dynamically similar to one another, while low-performing networks are more diverse. Lastly, we apply InputDSA to neural data recorded from rats performing a cognitive task, demonstrating that it identifies a transition from input-driven evidence accumulation to intrinsically-driven decision-making. Our work demonstrates that InputDSA is a robust and efficient method for comparing intrinsic dynamics and the effect of external input on dynamical systems.
Authors: Negar Foroutan, Paul Teiletche, Ayush Kumar Tarun, Antoine Bosselut
Abstract: The impact of different multilingual data mixtures in pretraining large language models (LLMs) has been a topic of ongoing debate, often raising concerns about potential trade-offs between language coverage and model performance (i.e., the curse of multilinguality). In this work, we investigate these assumptions by training 1.1B and 3B parameter LLMs on diverse multilingual corpora, varying the number of languages from 25 to 400. Our study challenges common beliefs surrounding multilingual training. First, we find that combining English and multilingual data does not necessarily degrade the in-language performance of either group, provided that languages have a sufficient number of tokens included in the pretraining corpus. Second, we observe that using English as a pivot language (i.e., a high-resource language that serves as a catalyst for multilingual generalization) yields benefits across language families, and contrary to expectations, selecting a pivot language from within a specific family does not consistently improve performance for languages within that family. Lastly, we do not observe a significant "curse of multilinguality" as the number of training languages increases in models at this scale. Our findings suggest that multilingual data, when balanced appropriately, can enhance language model capabilities without compromising performance, even in low-resource settings
Authors: Mengtian Guo, David Gotz, Yue Wang
Abstract: Predictive modeling has the potential to enhance human decision-making. However, many predictive models fail in practice due to problematic problem formulation in cases where the prediction target is an abstract concept or construct and practitioners need to define an appropriate target variable as a proxy to operationalize the construct of interest. The choice of an appropriate proxy target variable is rarely self-evident in practice, requiring both domain knowledge and iterative data modeling. This process is inherently collaborative, involving both domain experts and data scientists. In this work, we explore how human-machine teaming can support this process by accelerating iterations while preserving human judgment. We study the impact of two human-machine teaming strategies on proxy construction: 1) relevance-first: humans leading the process by selecting relevant proxies, and 2) performance-first: machines leading the process by recommending proxies based on predictive performance. Based on a controlled user study of a proxy construction task (N = 20), we show that the performance-first strategy facilitated faster iterations and decision-making, but also biased users towards well-performing proxies that are misaligned with the application goal. Our study highlights the opportunities and risks of human-machine teaming in operationalizing machine learning target variables, yielding insights for future research to explore the opportunities and mitigate the risks.
Authors: Dinghong Song (University of California, Merced, USA), Yuan Feng (University of California, Merced, USA), Yiwei Wang (University of California, Merced, USA), Shangye Chen (University of California, Merced, USA), Cyril Guyot (Western Digital Research, USA), Filip Blagojevic (Western Digital Research, USA), Hyeran Jeon (University of California, Merced, USA), Pengfei Su (University of California, Merced, USA), Dong Li (University of California, Merced, USA)
Abstract: Large Language Models (LLMs) are widely used in generative applications such as chatting, code generation, and reasoning. However, many realworld workloads such as classification, question answering, recommendation, and text embedding rely solely on the prefill stage of inference, where the model encodes input sequences without performing autoregressive decoding. In these prefill only scenarios, the self-attention computation becomes the primary performance bottleneck due to its quadratic complexity with respect to sequence length. In this paper, we observe that semantically different sentences often produce similar attention maps across layers and heads. Building on this insight, we propose AttnCache, a framework that accelerates the prefill stage of LLM inference by retrieving and reusing similar attention maps. Based on an attention map memorization database, AttnCache employs efficient caching and similarity search techniques to identify and reuse pre-cached attention maps during inference, thereby reducing the computational overhead of self-attention. Experimental results show that AttnCache achieves an average of 1.2x end-to-end and 2x attention speedup on CPU, and 1.6x end-to-end and 3x attention speedup on GPU, with negligible accuracy degradation.
Authors: Chaithanya Naik Mude, Linipun Phuttitarn, Satvik Maurya, Kunal Sinha, Mark Saffman, Swamit Tannu
Abstract: Neutral atom quantum computers hold promise for scaling up to hundreds of thousands of qubits, but their progress is constrained by slow qubit readout. Measuring qubits currently takes milliseconds-much longer than the underlying quantum gate operations-making readout the primary bottleneck in deploying quantum error correction. Because each round of QEC depends on measurement, long readout times increase cycle duration and slow down program execution. Reducing the readout duration speeds up cycles and reduces decoherence errors that accumulate while qubits idle, but it also lowers the number of collected photons, making measurements noisier and more error-prone. This tradeoff leaves neutral atom systems stuck between slow but accurate readout and fast but unreliable readout. We show that image denoising can resolve this tension. Our framework, GANDALF, uses explicit denoising using image translation to reconstruct clear signals from short, low-photon measurements, enabling reliable classification at up to 1.6x shorter readout times. Combined with lightweight classifiers and a pipelined readout design, our approach both reduces logical error rate by up to 35x and overall QEC cycle time up to 1.77x compared to state-of-the-art CNN-based readout for Cesium (Cs) Neutral Atom arrays.
Authors: Yihe Deng, I-Hung Hsu, Jun Yan, Zifeng Wang, Rujun Han, Gufeng Zhang, Yanfei Chen, Wei Wang, Tomas Pfister, Chen-Yu Lee
Abstract: Large Language Models (LLMs) often struggle with problems that require multi-step reasoning. For small-scale open-source models, Reinforcement Learning with Verifiable Rewards (RLVR) fails when correct solutions are rarely sampled even after many attempts, while Supervised Fine-Tuning (SFT) tends to overfit long demonstrations through rigid token-by-token imitation. To address this gap, we propose Supervised Reinforcement Learning (SRL), a framework that reformulates problem solving as generating a sequence of logical "actions". SRL trains the model to generate an internal reasoning monologue before committing to each action. It provides smoother rewards based on the similarity between the model's actions and expert actions extracted from the SFT dataset in a step-wise manner. This supervision offers richer learning signals even when all rollouts are incorrect, while encouraging flexible reasoning guided by expert demonstrations. As a result, SRL enables small models to learn challenging problems previously unlearnable by SFT or RLVR. Moreover, initializing training with SRL before refining with RLVR yields the strongest overall performance. Beyond reasoning benchmarks, SRL generalizes effectively to agentic software engineering tasks, establishing it as a robust and versatile training framework for reasoning-oriented LLMs.
Authors: Theresia Veronika Rampisela, Maria Maistro, Tuukka Ruotsalo, Christina Lioma
Abstract: The development of Artificial Intelligence (AI), including AI in Science (AIS), should be done following the principles of responsible AI. Progress in responsible AI is often quantified through evaluation metrics, yet there has been less work on assessing the robustness and reliability of the metrics themselves. We reflect on prior work that examines the robustness of fairness metrics for recommender systems as a type of AI application and summarise their key takeaways into a set of non-exhaustive guidelines for developing reliable metrics of responsible AI. Our guidelines apply to a broad spectrum of AI applications, including AIS.
Authors: Ziji Chen, Steven Chien, Peng Qian, Noa Zilberman
Abstract: Modern machine learning (ML) has grown into a tightly coupled, full-stack ecosystem that combines hardware, software, network, and applications. Many users rely on cloud providers for elastic, isolated, and cost-efficient resources. Unfortunately, these platforms as a service use virtualization, which means operators have little insight into the users' workloads. This hinders resource optimizations by the operator, which is essential to ensure cost efficiency and minimize execution time. In this paper, we argue that workload knowledge is unnecessary for system-level optimization. We propose System-X, which takes a \emph{hardware-centric} approach, relying only on hardware signals -- fully accessible by operators. Using low-level signals collected from the system, System-X detects anomalies through an unsupervised learning pipeline. The pipeline is developed by analyzing over 30 popular ML models on various hardware platforms, ensuring adaptability to emerging workloads and unknown deployment patterns. Using System-X, we successfully identified both network and system configuration issues, accelerating the DeepSeek model by 5.97%.
Authors: Feijie Wu, Weiwu Zhu, Yuxiang Zhang, Soumya Chatterjee, Jiarong Zhu, Fan Mo, Rodin Luo, Jing Gao
Abstract: Current tool-use large language models (LLMs) are trained on static datasets, enabling them to interact with external tools and perform multi-step, tool-integrated reasoning, which produces tool-call trajectories. However, these models imitate how a query is resolved in a generic tool-call routine, thereby failing to explore possible solutions and demonstrating limited performance in an evolved, dynamic tool-call environment. In this work, we propose PORTool, a reinforcement learning (RL) method that encourages a tool-use LLM to explore various trajectories yielding the correct answer. Specifically, this method starts with generating multiple rollouts for a given query, and some of them share the first few tool-call steps, thereby forming a tree-like structure. Next, we assign rewards to each step, based on its ability to produce a correct answer and make successful tool calls. A shared step across different trajectories receives the same reward, while different steps under the same fork receive different rewards. Finally, these step-wise rewards are used to calculate fork-relative advantages, blended with trajectory-relative advantages, to train the LLM for tool use. The experiments utilize 17 tools to address user queries, covering both time-sensitive and time-invariant topics. We conduct ablation studies to systematically justify the necessity and the design robustness of step-wise rewards. Furthermore, we compare the proposed PORTool with other training approaches and demonstrate significant improvements in final accuracy and the number of tool-call steps.
Authors: Feichen Gan, Youcun Lu, Yingying Zhang, Yukun Liu
Abstract: Reliable uncertainty quantification is crucial for reinforcement learning (RL) in high-stakes settings. We propose a unified conformal prediction framework for infinite-horizon policy evaluation that constructs distribution-free prediction intervals {for returns} in both on-policy and off-policy settings. Our method integrates distributional RL with conformal calibration, addressing challenges such as unobserved returns, temporal dependencies, and distributional shifts. We propose a modular pseudo-return construction based on truncated rollouts and a time-aware calibration strategy using experience replay and weighted subsampling. These innovations mitigate model bias and restore approximate exchangeability, enabling uncertainty quantification even under policy shifts. Our theoretical analysis provides coverage guarantees that account for model misspecification and importance weight estimation. Empirical results, including experiments in synthetic and benchmark environments like Mountain Car, show that our method significantly improves coverage and reliability over standard distributional RL baselines.
Authors: Emily Steiner, Daniel van der Spuy, Futian Zhou, Afereti Pama, Minas Liarokapis, Henry Williams
Abstract: While autonomous racing performance in Time-Trial scenarios has seen significant progress and development, autonomous wheel-to-wheel racing and overtaking are still severely limited. These limitations are particularly apparent in real-life driving scenarios where state-of-the-art algorithms struggle to safely or reliably complete overtaking manoeuvres. This is important, as reliable navigation around other vehicles is vital for safe autonomous wheel-to-wheel racing. The F1Tenth Competition provides a useful opportunity for developing wheel-to-wheel racing algorithms on a standardised physical platform. The competition format makes it possible to evaluate overtaking and wheel-to-wheel racing algorithms against the state-of-the-art. This research presents a novel racing and overtaking agent capable of learning to reliably navigate a track and overtake opponents in both simulation and reality. The agent was deployed on an F1Tenth vehicle and competed against opponents running varying competitive algorithms in the real world. The results demonstrate that the agent's training against opponents enables deliberate overtaking behaviours with an overtaking rate of 87% compared 56% for an agent trained just to race.
Authors: Shaoxin Wang, Hanjing Yao
Abstract: Kernel logistic regression (KLR) is a powerful classification method widely applied across diverse domains. In many real-world scenarios, indefinite kernels capture more domain-specific structural information than positive definite kernels. This paper proposes a novel $L_1$-norm regularized indefinite kernel logistic regression (RIKLR) model, which extends the existing IKLR framework by introducing sparsity via an $L_1$-norm penalty. The introduction of this regularization enhances interpretability and generalization while introducing nonsmoothness and nonconvexity into the optimization landscape. To address these challenges, a theoretically grounded and computationally efficient proximal linearized algorithm is developed. Experimental results on multiple benchmark datasets demonstrate the superior performance of the proposed method in terms of both accuracy and sparsity.
Authors: Pengfei Lyu, Zhengchi Ma, Linjun Zhang, Anru R. Zhang
Abstract: Imbalanced data, where the positive samples represent only a small proportion compared to the negative samples, makes it challenging for classification problems to balance the false positive and false negative rates. A common approach to addressing the challenge involves generating synthetic data for the minority group and then training classification models with both observed and synthetic data. However, since the synthetic data depends on the observed data and fails to replicate the original data distribution accurately, prediction accuracy is reduced when the synthetic data is naively treated as the true data. In this paper, we address the bias introduced by synthetic data and provide consistent estimators for this bias by borrowing information from the majority group. We propose a bias correction procedure to mitigate the adverse effects of synthetic data, enhancing prediction accuracy while avoiding overfitting. This procedure is extended to broader scenarios with imbalanced data, such as imbalanced multi-task learning and causal inference. Theoretical properties, including bounds on bias estimation errors and improvements in prediction accuracy, are provided. Simulation results and data analysis on handwritten digit datasets demonstrate the effectiveness of our method.
Authors: Tomoharu Iwata, Futoshi Futami
Abstract: We propose a data-driven framework for efficiently solving quadratic programming (QP) problems by reducing the number of variables in high-dimensional QPs using instance-specific projection. A graph neural network-based model is designed to generate projections tailored to each QP instance, enabling us to produce high-quality solutions even for previously unseen problems. The model is trained on heterogeneous QPs to minimize the expected objective value evaluated on the projected solutions. This is formulated as a bilevel optimization problem; the inner optimization solves the QP under a given projection using a QP solver, while the outer optimization updates the model parameters. We develop an efficient algorithm to solve this bilevel optimization problem, which computes parameter gradients without backpropagating through the solver. We provide a theoretical analysis of the generalization ability of solving QPs with projection matrices generated by neural networks. Experimental results demonstrate that our method produces high-quality feasible solutions with reduced computation time, outperforming existing methods.
Authors: Yuxin Li, Minghao Liu, Ruida Wang, Wenzhao Ji, Zhitao He, Rui Pan, Junming Huang, Tong Zhang, Yi R. Fung
Abstract: We present **Lean4PHYS**, a comprehensive reasoning framework for college-level physics problems in Lean4. **Lean4PHYS** includes *LeanPhysBench*, a college-level benchmark for formal physics reasoning in Lean4, which contains 200 hand-crafted and peer-reviewed statements derived from university textbooks and physics competition problems. To establish a solid foundation for formal reasoning in physics, we also introduce *PhysLib*, a community-driven repository containing fundamental unit systems and theorems essential for formal physics reasoning. Based on the benchmark and Lean4 repository we composed in **Lean4PHYS**, we report baseline results using major expert Math Lean4 provers and state-of-the-art closed-source models, with the best performance of DeepSeek-Prover-V2-7B achieving only 16% and Claude-Sonnet-4 achieving 35%. We also conduct a detailed analysis showing that our *PhysLib* can achieve an average improvement of 11.75% in model performance. This demonstrates the challenging nature of our *LeanPhysBench* and the effectiveness of *PhysLib*. To the best of our knowledge, this is the first study to provide a physics benchmark in Lean4.
Authors: Weifei Jin, Yuxin Cao, Junjie Su, Minhui Xue, Jie Hao, Ke Xu, Jin Song Dong, Derui Wang
Abstract: Recent advances in Audio-Language Models (ALMs) have significantly improved multimodal understanding capabilities. However, the introduction of the audio modality also brings new and unique vulnerability vectors. Previous studies have proposed jailbreak attacks that specifically target ALMs, revealing that defenses directly transferred from traditional audio adversarial attacks or text-based Large Language Model (LLM) jailbreaks are largely ineffective against these ALM-specific threats. To address this issue, we propose ALMGuard, the first defense framework tailored to ALMs. Based on the assumption that safety-aligned shortcuts naturally exist in ALMs, we design a method to identify universal Shortcut Activation Perturbations (SAPs) that serve as triggers that activate the safety shortcuts to safeguard ALMs at inference time. To better sift out effective triggers while preserving the model's utility on benign tasks, we further propose Mel-Gradient Sparse Mask (M-GSM), which restricts perturbations to Mel-frequency bins that are sensitive to jailbreaks but insensitive to speech understanding. Both theoretical analyses and empirical results demonstrate the robustness of our method against both seen and unseen attacks. Overall, \MethodName reduces the average success rate of advanced ALM-specific jailbreak attacks to 4.6% across four models, while maintaining comparable utility on benign benchmarks, establishing it as the new state of the art. Our code and data are available at https://github.com/WeifeiJin/ALMGuard.
Authors: Usman Akram, Fan Zhang, Yang Li, Haris Vikalo
Abstract: Accurate channel state information (CSI) is essential for reliable multiuser MIMO operation. In 5G NR, reciprocity-based beamforming via uplink Sounding Reference Signals (SRS) face resource and coverage constraints, motivating sparse non-uniform SRS allocation. Prior masked-autoencoder (MAE) approaches improve coverage but overfit to training masks and degrade under unseen distortions (e.g., additional masking, interference, clipping, non-Gaussian noise). We propose a diffusion-based channel inpainting framework that integrates system-model knowledge at inference via a likelihood-gradient term, enabling a single trained model to adapt across mismatched conditions. On standardized CDL channels, the score-based diffusion variant consistently outperforms a UNet score-model baseline and the one-step MAE under distribution shift, with improvements up to 14 dB NMSE in challenging settings (e.g., Laplace noise, user interference), while retaining competitive accuracy under matched conditions. These results demonstrate that diffusion-guided inpainting is a robust and generalizable approach for super-capacity SRS design in 5G NR systems.
Authors: Mara Daniels, Liam Hodgkinson, Michael Mahoney
Abstract: Physics-informed machine learning (PIML) integrates prior physical information, often in the form of differential equation constraints, into the process of fitting machine learning models to physical data. Popular PIML approaches, including neural operators, physics-informed neural networks, neural ordinary differential equations, and neural discrete equilibria, are typically fit to objectives that simultaneously include both data and physical constraints. However, the multi-objective nature of this approach creates ambiguity in the measurement of model quality. This is related to a poor understanding of epistemic uncertainty, and it can lead to surprising failure modes, even when existing statistical metrics suggest strong fits. Working within a Gaussian process regression framework, we introduce the Physics-Informed Log Evidence (PILE) score. Bypassing the ambiguities of test losses, the PILE score is a single, uncertainty-aware metric that provides a selection principle for hyperparameters of a PIML model. We show that PILE minimization yields excellent choices for a wide variety of model parameters, including kernel bandwidth, least squares regularization weights, and even kernel function selection. We also show that, even prior to data acquisition, a special 'data-free' case of the PILE score identifies a priori kernel choices that are 'well-adapted' to a given PDE. Beyond the kernel setting, we anticipate that the PILE score can be extended to PIML at large, and we outline approaches to do so.
Authors: Musfiqur Rahman, SayedHassan Khatoonabadi, Emad Shihab
Abstract: Large language models (LLMs) have advanced code generation at the function level, yet their ability to produce correct class-level implementations in authentic software projects remains poorly understood. This work introduces a novel benchmark derived from open-source repositories, comprising real-world classes divided into seen and unseen partitions to evaluate generalization under practical conditions. The evaluation examines multiple LLMs under varied input specifications, retrieval-augmented configurations, and documentation completeness levels. Results reveal a stark performance disparity: LLMs achieve 84% to 89% correctness on established synthetic benchmarks but only 25% to 34% on real-world class tasks, with negligible differences between familiar and novel codebases. Comprehensive docstrings yield modest gains of 1% to 3% in functional accuracy, though statistical significance is rare. Retrieval-augmented generation proves most effective with partial documentation, improving correctness by 4% to 7% by supplying concrete implementation patterns absent from specifications. Error profiling identifies AttributeError, TypeError, and AssertionError as dominant failure modes (84% of cases), with synthetic tests overemphasizing assertion issues and real-world scenarios highlighting type and attribute mismatches. Retrieval augmentation reduces logical flaws but can introduce dependency conflicts. The benchmark and analysis expose critical limitations in current LLM capabilities for class-level engineering, offering actionable insights for enhancing context modelling, documentation strategies, and retrieval integration in production code assistance tools.
Authors: Haohua Duan, Liyao Xiang, Xin Zhang
Abstract: Watermarking schemes for large language models (LLMs) have been proposed to identify the source of the generated text, mitigating the potential threats emerged from model theft. However, current watermarking solutions hardly resolve the trust issue: the non-public watermark detection cannot prove itself faithfully conducting the detection. We observe that it is attributed to the secret key mostly used in the watermark detection -- it cannot be public, or the adversary may launch removal attacks provided the key; nor can it be private, or the watermarking detection is opaque to the public. To resolve the dilemma, we propose PVMark, a plugin based on zero-knowledge proof (ZKP), enabling the watermark detection process to be publicly verifiable by third parties without disclosing any secret key. PVMark hinges upon the proof of `correct execution' of watermark detection on which a set of ZKP constraints are built, including mapping, random number generation, comparison, and summation. We implement multiple variants of PVMark in Python, Rust and Circom, covering combinations of three watermarking schemes, three hash functions, and four ZKP protocols, to show our approach effectively works under a variety of circumstances. By experimental results, PVMark efficiently enables public verifiability on the state-of-the-art LLM watermarking schemes yet without compromising the watermarking performance, promising to be deployed in practice.
Authors: Domenico Amalfitano, Andreas Metzger, Marco Autili, Tommaso Fulcini, Tobias Hey, Jan Keim, Patrizio Pelliccione, Vincenzo Scotti, Anne Koziolek, Raffaela Mirandola, Andreas Vogelsang
Abstract: Generative AI (GenAI) is rapidly transforming software engineering (SE) practices, influencing how SE processes are executed, as well as how software systems are developed, operated, and evolved. This paper applies design science research to build a roadmap for GenAI-augmented SE. The process consists of three cycles that incrementally integrate multiple sources of evidence, including collaborative discussions from the FSE 2025 "Software Engineering 2030" workshop, rapid literature reviews, and external feedback sessions involving peers. McLuhan's tetrads were used as a conceptual instrument to systematically capture the transforming effects of GenAI on SE processes and software products.The resulting roadmap identifies four fundamental forms of GenAI augmentation in SE and systematically characterizes their related research challenges and opportunities. These insights are then consolidated into a set of future research directions. By grounding the roadmap in a rigorous multi-cycle process and cross-validating it among independent author teams and peers, the study provides a transparent and reproducible foundation for analyzing how GenAI affects SE processes, methods and tools, and for framing future research within this rapidly evolving area. Based on these findings, the article finally makes ten predictions for SE in the year 2030.
Authors: Michal \v{S}tef\'anik, Timothee Mickus, Marek Kadl\v{c}\'ik, Bertram H{\o}jer, Michal Spiegel, Ra\'ul V\'azquez, Aman Sinha, Josef Kucha\v{r}, Philipp Mondorf
Abstract: Recent work has shown that different large language models (LLMs) converge to similar and accurate input embedding representations for numbers. These findings conflict with the documented propensity of LLMs to produce erroneous outputs when dealing with numeric information. In this work, we aim to explain this conflict by exploring how language models manipulate numbers and quantify the lower bounds of accuracy of these mechanisms. We find that despite surfacing errors, different language models learn interchangeable representations of numbers that are systematic, highly accurate and universal across their hidden states and the types of input contexts. This allows us to create universal probes for each LLM and to trace information -- including the causes of output errors -- to specific layers. Our results lay a fundamental understanding of how pre-trained LLMs manipulate numbers and outline the potential of more accurate probing techniques in addressed refinements of LLMs' architectures.
Authors: Laura Jiang, Reza Ryan, Qian Li, Nasim Ferdosian
Abstract: Anomaly detection is a critical task in cybersecurity, where identifying insider threats, access violations, and coordinated attacks is essential for ensuring system resilience. Graph-based approaches have become increasingly important for modeling entity interactions, yet most rely on homogeneous and static structures, which limits their ability to capture the heterogeneity and temporal evolution of real-world environments. Heterogeneous Graph Neural Networks (HGNNs) have emerged as a promising paradigm for anomaly detection by incorporating type-aware transformations and relation-sensitive aggregation, enabling more expressive modeling of complex cyber data. However, current research on HGNN-based anomaly detection remains fragmented, with diverse modeling strategies, limited comparative evaluation, and an absence of standardized benchmarks. To address this gap, we provide a comprehensive survey of HGNN-based anomaly detection methods in cybersecurity. We introduce a taxonomy that classifies approaches by anomaly type and graph dynamics, analyze representative models, and map them to key cybersecurity applications. We also review commonly used benchmark datasets and evaluation metrics, highlighting their strengths and limitations. Finally, we identify key open challenges related to modeling, data, and deployment, and outline promising directions for future research. This survey aims to establish a structured foundation for advancing HGNN-based anomaly detection toward scalable, interpretable, and practically deployable solutions.
Authors: Shih-Kai Chou, Mengran Zhao, Cheng-Nan Hu, Kuang-Chung Chou, Carolina Fortuna, Jernej Hribar
Abstract: Accurate Angle-of-arrival (AoA) estimation is essential for next-generation wireless communication systems to enable reliable beamforming, high-precision localization, and integrated sensing. Unfortunately, classical high-resolution techniques require multi-element arrays and extensive snapshot collection, while generic Machine Learning (ML) approaches often yield black-box models that lack physical interpretability. To address these limitations, we propose a Symbolic Regression (SR)-based ML framework. Namely, Symbolic Regression-based Angle of Arrival and Beam Pattern Estimator (SABER), a constrained symbolic-regression framework that automatically discovers closed-form beam pattern and AoA models from path loss measurements with interpretability. SABER achieves high accuracy while bridging the gap between opaque ML methods and interpretable physics-driven estimators. First, we validate our approach in a controlled free-space anechoic chamber, showing that both direct inversion of the known $\cos^n$ beam and a low-order polynomial surrogate achieve sub-0.5 degree Mean Absolute Error (MAE). A purely unconstrained SR method can further reduce the error of the predicted angles, but produces complex formulas that lack physical insight. Then, we implement the same SR-learned inversions in a real-world, Reconfigurable Intelligent Surface (RIS)-aided indoor testbed. SABER and unconstrained SR models accurately recover the true AoA with near-zero error. Finally, we benchmark SABER against the Cram\'er-Rao Lower Bounds (CRLBs). Our results demonstrate that SABER is an interpretable and accurate alternative to state-of-the-art and black-box ML-based methods for AoA estimation.
Authors: Mykhailo Poliakov, Nadiya Shvai
Abstract: Health-related misinformation is very prevalent and potentially harmful. It is difficult to identify, especially when claims distort or misinterpret scientific findings. We investigate the impact of synthetic data generation and lightweight fine-tuning techniques on the ability of large language models (LLMs) to recognize fallacious arguments using the MISSCI dataset and framework. In this work, we propose MisSynth, a pipeline that applies retrieval-augmented generation (RAG) to produce synthetic fallacy samples, which are then used to fine-tune an LLM model. Our results show substantial accuracy gains with fine-tuned models compared to vanilla baselines. For instance, the LLaMA 3.1 8B fine-tuned model achieved an over 35% F1-score absolute improvement on the MISSCI test split over its vanilla baseline. We demonstrate that introducing synthetic fallacy data to augment limited annotated resources can significantly enhance zero-shot LLM classification performance on real-world scientific misinformation tasks, even with limited computational resources. The code and synthetic dataset are available on https://github.com/mxpoliakov/MisSynth.
Authors: Andrew M. Bean, Nabeel Seedat, Shengzhuang Chen, Jonathan Richard Schwarz
Abstract: The prohibitive cost of evaluating large language models (LLMs) on comprehensive benchmarks necessitates the creation of small yet representative data subsets (i.e., tiny benchmarks) that enable efficient assessment while retaining predictive fidelity. Current methods for this task operate under a model-centric paradigm, selecting benchmarking items based on the collective performance of existing models. Such approaches are limited by large upfront costs, an inability to immediately handle new benchmarks (`cold-start'), and the fragile assumption that future models will share the failure patterns of their predecessors. In this work, we challenge this paradigm and propose a item-centric approach to benchmark subset selection, arguing that selection should be based on the intrinsic properties of the task items themselves, rather than on model-specific failure patterns. We instantiate this item-centric efficient benchmarking approach via a novel method, Scales++, where data selection is based on the cognitive demands of the benchmark samples. Empirically, we show Scales++ reduces the upfront selection cost by over 18x while achieving competitive predictive fidelity. On the Open LLM Leaderboard, using just a 0.5\% data subset, we predict full benchmark scores with a 2.9% mean absolute error. We demonstrate that this item-centric approach enables more efficient model evaluation without significant fidelity degradation, while also providing better cold-start performance and more interpretable benchmarking.
Authors: Joshua Rooijakkers, Leiv R{\o}nneberg, Fran\c{c}ois-Xavier Briol, Jeremias Knoblauch, Matias Altamirano
Abstract: Multi-output Gaussian process (MOGP) regression allows modelling dependencies among multiple correlated response variables. Similarly to standard Gaussian processes, MOGPs are sensitive to model misspecification and outliers, which can distort predictions within individual outputs. This situation can be further exacerbated by multiple anomalous response variables whose errors propagate due to correlations between outputs. To handle this situation, we extend and generalise the robust and conjugate Gaussian process (RCGP) framework introduced by Altamirano et al. (2024). This results in the multi-output RCGP (MO-RCGP): a provably robust MOGP that is conjugate, and jointly captures correlations across outputs. We thoroughly evaluate our approach through applications in finance and cancer research.
Authors: Vikrant Sahu, Gagan Raj Gupta, Raghav Borikar, Nitin Mane
Abstract: The rapid growth of programming education has outpaced traditional assessment tools, leaving faculty with limited means to provide meaningful, scalable feedback. Conventional autograders, while efficient, act as black-box systems that simply return pass/fail results, offering little insight into student thinking or learning needs. Autograder+ is designed to shift autograding from a purely summative process to a formative learning experience. It introduces two key capabilities: automated feedback generation using a fine-tuned Large Language Model, and visualization of student code submissions to uncover learning patterns. The model is fine-tuned on curated student code and expert feedback to ensure pedagogically aligned, context-aware guidance. In evaluation across 600 student submissions from multiple programming tasks, the system produced feedback with strong semantic alignment to instructor comments. For visualization, contrastively learned code embeddings trained on 1,000 annotated submissions enable grouping solutions into meaningful clusters based on functionality and approach. The system also supports prompt-pooling, allowing instructors to guide feedback style through selected prompt templates. By integrating AI-driven feedback, semantic clustering, and interactive visualization, Autograder+ reduces instructor workload while supporting targeted instruction and promoting stronger learning outcomes.
Authors: Danial Ebrat, Sepideh Ahmadian, Luis Rueda
Abstract: Recommender systems often struggle with data sparsity and cold-start scenarios, limiting their ability to provide accurate suggestions for new or infrequent users. This paper presents a Graph Attention Network (GAT) based Collaborative Filtering (CF) framework enhanced with Large Language Model (LLM) driven context aware embeddings. Specifically, we generate concise textual user profiles and unify item metadata (titles, genres, overviews) into rich textual embeddings, injecting these as initial node features in a bipartite user item graph. To further optimize ranking performance, we introduce a hybrid loss function that combines Bayesian Personalized Ranking (BPR) with a cosine similarity term and robust negative sampling, ensuring explicit negative feedback is distinguished from unobserved data. Experiments on the MovieLens 100k and 1M datasets show consistent improvements over state-of-the-art baselines in Precision, NDCG, and MAP while demonstrating robustness for users with limited interaction history. Ablation studies confirm the critical role of LLM-augmented embeddings and the cosine similarity term in capturing nuanced semantic relationships. Our approach effectively mitigates sparsity and cold-start limitations by integrating LLM-derived contextual understanding into graph-based architectures. Future directions include balancing recommendation accuracy with coverage and diversity, and introducing fairness-aware constraints and interpretability features to enhance system performance further.
Authors: Pei Peng, MingKun Xie, Hang Hao, Tong Jin, ShengJun Huang
Abstract: Object-context shortcuts remain a persistent challenge in vision-language models, undermining zero-shot reliability when test-time scenes differ from familiar training co-occurrences. We recast this issue as a causal inference problem and ask: Would the prediction remain if the object appeared in a different environment? To answer this at inference time, we estimate object and background expectations within CLIP's representation space, and synthesize counterfactual embeddings by recombining object features with diverse alternative contexts sampled from external datasets, batch neighbors, or text-derived descriptions. By estimating the Total Direct Effect and simulating intervention, we further subtract background-only activation, preserving beneficial object-context interactions while mitigating hallucinated scores. Without retraining or prompt design, our method substantially improves both worst-group and average accuracy on context-sensitive benchmarks, establishing a new zero-shot state of the art. Beyond performance, our framework provides a lightweight representation-level counterfactual approach, offering a practical causal avenue for debiased and reliable multimodal reasoning.
Authors: Xin Guo, Zhiheng Xi, Yiwen Ding, Yitao Zhai, Xiaowei Shi, Xunliang Cai, Tao Gui, Qi Zhang, Xuanjing Huang
Abstract: Self-improvement has emerged as a mainstream paradigm for advancing the reasoning capabilities of large vision-language models (LVLMs), where models explore and learn from successful trajectories iteratively. However, we identify a critical issue during this process: the model excels at generating high-quality trajectories for simple queries (i.e., head data) but struggles with more complex ones (i.e., tail data). This leads to an imbalanced optimization that drives the model to prioritize simple reasoning skills, while hindering its ability to tackle more complex reasoning tasks. Over iterations, this imbalance becomes increasingly pronounced--a dynamic we term the "Matthew effect"--which ultimately hinders further model improvement and leads to performance bottlenecks. To counteract this challenge, we introduce four efficient strategies from two perspectives: distribution-reshaping and trajectory-resampling, to achieve head-tail re-balancing during the exploration-and-learning self-improvement process. Extensive experiments on Qwen2-VL-7B-Instruct and InternVL2.5-4B models across visual reasoning tasks demonstrate that our methods consistently improve visual reasoning capabilities, outperforming vanilla self-improvement by 3.86 points on average.
Authors: Dipak Meher, Carlotta Domeniconi, Guadalupe Correa-Cabrera
Abstract: Human smuggling networks are complex and constantly evolving, making them difficult to analyze comprehensively. Legal case documents offer rich factual and procedural insights into these networks but are often long, unstructured, and filled with ambiguous or shifting references, posing significant challenges for automated knowledge graph (KG) construction. Existing methods either overlook coreference resolution or fail to scale beyond short text spans, leading to fragmented graphs and inconsistent entity linking. We propose LINK-KG, a modular framework that integrates a three-stage, LLM-guided coreference resolution pipeline with downstream KG extraction. At the core of our approach is a type-specific Prompt Cache, which consistently tracks and resolves references across document chunks, enabling clean and disambiguated narratives for structured knowledge graph construction from both short and long legal texts. LINK-KG reduces average node duplication by 45.21% and noisy nodes by 32.22% compared to baseline methods, resulting in cleaner and more coherent graph structures. These improvements establish LINK-KG as a strong foundation for analyzing complex criminal networks.
Authors: Dipak Meher, Carlotta Domeniconi
Abstract: Human smuggling networks are increasingly adaptive and difficult to analyze. Legal case documents offer critical insights but are often unstructured, lexically dense, and filled with ambiguous or shifting references, which pose significant challenges for automated knowledge graph (KG) construction. While recent LLM-based approaches improve over static templates, they still generate noisy, fragmented graphs with duplicate nodes due to the absence of guided extraction and coreference resolution. The recently proposed CORE-KG framework addresses these limitations by integrating a type-aware coreference module and domain-guided structured prompts, significantly reducing node duplication and legal noise. In this work, we present a systematic ablation study of CORE-KG to quantify the individual contributions of its two key components. Our results show that removing coreference resolution results in a 28.32% increase in node duplication and a 4.32% increase in noisy nodes, while removing structured prompts leads to a 4.34% increase in node duplication and a 73.33% increase in noisy nodes. These findings offer empirical insights for designing robust LLM-based pipelines for extracting structured representations from complex legal texts.
Authors: Miranda Anna Christ, Adri\'an Csisz\'arik, Gergely Becs\'o, D\'aniel Varga
Abstract: This paper investigates the structure of linear operators introduced in Hernandez et al. [2023] that decode specific relational facts in transformer language models. We extend their single-relation findings to a collection of relations and systematically chart their organization. We show that such collections of relation decoders can be highly compressed by simple order-3 tensor networks without significant loss in decoding accuracy. To explain this surprising redundancy, we develop a cross-evaluation protocol, in which we apply each linear decoder operator to the subjects of every other relation. Our results reveal that these linear maps do not encode distinct relations, but extract recurring, coarse-grained semantic properties (e.g., country of capital city and country of food are both in the country-of-X property). This property-centric structure clarifies both the operators' compressibility and highlights why they generalize only to new relations that are semantically close. Our findings thus interpret linear relational decoding in transformer language models as primarily property-based, rather than relation-specific.
Authors: Prathamesh Kothavale, Sravani Boddepalli
Abstract: Conventional robots possess a limited understanding of their kinematics and are confined to preprogrammed tasks, hindering their ability to leverage tools efficiently. Driven by the essential components of tool usage - grasping the desired outcome, selecting the most suitable tool, determining optimal tool orientation, and executing precise manipulations - we introduce a pioneering framework. Our novel approach expands the capabilities of the robot's inverse kinematics solver, empowering it to acquire a sequential repertoire of actions using tools of varying lengths. By integrating a simulation-learned action trajectory with the tool, we showcase the practicality of transferring acquired skills from simulation to real-world scenarios through comprehensive experimentation. Remarkably, our extended inverse kinematics solver demonstrates an impressive error rate of less than 1 cm. Furthermore, our trained policy achieves a mean error of 8 cm in simulation. Noteworthy, our model achieves virtually indistinguishable performance when employing two distinct tools of different lengths. This research provides an indication of potential advances in the exploration of all four fundamental aspects of tool usage, enabling robots to master the intricate art of tool manipulation across diverse tasks.
Authors: Yinrong Hong, Zhiquan Tan, Kai Hu
Abstract: Large Language Models (LLMs) face significant inference latency challenges stemming from their autoregressive design and large size. To address this, speculative decoding emerges as a solution, enabling the simultaneous generation and validation of multiple tokens. While recent approaches like EAGLE-2 and EAGLE-3 improve speculative decoding using dynamic tree structures, they often neglect the impact of crucial system variables such as GPU devices and batch sizes. Therefore, we introduce a new dynamic tree decoding approach called CAST that takes into account inference costs, including factors such as GPU configurations and batch sizes, to dynamically refine the tree structure. Through comprehensive experimentation across six diverse tasks and utilizing six distinct LLMs, our methodology demonstrates remarkable results, achieving speeds up to 5.2 times faster than conventional decoding methods. Moreover, it generally outperforms existing state-of-the-art techniques from 5% to 20%.
Authors: Sebastian Basterrech, Shuo Shan, Debabrata Adhikari, Sankhya Mohanty
Abstract: In this study, we leverage a mixture model learning approach to identify defects in laser-based Additive Manufacturing (AM) processes. By incorporating physics based principles, we also ensure that the model is sensitive to meaningful physical parameter variations. The empirical evaluation was conducted by analyzing real-world data from two AM processes: Directed Energy Deposition and Laser Powder Bed Fusion. In addition, we also studied the performance of the developed framework over public datasets with different alloy type and experimental parameter information. The results show the potential of physics-guided mixture models to examine the underlying physical behavior of an AM system.
Authors: Arne Thomsen, Tilman Tr\"oster, Fran\c{c}ois Lanusse
Abstract: Cosmological field-level inference requires differentiable forward models that solve the challenging dynamics of gas and dark matter under hydrodynamics and gravity. We propose a hybrid approach where gravitational forces are computed using a differentiable particle-mesh solver, while the hydrodynamics are parametrized by a neural network that maps local quantities to an effective pressure field. We demonstrate that our method improves upon alternative approaches, such as an Enthalpy Gradient Descent baseline, both at the field and summary-statistic level. The approach is furthermore highly data efficient, with a single reference simulation of cosmological structure formation being sufficient to constrain the neural pressure model. This opens the door for future applications where the model is fit directly to observational data, rather than a training set of simulations.
Authors: Shayan Nejadshamsi, Yuanyuan Zhang, Shadi Zaki, Brock Porth, Lysa Porth, Vahab Khoshdel
Abstract: Accurate and timely crop yield prediction is crucial for global food security and modern agricultural management. Traditional methods often lack the scalability and granularity required for precision farming. This paper introduces CYPRESS (Crop Yield Prediction via Regression on Prithvi's Encoder for Satellite Sensing), a deep learning model designed for high-resolution, intra-field canola yield prediction. CYPRESS leverages a pre-trained, large-scale geospatial foundation model (Prithvi-EO-2.0-600M) and adapts it for a continuous regression task, transforming multi-temporal satellite imagery into dense, pixel-level yield maps. Evaluated on a comprehensive dataset from the Canadian Prairies, CYPRESS demonstrates superior performance over existing deep learning-based yield prediction models, highlighting the effectiveness of fine-tuning foundation models for specialized agricultural applications. By providing a continuous, high-resolution output, CYPRESS offers a more actionable tool for precision agriculture than conventional classification or county-level aggregation methods. This work validates a novel approach that bridges the gap between large-scale Earth observation and on-farm decision-making, offering a scalable solution for detailed agricultural monitoring.
Authors: Xiaoyi He, Danggui Chen, Zhenshuo Zhang, Zimeng Bai
Abstract: This paper presents a hierarchical path-planning and control framework that combines a high-level Deep Q-Network (DQN) for discrete sub-goal selection with a low-level Twin Delayed Deep Deterministic Policy Gradient (TD3) controller for continuous actuation. The high-level module selects behaviors and sub-goals; the low-level module executes smooth velocity commands. We design a practical reward shaping scheme (direction, distance, obstacle avoidance, action smoothness, collision penalty, time penalty, and progress), together with a LiDAR-based safety gate that prevents unsafe motions. The system is implemented in ROS + Gazebo (TurtleBot3) and evaluated with PathBench metrics, including success rate, collision rate, path efficiency, and re-planning efficiency, in dynamic and partially observable environments. Experiments show improved success rate and sample efficiency over single-algorithm baselines (DQN or TD3 alone) and rule-based planners, with better generalization to unseen obstacle configurations and reduced abrupt control changes. Code and evaluation scripts are available at the project repository.
Authors: Georgios Kamaras, Craig Innes, Subramanian Ramamoorthy
Abstract: In robotics, likelihood-free inference (LFI) can provide the domain distribution that adapts a learnt agent in a parametric set of deployment conditions. LFI assumes an arbitrary support for sampling, which remains constant as the initial generic prior is iteratively refined to more descriptive posteriors. However, a potentially misspecified support can lead to suboptimal, yet falsely certain, posteriors. To address this issue, we propose three heuristic LFI variants: EDGE, MODE, and CENTRE. Each interprets the posterior mode shift over inference steps in its own way and, when integrated into an LFI step, adapts the support alongside posterior inference. We first expose the support misspecification issue and evaluate our heuristics using stochastic dynamical benchmarks. We then evaluate the impact of heuristic support adaptation on parameter inference and policy learning for a dynamic deformable linear object (DLO) manipulation task. Inference results in a finer length and stiffness classification for a parametric set of DLOs. When the resulting posteriors are used as domain distributions for sim-based policy learning, they lead to more robust object-centric agent performance.
Authors: Ruimin He, Shaowei Lin
Abstract: At the heart of reinforcement learning are actions - decisions made in response to observations of the environment. Actions are equally fundamental in the modeling of stochastic processes, as they trigger discontinuous state transitions and enable the flow of information through large, complex systems. In this paper, we unify the perspectives of stochastic processes and reinforcement learning through action- driven processes, and illustrate their application to spiking neural networks. Leveraging ideas from control-as-inference, we show that minimizing the Kullback-Leibler divergence between a policy-driven true distribution and a reward-driven model distribution for a suitably defined action-driven process is equivalent to maximum entropy reinforcement learning.
Authors: Jun Dai, Michael Rizvi-Martel, Guillaume Rabusseau
Abstract: Designing efficient quantum circuits is a central bottleneck to exploring the potential of quantum computing, particularly for noisy intermediate-scale quantum (NISQ) devices, where circuit efficiency and resilience to errors are paramount. The search space of gate sequences grows combinatorially, and handcrafted templates often waste scarce qubit and depth budgets. We introduce \textsc{FlowQ-Net} (Flow-based Quantum design Network), a generative framework for automated quantum circuit synthesis based on Generative Flow Networks (GFlowNets). This framework learns a stochastic policy to construct circuits sequentially, sampling them in proportion to a flexible, user-defined reward function that can encode multiple design objectives such as performance, depth, and gate count. This approach uniquely enables the generation of a diverse ensemble of high-quality circuits, moving beyond single-solution optimization. We demonstrate the efficacy of \textsc{FlowQ-Net} through an extensive set of simulations. We apply our method to Variational Quantum Algorithm (VQA) ansatz design for molecular ground state estimation, Max-Cut, and image classification, key challenges in near-term quantum computing. Circuits designed by \textsc{FlowQ-Net} achieve significant improvements, yielding circuits that are 10$\times$-30$\times$ more compact in terms of parameters, gates, and depth compared to commonly used unitary baselines, without compromising accuracy. This trend holds even when subjected to error profiles from real-world quantum devices. Our results underline the potential of generative models as a general-purpose methodology for automated quantum circuit design, offering a promising path towards more efficient quantum algorithms and accelerating scientific discovery in the quantum domain.
Authors: Kimi Team, Yu Zhang, Zongyu Lin, Xingcheng Yao, Jiaxi Hu, Fanqing Meng, Chengyin Liu, Xin Men, Songlin Yang, Zhiyuan Li, Wentao Li, Enzhe Lu, Weizhou Liu, Yanru Chen, Weixin Xu, Longhui Yu, Yejie Wang, Yu Fan, Longguang Zhong, Enming Yuan, Dehao Zhang, Yizhi Zhang, T. Y. Liu, Haiming Wang, Shengjun Fang, Weiran He, Shaowei Liu, Yiwei Li, Jianlin Su, Jiezhong Qiu, Bo Pang, Junjie Yan, Zhejun Jiang, Weixiao Huang, Bohong Yin, Jiacheng You, Chu Wei, Zhengtao Wang, Chao Hong, Yutian Chen, Guanduo Chen, Yucheng Wang, Huabin Zheng, Feng Wang, Yibo Liu, Mengnan Dong, Zheng Zhang, Siyuan Pan, Wenhao Wu, Yuhao Wu, Longyu Guan, Jiawen Tao, Guohong Fu, Xinran Xu, Yuzhi Wang, Guokun Lai, Yuxin Wu, Xinyu Zhou, Zhilin Yang, Yulun Du
Abstract: We introduce Kimi Linear, a hybrid linear attention architecture that, for the first time, outperforms full attention under fair comparisons across various scenarios -- including short-context, long-context, and reinforcement learning (RL) scaling regimes. At its core lies Kimi Delta Attention (KDA), an expressive linear attention module that extends Gated DeltaNet with a finer-grained gating mechanism, enabling more effective use of limited finite-state RNN memory. Our bespoke chunkwise algorithm achieves high hardware efficiency through a specialized variant of the Diagonal-Plus-Low-Rank (DPLR) transition matrices, which substantially reduces computation compared to the general DPLR formulation while remaining more consistent with the classical delta rule. We pretrain a Kimi Linear model with 3B activated parameters and 48B total parameters, based on a layerwise hybrid of KDA and Multi-Head Latent Attention (MLA). Our experiments show that with an identical training recipe, Kimi Linear outperforms full MLA with a sizeable margin across all evaluated tasks, while reducing KV cache usage by up to 75% and achieving up to 6 times decoding throughput for a 1M context. These results demonstrate that Kimi Linear can be a drop-in replacement for full attention architectures with superior performance and efficiency, including tasks with longer input and output lengths. To support further research, we open-source the KDA kernel and vLLM implementations, and release the pre-trained and instruction-tuned model checkpoints.
Authors: Gerard T. Portela, Jason B. Gibbons, Sebastian Schneeweiss, Rishi J. Desai
Abstract: Observational studies developing causal machine learning (ML) models for the prediction of individualized treatment effects (ITEs) seldom conduct empirical evaluations to assess the conditional exchangeability assumption. We aimed to evaluate the performance of these models under conditional exchangeability violations and the utility of negative control outcomes (NCOs) as a diagnostic. We conducted a simulation study to examine confounding bias in ITE estimates generated by causal forest and X-learner models under varying conditions, including the presence or absence of true heterogeneity. We simulated data to reflect real-world scenarios with differing levels of confounding, sample size, and NCO confounding structures. We then estimated and compared subgroup-level treatment effects on the primary outcome and NCOs across settings with and without unmeasured confounding. When conditional exchangeability was violated, causal forest and X-learner models failed to recover true treatment effect heterogeneity and, in some cases, falsely indicated heterogeneity when there was none. NCOs successfully identified subgroups affected by unmeasured confounding. Even when NCOs did not perfectly satisfy its ideal assumptions, it remained informative, flagging potential bias in subgroup level estimates, though not always pinpointing the subgroup with the largest confounding. Violations of conditional exchangeability substantially limit the validity of ITE estimates from causal ML models in routinely collected observational data. NCOs serve a useful empirical diagnostic tool for detecting subgroup-specific unmeasured confounding and should be incorporated into causal ML workflows to support the credibility of individualized inference.
Authors: Mehar Bhatia, Shravan Nayak, Gaurav Kamath, Marius Mosbach, Karolina Sta\'nczak, Vered Shwartz, Siva Reddy
Abstract: As LLMs occupy an increasingly important role in society, they are more and more confronted with questions that require them not only to draw on their general knowledge but also to align with certain human value systems. Therefore, studying the alignment of LLMs with human values has become a crucial field of inquiry. Prior work, however, mostly focuses on evaluating the alignment of fully trained models, overlooking the training dynamics by which models learn to express human values. In this work, we investigate how and at which stage value alignment arises during the course of a model's post-training. Our analysis disentangles the effects of post-training algorithms and datasets, measuring both the magnitude and time of value drifts during training. Experimenting with Llama-3 and Qwen-3 models of different sizes and popular supervised fine-tuning (SFT) and preference optimization datasets and algorithms, we find that the SFT phase generally establishes a model's values, and subsequent preference optimization rarely re-aligns these values. Furthermore, using a synthetic preference dataset that enables controlled manipulation of values, we find that different preference optimization algorithms lead to different value alignment outcomes, even when preference data is held constant. Our findings provide actionable insights into how values are learned during post-training and help to inform data curation, as well as the selection of models and algorithms for preference optimization to improve model alignment to human values.
Authors: Masahiro Kato
Abstract: The goal of policy learning is to train a policy function that recommends a treatment given covariates to maximize population welfare. There are two major approaches in policy learning: the empirical welfare maximization (EWM) approach and the plug-in approach. The EWM approach is analogous to a classification problem, where one first builds an estimator of the population welfare, which is a functional of policy functions, and then trains a policy by maximizing the estimated welfare. In contrast, the plug-in approach is based on regression, where one first estimates the conditional average treatment effect (CATE) and then recommends the treatment with the highest estimated outcome. This study bridges the gap between the two approaches by showing that both are based on essentially the same optimization problem. In particular, we prove an exact equivalence between EWM and least squares over a reparameterization of the policy class. As a consequence, the two approaches are interchangeable in several respects and share the same theoretical guarantees under common conditions. Leveraging this equivalence, we propose a novel regularization method for policy learning. Our findings yield a convex and computationally efficient training procedure that avoids the NP-hard combinatorial step typically required in EWM.
Authors: William Overman, Mohsen Bayati
Abstract: As increasingly capable agents are deployed, a central safety question is how to retain meaningful human control without modifying the underlying system. We study a minimal control interface where an agent chooses whether to act autonomously (play) or defer (ask), while a human simultaneously chooses whether to be permissive (trust) or to engage in oversight (oversee). If the agent defers, the human's choice determines the outcome, potentially leading to a corrective action or a system shutdown. We model this interaction as a two-player Markov Game. Our analysis focuses on cases where this game qualifies as a Markov Potential Game (MPG), a class of games where we can provide an alignment guarantee: under a structural assumption on the human's value function, any decision by the agent to act more autonomously that benefits itself cannot harm the human's value. We also analyze extensions to this MPG framework. Theoretically, this perspective provides conditions for a specific form of intrinsic alignment. If the reward structures of the human-agent game meet these conditions, we have a formal guarantee that the agent improving its own outcome will not harm the human's. Practically, this model motivates a transparent control layer with predictable incentives where the agent learns to defer when risky and act when safe, while its pretrained policy and the environment's reward structure remain untouched. Our gridworld simulation shows that through independent learning, the agent and human discover their optimal oversight roles. The agent learns to ask when uncertain and the human learns when to oversee, leading to an emergent collaboration that avoids safety violations introduced post-training. This demonstrates a practical method for making misaligned models safer after deployment.
Authors: Anushka Sivakumar, Andrew Zhang, Zaber Hakim, Chris Thomas
Abstract: This work introduces SteerVLM, a lightweight steering module designed to guide Vision-Language Models (VLMs) towards outputs that better adhere to desired instructions. Our approach learns from the latent embeddings of paired prompts encoding target and converse behaviors to dynamically adjust activations connecting the language modality with image context. This allows for fine-grained, inference-time control over complex output semantics without modifying model weights while preserving performance on off-target tasks. Our steering module requires learning parameters equal to 0.14% of the original VLM's size. Our steering module gains model control through dimension-wise activation modulation and adaptive steering across layers without requiring pre-extracted static vectors or manual tuning of intervention points. Furthermore, we introduce VNIA (Visual Narrative Intent Alignment), a multimodal dataset specifically created to facilitate the development and evaluation of VLM steering techniques. Our method outperforms existing intervention techniques on steering and hallucination mitigation benchmarks for VLMs and proposes a robust solution for multimodal model control through activation engineering.
Authors: Valentyna Starodub, Mantas Luko\v{s}evi\v{c}ius
Abstract: Age-related macular degeneration (AMD) is one of the leading causes of irreversible vision impairment in people over the age of 60. This research focuses on semantic segmentation for AMD lesion detection in RGB fundus images, a non-invasive and cost-effective imaging technique. The results of the ADAM challenge - the most comprehensive AMD detection from RGB fundus images research competition and open dataset to date - serve as a benchmark for our evaluation. Taking the U-Net connectivity as a base of our framework, we evaluate and compare several approaches to improve the segmentation model's architecture and training pipeline, including pre-processing techniques, encoder (backbone) deep network types of varying complexity, and specialized loss functions to mitigate class imbalances on image and pixel levels. The main outcome of this research is the final configuration of the AMD detection framework, which outperforms all the prior ADAM challenge submissions on the multi-class segmentation of different AMD lesion types in non-invasive RGB fundus images. The source code used to conduct the experiments presented in this paper is made freely available.
Authors: Masahiro Kato
Abstract: This note introduces a unified theory for causal inference that integrates Riesz regression, covariate balancing, density-ratio estimation (DRE), targeted maximum likelihood estimation (TMLE), and the matching estimator in average treatment effect (ATE) estimation. In ATE estimation, the balancing weights and the regression functions of the outcome play important roles, where the balancing weights are referred to as the Riesz representer, bias-correction term, and clever covariates, depending on the context. Riesz regression, covariate balancing, DRE, and the matching estimator are methods for estimating the balancing weights, where Riesz regression is essentially equivalent to DRE in the ATE context, the matching estimator is a special case of DRE, and DRE is in a dual relationship with covariate balancing. TMLE is a method for constructing regression function estimators such that the leading bias term becomes zero. Nearest Neighbor Matching is equivalent to Least Squares Density Ratio Estimation and Riesz Regression.
Authors: Cheng Zheng, William Koch, Baiang Li, Felix Heide
Abstract: Hierarchical structures of motion exist across research fields, including computer vision, graphics, and robotics, where complex dynamics typically arise from coordinated interactions among simpler motion components. Existing methods to model such dynamics typically rely on manually-defined or heuristic hierarchies with fixed motion primitives, limiting their generalizability across different tasks. In this work, we propose a general hierarchical motion modeling method that learns structured, interpretable motion relationships directly from data. Our method represents observed motions using graph-based hierarchies, explicitly decomposing global absolute motions into parent-inherited patterns and local motion residuals. We formulate hierarchy inference as a differentiable graph learning problem, where vertices represent elemental motions and directed edges capture learned parent-child dependencies through graph neural networks. We evaluate our hierarchical reconstruction approach on three examples: 1D translational motion, 2D rotational motion, and dynamic 3D scene deformation via Gaussian splatting. Experimental results show that our method reconstructs the intrinsic motion hierarchy in 1D and 2D cases, and produces more realistic and interpretable deformations compared to the baseline on dynamic 3D Gaussian splatting scenes. By providing an adaptable, data-driven hierarchical modeling paradigm, our method offers a formulation applicable to a broad range of motion-centric tasks. Project Page: https://light.princeton.edu/HEIR/
Authors: Philipp Lindenberger, Paul-Edouard Sarlin, Jan Hosang, Matteo Balice, Marc Pollefeys, Simon Lynen, Eduard Trulls
Abstract: Determining the precise geographic location of an image at a global scale remains an unsolved challenge. Standard image retrieval techniques are inefficient due to the sheer volume of images (>100M) and fail when coverage is insufficient. Scalable solutions, however, involve a trade-off: global classification typically yields coarse results (10+ kilometers), while cross-view retrieval between ground and aerial imagery suffers from a domain gap and has been primarily studied on smaller regions. This paper introduces a hybrid approach that achieves fine-grained geo-localization across a large geographic expanse the size of a continent. We leverage a proxy classification task during training to learn rich feature representations that implicitly encode precise location information. We combine these learned prototypes with embeddings of aerial imagery to increase robustness to the sparsity of ground-level data. This enables direct, fine-grained retrieval over areas spanning multiple countries. Our extensive evaluation demonstrates that our approach can localize within 200m more than 68\% of queries of a dataset covering a large part of Europe. The code is publicly available at https://scaling-geoloc.github.io.
Authors: Yukun Huang, Jiwen Yu, Yanning Zhou, Jianan Wang, Xintao Wang, Pengfei Wan, Xihui Liu
Abstract: There are two prevalent ways to constructing 3D scenes: procedural generation and 2D lifting. Among them, panorama-based 2D lifting has emerged as a promising technique, leveraging powerful 2D generative priors to produce immersive, realistic, and diverse 3D environments. In this work, we advance this technique to generate graphics-ready 3D scenes suitable for physically based rendering (PBR), relighting, and simulation. Our key insight is to repurpose 2D generative models for panoramic perception of geometry, textures, and PBR materials. Unlike existing 2D lifting approaches that emphasize appearance generation and ignore the perception of intrinsic properties, we present OmniX, a versatile and unified framework. Based on a lightweight and efficient cross-modal adapter structure, OmniX reuses 2D generative priors for a broad range of panoramic vision tasks, including panoramic perception, generation, and completion. Furthermore, we construct a large-scale synthetic panorama dataset containing high-quality multimodal panoramas from diverse indoor and outdoor scenes. Extensive experiments demonstrate the effectiveness of our model in panoramic visual perception and graphics-ready 3D scene generation, opening new possibilities for immersive and physically realistic virtual world generation.
Authors: Ziang Song, Tianle Cai, Jason D. Lee, Weijie J. Su
Abstract: The extraordinary capabilities of large language models (LLMs) such as ChatGPT and GPT-4 are in part unleashed by aligning them with reward models that are trained on human preferences, which are often represented as rankings of responses to prompts. In this paper, we document the phenomenon of \textit{reward collapse}, an empirical observation where the prevailing ranking-based approach results in an \textit{identical} reward distribution \textit{regardless} of the prompts during the terminal phase of training. This outcome is undesirable as open-ended prompts like ``write a short story about your best friend'' should yield a continuous range of rewards for their completions, while specific prompts like ``what is the capital of New Zealand'' should generate either high or low rewards. Our theoretical investigation reveals that reward collapse is primarily due to the insufficiency of the ranking-based objective function to incorporate prompt-related information during optimization. This insight allows us to derive closed-form expressions for the reward distribution associated with a set of utility functions in an asymptotic regime. To overcome reward collapse, we introduce a prompt-aware optimization scheme that provably admits a prompt-dependent reward distribution within the interpolating regime. Our experimental results suggest that our proposed prompt-aware utility functions significantly alleviate reward collapse during the training of reward models.
Authors: Sai Shankar Narasimhan, Shubhankar Agarwal, Oguzhan Akcin, Sujay Sanghavi, Sandeep Chinchali
Abstract: Imagine generating a city's electricity demand pattern based on weather, the presence of an electric vehicle, and location, which could be used for capacity planning during a winter freeze. Such real-world time series are often enriched with paired heterogeneous contextual metadata (e.g., weather and location). Current approaches to time series generation often ignore this paired metadata. Additionally, the heterogeneity in metadata poses several practical challenges in adapting existing conditional generation approaches from the image, audio, and video domains to the time series domain. To address this gap, we introduce TIME WEAVER, a novel diffusion-based model that leverages the heterogeneous metadata in the form of categorical, continuous, and even time-variant variables to significantly improve time series generation. Additionally, we show that naive extensions of standard evaluation metrics from the image to the time series domain are insufficient. These metrics do not penalize conditional generation approaches for their poor specificity in reproducing the metadata-specific features in the generated time series. Thus, we innovate a novel evaluation metric that accurately captures the specificity of conditional generation and the realism of the generated time series. We show that TIME WEAVER outperforms state-of-the-art benchmarks, such as Generative Adversarial Networks (GANs), by up to 30% in downstream classification tasks on real-world energy, medical, air quality, and traffic datasets.
Authors: Toshitaka Matsuki, Yusuke Sakemi, Kazuyuki Aihara
Abstract: Chaos-based reinforcement learning (CBRL) is a method in which the agent's internal chaotic dynamics drives exploration. However, the learning algorithms in CBRL have not been thoroughly developed in previous studies, nor have they incorporated recent advances in reinforcement learning. This study introduced Twin Delayed Deep Deterministic Policy Gradients (TD3), which is one of the state-of-the-art deep reinforcement learning algorithms that can treat deterministic and continuous action spaces, to CBRL. The validation results provide several insights. First, TD3 works as a learning algorithm for CBRL in a simple goal-reaching task. Second, CBRL agents with TD3 can autonomously suppress their exploratory behavior as learning progresses and resume exploration when the environment changes. Finally, examining the effect of the agent's chaoticity on learning shows that there exists a suitable range of chaos strength in the agent's model to flexibly switch between exploration and exploitation and adapt to environmental changes.
Authors: Alejandro Polo-Molina, David Alfaya, Jose Portela
Abstract: Artificial Neural Networks (ANNs) have become a powerful tool for modeling complex relationships in large-scale datasets. However, their black-box nature poses trustworthiness challenges. In certain situations, ensuring trust in predictions might require following specific partial monotonicity constraints. However, certifying if an already-trained ANN is partially monotonic is challenging. Therefore, ANNs are often disregarded in some critical applications, such as credit scoring, where partial monotonicity is required. To address this challenge, this paper presents a novel algorithm (LipVor) that certifies if a black-box model, such as an ANN, is positive based on a finite number of evaluations. Consequently, since partial monotonicity can be expressed as a positivity condition on partial derivatives, LipVor can certify whether an ANN is partially monotonic. To do so, for every positively evaluated point, the Lipschitzianity of the black-box model is used to construct a specific neighborhood where the function remains positive. Next, based on the Voronoi diagram of the evaluated points, a sufficient condition is stated to certify if the function is positive in the domain. Unlike prior methods, our approach certifies partial monotonicity without constrained architectures or piece-wise linear activations. Therefore, LipVor could open up the possibility of using unconstrained ANN in some critical fields. Moreover, some other properties of an ANN, such as convexity, can be posed as positivity conditions, and therefore, LipVor could also be applied.
Authors: Xiao Liu, Mingyuan Li, Guangsheng Yu, Lixiang Li, Haipeng Peng, Ren Ping Liu
Abstract: Unlearning is challenging in generic learning frameworks with the continuous growth and updates of models exhibiting complex inheritance relationships. This paper presents a novel unlearning framework that enables fully parallel unlearning among models exhibiting inheritance. We use a chronologically Directed Acyclic Graph (DAG) to capture various unlearning scenarios occurring in model inheritance networks. Central to our framework is the Fisher Inheritance Unlearning (FIUn) method, designed to enable efficient parallel unlearning within the DAG. FIUn utilizes the Fisher Information Matrix (FIM) to assess the significance of model parameters for unlearning tasks and adjusts them accordingly. To handle multiple unlearning requests simultaneously, we propose the Merging-FIM (MFIM) function, which consolidates FIMs from multiple upstream models into a unified matrix. This design supports all unlearning scenarios captured by the DAG, enabling one-shot removal of inherited knowledge while significantly reducing computational overhead. Experiments confirm the effectiveness of our unlearning framework. For single-class tasks, it achieves complete unlearning with 0% accuracy for unlearned labels while maintaining 94.53% accuracy for retained labels. For multi-class tasks, the accuracy is 1.07% for unlearned labels and 84.77% for retained labels. Our framework accelerates unlearning by 99% compared to alternative methods. Code is in https://github.com/MJLee00/Parallel-Unlearning-in-Inherited-Model-Networks.
URLs: https://github.com/MJLee00/Parallel-Unlearning-in-Inherited-Model-Networks.
Authors: Aaditya Naik, Jason Liu, Claire Wang, Amish Sethi, Saikat Dutta, Mayur Naik, Eric Wong
Abstract: Neurosymbolic learning enables the integration of symbolic reasoning with deep learning but faces significant challenges in scaling to complex symbolic programs, large datasets, or both. We introduce DOLPHIN, a framework that tackles these challenges by supporting neurosymbolic programs in Python, executing complex symbolic reasoning on the CPU while vectorizing probabilistic computations and gradient propagation on the GPU. Across 13 benchmarks spanning tasks over text, image, and video data, with symbolic reasoning features like recursion and black-box functions, DOLPHIN converges to state-of-the-art accuracies on the more complex benchmarks while existing frameworks such as Scallop, ISED, and IndeCateR+ fail to converge within the time limit. On simpler benchmarks, DOLPHIN matches their performance, while achieving these results 1.71x to 62x faster than the baselines. Overall, DOLPHIN advances the scalability of neurosymbolic frameworks, achieving state-of-the-art efficiency and convergence on difficult benchmarks where existing frameworks struggle. The code is published at https://github.com/Dolphin-NeSy/Dolphin.
Authors: Connor W. Magoon, Fengyu Yang, Noam Aigerman, Shahar Z. Kovalsky
Abstract: Differentiable optimization has attracted significant research interest, particularly for quadratic programming (QP). Existing approaches for differentiating the solution of a QP with respect to its defining parameters often rely on specific integrated solvers. This integration limits their applicability, including their use in neural network architectures and bi-level optimization tasks, restricting users to a narrow selection of solver choices. To address this limitation, we introduce dQP, a modular and solver-agnostic framework for plug-and-play differentiation of virtually any QP solver. A key insight we leverage to achieve modularity is that, once the active set of inequality constraints is known, both the solution and its derivative can be expressed using simplified linear systems that share the same matrix. This formulation fully decouples the computation of the QP solution from its differentiation. Building on this result, we provide a minimal-overhead, open-source implementation ( https://github.com/cwmagoon/dQP ) that seamlessly integrates with over 15 state-of-the-art solvers. Comprehensive benchmark experiments demonstrate dQP's robustness and scalability, particularly highlighting its advantages in large-scale sparse problems.
Authors: Bowei Zhu, Shaojie Li, Mingyang Yi, Yong Liu
Abstract: Prior work (Klochkov $\&$ Zhivotovskiy, 2021) establishes at most $O\left(\log (n)/n\right)$ excess risk bounds via algorithmic stability for strongly-convex learners with high probability. We show that under the similar common assumptions -- - Polyak-Lojasiewicz condition, smoothness, and Lipschitz continous for losses -- - rates of $O\left(\log^2(n)/n^2\right)$ are at most achievable. To our knowledge, our analysis also provides the tightest high-probability bounds for gradient-based generalization gaps in nonconvex settings.
Authors: Sai Shankar Narasimhan, Shubhankar Agarwal, Litu Rout, Sanjay Shakkottai, Sandeep P. Chinchali
Abstract: Generating realistic time series samples is crucial for stress-testing models and protecting user privacy by using synthetic data. In engineering and safety-critical applications, these samples must meet certain hard constraints that are domain-specific or naturally imposed by physics or nature. Consider, for example, generating electricity demand patterns with constraints on peak demand times. This can be used to stress-test the functioning of power grids during adverse weather conditions. Existing approaches for generating constrained time series are either not scalable or degrade sample quality. To address these challenges, we introduce Constrained Posterior Sampling (CPS), a diffusion-based sampling algorithm that aims to project the posterior mean estimate into the constraint set after each denoising update. Notably, CPS scales to a large number of constraints ($\sim100$) without requiring additional training. We provide theoretical justifications highlighting the impact of our projection step on sampling. Empirically, CPS outperforms state-of-the-art methods in sample quality and similarity to real time series by around 70\% and 22\%, respectively, on real-world stocks, traffic, and air quality datasets.
Authors: Moshe Kimhi, Idan Kashani, Avi Mendelson, Chaim Baskin
Abstract: The widely used ReLU is favored for its hardware efficiency, {as the implementation at inference is a one bit sign case,} yet suffers from issues such as the ``dying ReLU'' problem, where during training, neurons fail to activate and constantly remain at zero, as highlighted by Lu et al. Traditional approaches to mitigate this issue often introduce more complex and less hardware-friendly activation functions. In this work, we propose a Hysteresis Rectified Linear Unit (HeLU), an efficient activation function designed to address the ``dying ReLU'' problem with minimal complexity. Unlike traditional activation functions with fixed thresholds for training and inference, HeLU employs a variable threshold that refines the backpropagation. This refined mechanism allows simpler activation functions to achieve competitive performance comparable to their more complex counterparts without introducing unnecessary complexity or requiring inductive biases. Empirical evaluations demonstrate that HeLU enhances model generalization across diverse datasets, offering a promising solution for efficient and effective inference suitable for a wide range of neural network architectures.
Authors: Thomas Bailie, Yun Sing Koh, Karthik Mukkavilli
Abstract: Graphs model latent variable relationships in many real-world systems, and Message Passing Neural Networks (MPNNs) are widely used to learn such structures for downstream tasks. While edge-based MPNNs effectively capture local interactions, their expressive power is theoretically bounded, limiting the discovery of higher-order relationships. We introduce the Higher-Order Graph Attention (HoGA) module, which constructs a k-order attention matrix by sampling subgraphs to maximize diversity among feature vectors. Unlike existing higher-order attention methods that greedily resample similar k-order relationships, HoGA targets diverse modalities in higher-order topology, reducing redundancy and expanding the range of captured substructures. Applied to two single-hop attention models, HoGA achieves at least a 5% accuracy gain on all benchmark node classification datasets and outperforms recent baselines on six of eight datasets. Code is available at https://github.com/TB862/Higher_Order.
Authors: Nick Oh
Abstract: This position paper defends post-hoc explainability methods as legitimate tools for scientific knowledge production in machine learning. Addressing criticism of these methods' reliability and epistemic status, we develop a philosophical framework grounded in mediated understanding and bounded factivity. We argue that scientific insights can emerge through structured interpretation of model behaviour without requiring complete mechanistic transparency, provided explanations acknowledge their approximative nature and undergo rigorous empirical validation. Through analysis of recent biomedical ML applications, we demonstrate how post-hoc methods, when properly integrated into scientific practice, generate novel hypotheses and advance phenomenal understanding.
Authors: Chengxin Hu, Hao Li, Yihe Yuan, Zezheng Song, Chenyang Zhao, Haixin Wang
Abstract: In the molecular domain, numerous studies have explored the use of multimodal large language models (LLMs) to construct a general-purpose, multi-task molecular model. However, these efforts are still far from achieving a truly universal molecular model. We identify three key challenges in this endeavor: (1) Existing molecular task datasets are typically small in scale and lack comprehensive domain coverage. (2) Tasks from different molecular subfields are difficult to effectively learn jointly through LLMs due to significant distributional shifts and competition among tasks, which introduces instability in the learning process. (3) Both inter-task and intra-task molecular representations demand different intrinsic dimensions in the language space, making it challenging to balance between redundancy and insufficiency in language model representations. To address these challenges, we innovatively categorize existing small-molecule tasks into four types: Mol2Mol, Mol2Text, Mol2Num, and Text2Mol. We then collect a dataset encompassing over 16 tasks with more than 1.4 million samples, making it the largest molecular instruction-tuning dataset to date. Leveraging the extensive pretraining of LLMs on existing chemical literature, we propose a novel multimodal LLM framework, named Omni-Mol, which unifies all small-molecule tasks and supports both molecular generation and understanding. The core of Omni-Mol is our proposed MoGE, which dynamically adapts to the intrinsic rank of different tasks. This mixture-of-experts architecture enhances the model's ability to handle diverse tasks and modalities effectively. Our model achieves unified instruction tuning across 16 tasks and attains state-of-the-art performance on 13 of them. Extensive experiments further demonstrate the scalability and versatility of Omni-Mol.
Authors: Ishika Agarwal, Dilek Hakkani-T\"ur
Abstract: Influence functions provide crucial insights into model training, but existing methods suffer from large computational costs and limited generalization. Particularly, recent works have proposed various metrics and algorithms to calculate the influence of data using language models, which do not scale well with large models and datasets. This is because of the expensive forward and backward passes required for computation, substantial memory requirements to store large models, and poor generalization of influence estimates to new data. In this paper, we explore the use of small neural networks -- which we refer to as the InfluenceNetwork -- to estimate influence values, achieving up to 99% cost reduction. Our evaluation demonstrates that influence values can be estimated with models just 0.0027% the size of full language models (we use 7B and 8B versions). We apply our algorithm of estimating influence values (called NN-CIFT: Neural Networks for effiCient Instruction Fine-Tuning) to the downstream task of subset selection for general instruction fine-tuning. In our study, we include four state-of-the-art influence functions and show no compromise in performance, despite large speedups, between NN-CIFT and the original influence functions. We provide an in-depth hyperparameter analyses of NN-CIFT. The code for our method can be found here: https://github.com/agarwalishika/NN-CIFT.
Authors: Yongli Yan, Linglong Dai
Abstract: Neural network-based decoding methods show promise in enhancing error correction performance but face challenges with punctured codes. In particular, existing methods struggle to adapt to variable code rates or meet protocol compatibility requirements. This paper proposes a unified long short-term memory (LSTM)-based neural decoder for punctured convolutional and Turbo codes to address these challenges. The key component of the proposed LSTM-based neural decoder is puncturing-aware embedding, which integrates puncturing patterns directly into the neural network to enable seamless adaptation to different code rates. Moreover, a balanced bit error rate training strategy is designed to ensure the decoder's robustness across various code lengths, rates, and channels. In this way, the protocol compatibility requirement can be realized. Extensive simulations in both additive white Gaussian noise (AWGN) and Rayleigh fading channels demonstrate that the proposed neural decoder outperforms conventional decoding techniques, offering significant improvements in decoding accuracy and robustness.
Authors: Ethan Mendes, Alan Ritter
Abstract: Collecting ground-truth rewards or human demonstrations for multi-step reasoning tasks is often prohibitively expensive, particularly in interactive domains such as web tasks. We introduce Self-Taught Lookahead (STL), a reward-free framework that improves language model-based value functions by reasoning explicitly about state transitions. STL can be viewed as a chain-of-thought analogue of the value iteration algorithm: instead of regressing directly on numeric values, a value LLM is trained to simulate a step of lookahead in natural language - predicting the next action, resulting state, and rationale for its value, thereby refining value estimates without any labeled data. This self-supervised procedure yields more accurate state-value predictions, which in turn enable lightweight search algorithms to expand fewer states while maintaining strong performance. Empirically, STL-trained value models built on moderately sized (8B parameter) open-weight LLMs boost web agent success rates by 39%, achieving comparable performance with proprietary models. STL also generalizes to multi-hop QA and math puzzles. We find that STL enables small open-source models to guide efficient search, reducing inference costs by integrating explicit reasoning with value learning.
Authors: Sabino Francesco Roselli, Eibe Frank
Abstract: Model trees provide an appealing way to perform interpretable machine learning for both classification and regression problems. In contrast to ``classic'' decision trees with constant values in their leaves, model trees can use linear combinations of predictor variables in their leaf nodes to form predictions, which can help achieve higher accuracy and smaller trees. Typical algorithms for learning model trees from training data work in a greedy fashion, growing the tree in a top-down manner by recursively splitting the data into smaller and smaller subsets. Crucially, the selected splits are only locally optimal, potentially rendering the tree overly complex and less accurate than a tree whose structure is globally optimal for the training data. In this paper, we empirically investigate the effect of constructing globally optimal model trees for classification and regression with linear support vector machines at the leaf nodes. To this end, we present mixed-integer linear programming formulations to learn optimal trees, compute such trees for a large collection of benchmark data sets, and compare their performance against greedily grown model trees in terms of interpretability and accuracy. We also compare to classic optimal and greedily grown decision trees, random forests, and support vector machines. Our results show that optimal model trees can achieve competitive accuracy with very small trees. We also investigate the effect on the accuracy of replacing axis-parallel splits with multivariate ones, foregoing interpretability while potentially obtaining greater accuracy.
Authors: Junyi Zhu, Ruicong Yao, Taha Ceritli, Savas Ozkan, Matthew B. Blaschko, Eunchung Noh, Jeongwon Min, Cho Jung Min, Mete Ozay
Abstract: Current network training paradigms primarily focus on either centralized or decentralized data regimes. However, in practice, data availability often exhibits a hybrid nature, where both regimes coexist. This hybrid setting presents new opportunities for model training, as the two regimes offer complementary trade-offs: decentralized data is abundant but subject to heterogeneity and communication constraints, while centralized data, though limited in volume and potentially unrepresentative, enables better curation and high-throughput access. Despite its potential, effectively combining these paradigms remains challenging, and few frameworks are tailored to hybrid data regimes. To address this, we propose a novel framework that constructs a model atlas from decentralized models and leverages centralized data to refine a global model within this structured space. The refined model is then used to reinitialize the decentralized models. Our method synergizes federated learning (to exploit decentralized data) and model merging (to utilize centralized data), enabling effective training under hybrid data availability. Theoretically, we show that our approach achieves faster convergence than methods relying solely on decentralized data, due to variance reduction in the merging process. Extensive experiments demonstrate that our framework consistently outperforms purely centralized, purely decentralized, and existing hybrid-adaptable methods. Notably, our method remains robust even when the centralized and decentralized data domains differ or when decentralized data contains noise, significantly broadening its applicability.
Authors: Ryan Franks, Alexey Miroshnikov, Konstandinos Kotsiopoulos
Abstract: We develop a novel bias mitigation framework with distribution-based fairness constraints suitable for producing demographically blind and explainable machine-learning models across a wide range of fairness levels. This is accomplished through post-processing, allowing fairer models to be generated efficiently without retraining the underlying model. Our framework, which is based on stochastic gradient descent, can be applied to a wide range of model types, with a particular emphasis on the post-processing of gradient-boosted decision trees. Additionally, we design a broad family of global fairness metrics, along with differentiable and consistent estimators compatible with our framework, building on previous work. We empirically test our methodology on a variety of datasets and compare it with alternative post-processing approaches, including Bayesian search, optimal transport projection, and direct neural network training.
Authors: J. S. van Hulst, W. P. M. H. Heemels, D. J. Antunes
Abstract: Reinforcement learning (RL) is a powerful framework for decision-making in uncertain environments, but it often requires large amounts of data to learn an optimal policy. We address this challenge by incorporating prior model knowledge to guide exploration and accelerate the learning process. Specifically, we assume access to a model set that contains the true transition kernel and reward function. We optimize over this model set to obtain upper and lower bounds on the Q-function, which are then used to guide the exploration of the agent. We provide theoretical guarantees on the convergence of the Q-function to the optimal Q-function under the proposed class of exploring policies. Furthermore, we also introduce a data-driven regularized version of the model set optimization problem that ensures the convergence of the class of exploring policies to the optimal policy. Lastly, we show that when the model set has a specific structure, namely the bounded-parameter MDP (BMDP) framework, the regularized model set optimization problem becomes convex and simple to implement. In this setting, we also prove finite-time convergence to the optimal policy under mild assumptions. We demonstrate the effectiveness of the proposed exploration strategy, which we call BUMEX (Bounded Uncertainty Model-based Exploration), in a simulation study. The results indicate that the proposed method can significantly accelerate learning in benchmark examples. A toolbox is available at https://github.com/JvHulst/BUMEX.
Authors: Zhaiming Shen, Sung Ha Kang
Abstract: Local clustering aims to identify specific substructures within a large graph without any additional structural information of the graph. These substructures are typically small compared to the overall graph, enabling the problem to be approached by finding a sparse solution to a linear system associated with the graph Laplacian. In this work, we first propose a method for identifying specific local clusters when very few labeled data are given, which we term semi-supervised local clustering. We then extend this approach to the unsupervised setting when no prior information on labels is available. The proposed methods involve randomly sampling the graph, applying diffusion through local cluster extraction, then examining the overlap among the results to find each cluster. We establish the co-membership conditions for any pair of nodes, and rigorously prove the correctness of our methods. Additionally, we conduct extensive experiments to demonstrate that the proposed methods achieve state of the art results in the low-label rates regime.
Authors: Pablo G\'omez, Laslo E. Ruhberg, Maria Teresa Nardone, David O'Ryan
Abstract: Anomaly detection in large datasets is essential in astronomy and computer vision. However, due to a scarcity of labelled data, it is often infeasible to apply supervised methods to anomaly detection. We present AnomalyMatch, an anomaly detection framework combining the semi-supervised FixMatch algorithm using EfficientNet classifiers with active learning. AnomalyMatch is tailored for large-scale applications and integrated into the ESA Datalabs science platform. In this method, we treat anomaly detection as a binary classification problem and efficiently utilise limited labelled and abundant unlabelled images for training. We enable active learning via a user interface for verification of high-confidence anomalies and correction of false positives. Evaluations on the GalaxyMNIST astronomical dataset and the miniImageNet natural-image benchmark under severe class imbalance display strong performance. Starting from five to ten labelled anomalies, we achieve an average AUROC of 0.96 (miniImageNet) and 0.89 (GalaxyMNIST), with respective AUPRC of 0.82 and 0.77. After three active learning cycles, anomalies are ranked with 76% (miniImageNet) to 94% (GalaxyMNIST) precision in the top 1% of the highest-ranking images by score. We compare to the established Astronomaly software on selected 'odd' galaxies from the 'Galaxy Zoo - The Galaxy Challenge' dataset, achieving comparable performance with an average AUROC of 0.83. Our results underscore the exceptional utility and scalability of this approach for anomaly discovery, highlighting the value of specialised approaches for domains characterised by severe label scarcity.
Authors: Hiroki Hasegawa, Yukihiko Okada
Abstract: Recent advances in IoT and biometric sensing technologies have led to the generation of massive and high-dimensional tensor data, yet achieving accurate and efficient low-rank approximation remains a major challenge. Existing tensor decomposition methods typically require prior specification of the tensor rank and rely on iterative optimization, which often results in heavy computational costs and dependence on the analyst's expertise. In this study, we propose a novel low-rank approximation method for tensor data that requires neither prior rank specification nor iterative optimization. The proposed method performs statistical singular value hard thresholding on the mode-wise unfolded matrices to automatically extract only statistically significant components, thereby achieving noise reduction while preserving the intrinsic tensor structure. Theoretically, the optimal threshold for each mode is derived based on the asymptotic properties of the Mar\v{c}enko--Pastur distribution. Simulation experiments demonstrate that the proposed method outperforms conventional approaches such as Higher-Order Singular Value Decomposition, Higher-Order Orthogonal Iteration, and Tucker-L2E in terms of both estimation accuracy and computational efficiency. These results indicate that our method provides an effective and theoretically grounded framework for automatic, non-iterative, and analyst-independent tensor decomposition.
Authors: Zichen Liu, Wei Zhang, Tiejun Li
Abstract: Euclidean diffusion models have achieved remarkable success in generative modeling across diverse domains, and they have been extended to manifold cases in recent advances. Instead of explicitly utilizing the structure of special manifolds as studied in previous works, in this paper we investigate direct sampling of the Euclidean diffusion models for general manifold-structured data. We reveal the multiscale singularity of the score function in the ambient space, which hinders the accuracy of diffusion-generated samples. We then present an elaborate theoretical analysis of the singularity structure of the score function by decomposing it along the tangential and normal directions of the manifold. To mitigate the singularity and improve the sampling accuracy, we propose two novel methods: (1) Niso-DM, which reduces the scale discrepancies in the score function by utilizing a non-isotropic noise, and (2) Tango-DM, which trains only the tangential component of the score function using a tangential-only loss function. Numerical experiments demonstrate that our methods achieve superior performance on distributions over various manifolds with complex geometries.
Authors: Xiaotian Zhang, Yue Shang, Entao Yang, Ge Zhang
Abstract: Understanding neural network's (NN) generalizability remains a central question in deep learning research. The special phenomenon of grokking, where NNs abruptly generalize long after the training performance reaches a near-perfect level, offers a unique window to investigate the underlying mechanisms of NNs' generalizability. Here we propose an interpretation for grokking by framing it as a computational glass relaxation: viewing NNs as a physical system where parameters are the degrees of freedom and train loss is the system energy, we find memorization process resembles a rapid cooling of liquid into non-equilibrium glassy state at low temperature and the later generalization is like a slow relaxation towards a more stable configuration. This mapping enables us to sample NNs' Boltzmann entropy (states of density) landscape as a function of training loss and test accuracy. Our experiments in transformers on arithmetic tasks suggests that there is NO entropy barrier in the memorization-to-generalization transition of grokking, challenging previous theory that defines grokking as a first-order phase transition. We identify a high-entropy advantage under grokking, an extension of prior work linking entropy to generalizability but much more significant. Inspired by grokking's far-from-equilibrium nature, we develop a toy optimizer WanD based on Wang-landau molecular dynamics, which can eliminate grokking without any constraints and find high-norm generalizing solutions. This provides strictly-defined counterexamples to theory attributing grokking solely to weight norm evolution towards the Goldilocks zone and also suggests new potential ways for optimizer design.
Authors: Wen-Chao Hu, Qi-Jie Li, Lin-Han Jia, Cunjing Ge, Yu-Feng Li, Yuan Jiang, Zhi-Hua Zhou
Abstract: Abductive Learning (ABL) integrates machine learning with logical reasoning in a loop: a learning model predicts symbolic concept labels from raw inputs, which are revised through abduction using domain knowledge and then fed back for retraining. However, due to the nondeterminism of abduction, the training process often suffers from instability, especially when the knowledge base is large and complex, resulting in a prohibitively large abduction space. While prior works focus on improving candidate selection within this space, they typically treat the knowledge base as a static black box. In this work, we propose Curriculum Abductive Learning (C-ABL), a method that explicitly leverages the internal structure of the knowledge base to address the ABL training challenges. C-ABL partitions the knowledge base into a sequence of sub-bases, progressively introduced during training. This reduces the abduction space throughout training and enables the model to incorporate logic in a stepwise, smooth way. Experiments across multiple tasks show that C-ABL outperforms previous ABL implementations, significantly improves training stability, convergence speed, and final accuracy, especially under complex knowledge setting.
Authors: Emile van Krieken, Pasquale Minervini, Edoardo Ponti, Antonio Vergari
Abstract: Neurosymbolic (NeSy) predictors combine neural perception with symbolic reasoning to solve tasks like visual reasoning. However, standard NeSy predictors assume conditional independence between the symbols they extract, thus limiting their ability to model interactions and uncertainty - often leading to overconfident predictions and poor out-of-distribution generalisation. To overcome the limitations of the independence assumption, we introduce neurosymbolic diffusion models (NeSyDMs), a new class of NeSy predictors that use discrete diffusion to model dependencies between symbols. Our approach reuses the independence assumption from NeSy predictors at each step of the diffusion process, enabling scalable learning while capturing symbol dependencies and uncertainty quantification. Across both synthetic and real-world benchmarks - including high-dimensional visual path planning and rule-based autonomous driving - NeSyDMs achieve state-of-the-art accuracy among NeSy predictors and demonstrate strong calibration.
Authors: Hengli Li, Chenxi Li, Tong Wu, Xuekai Zhu, Yuxuan Wang, Zhaoxin Yu, Eric Hanchen Jiang, Song-Chun Zhu, Zixia Jia, Ying Nian Wu, Zilong Zheng
Abstract: Reasoning ability, a core component of human intelligence, continues to pose a significant challenge for Large Language Models (LLMs) in the pursuit of AGI. Although model performance has improved under the training scaling law, significant challenges remain, particularly with respect to training algorithms, such as catastrophic forgetting, and the limited availability of novel training data. As an alternative, test-time scaling enhances reasoning performance by increasing test-time computation without parameter updating. Unlike prior methods in this paradigm focused on token space, we propose leveraging latent space for more effective reasoning and better adherence to the test-time scaling law. We introduce LatentSeek, a novel framework that enhances LLM reasoning through Test-Time Instance-level Adaptation (TTIA) within the model's latent space. Specifically, LatentSeek leverages policy gradient to iteratively update latent representations, guided by self-generated reward signals. LatentSeek is evaluated on a range of reasoning benchmarks, including GSM8K, MATH-500, and AIME2024, across multiple LLM architectures. Results show that LatentSeek consistently outperforms strong baselines, such as Chain-of-Thought prompting and fine-tuning-based methods. Furthermore, our analysis demonstrates that LatentSeek is highly efficient, typically converging within a few iterations for problems of average complexity, while also benefiting from additional iterations, thereby highlighting the potential of test-time scaling in the latent space. These findings position LatentSeek as a lightweight, scalable, and effective solution for enhancing the reasoning capabilities of LLMs.
Authors: Fu Luo, Xi Lin, Mengyuan Zhong, Fei Liu, Zhenkun Wang, Jianyong Sun, Qingfu Zhang
Abstract: Neural Combinatorial Optimisation (NCO) is a promising learning-based approach for solving Vehicle Routing Problems (VRPs) without extensive manual design. While existing constructive NCO methods typically follow an appending-based paradigm that sequentially adds unvisited nodes to partial solutions, this rigid approach often leads to suboptimal results. To overcome this limitation, we explore the idea of insertion-based paradigm and propose Learning to Construct with Insertion-based Paradigm (L2C-Insert), a novel learning-based method for constructive NCO. Unlike traditional approaches, L2C-Insert builds solutions by strategically inserting unvisited nodes at any valid position in the current partial solution, which can significantly enhance the flexibility and solution quality. The proposed framework introduces three key components: a novel model architecture for precise insertion position prediction, an efficient training scheme for model optimization, and an advanced inference technique that fully exploits the insertion paradigm's flexibility. Extensive experiments on both synthetic and real-world instances of the Travelling Salesman Problem (TSP) and Capacitated Vehicle Routing Problem (CVRP) demonstrate that L2C-Insert consistently achieves superior performance across various problem sizes.
Authors: Christian Walder, Deep Karkhanis
Abstract: Reinforcement Learning (RL) algorithms sample multiple n>1 solution attempts for each problem and reward them independently. This optimizes for pass@1 performance and prioritizes the strength of isolated samples at the expense of the diversity and collective utility of sets of samples. This under-utilizes the sampling capacity, limiting exploration and eventual improvement on harder examples. As a fix, we propose Pass-at-k Policy Optimization (PKPO), a transformation on the final rewards which leads to direct optimization of pass@k performance, thus optimizing for sets of samples that maximize reward when considered jointly. Our contribution is to derive novel low variance unbiased estimators for pass@k and its gradient, in both the binary and continuous reward settings. We show optimization with our estimators reduces to standard RL with rewards that have been jointly transformed by a stable and efficient transformation function. While previous efforts are restricted to k=n, ours is the first to enable robust optimization of pass@k for any arbitrary k <= n. Moreover, instead of trading off pass@1 performance for pass@k gains, our method allows annealing k during training, optimizing both metrics and often achieving strong pass@1 numbers alongside significant pass@k gains. We validate our reward transformations on toy experiments, which reveal the variance reducing properties of our formulations. We also include real-world examples using the open-source LLM, GEMMA-2. We find that our transformation effectively optimizes for the target k. Furthermore, higher k values enable solving more and harder problems, while annealing k boosts both the pass@1 and pass@k . Crucially, for challenging task sets where conventional pass@1 optimization stalls, our pass@k approach unblocks learning, likely due to better exploration by prioritizing joint utility over the utility of individual samples.
Authors: Eric J. Michaud, Asher Parker-Sartori, Max Tegmark
Abstract: We study the problem of creating strong, yet narrow, AI systems. While recent AI progress has been driven by the training of large general-purpose foundation models, the creation of smaller models specialized for narrow domains could be valuable for both efficiency and safety. In this work, we explore two challenges involved in creating such systems, having to do with basic properties of how neural networks learn and structure their representations. The first challenge regards when it is possible to train narrow models from scratch. Through experiments on a synthetic task, we find that it is sometimes necessary to train networks on a wide distribution of data to learn certain narrow skills within that distribution. This effect arises when skills depend on each other hierarchically, and training on a broad distribution introduces a curriculum which substantially accelerates learning. The second challenge regards how to transfer particular skills from large general models into small specialized models. We find that model skills are often not perfectly localized to a particular set of prunable components. However, we find that methods based on pruning can still outperform distillation. We investigate the use of a regularization objective to align desired skills with prunable components while unlearning unnecessary skills.
Authors: Amir Hossein Rahmati, Sanket Jantre, Weifeng Zhang, Yucheng Wang, Byung-Jun Yoon, Nathan M. Urban, Xiaoning Qian
Abstract: Low-Rank Adaptation (LoRA) offers a cost-effective solution for fine-tuning large language models (LLMs), but it often produces overconfident predictions in data-scarce few-shot settings. To address this issue, several classical statistical learning approaches have been repurposed for scalable uncertainty-aware LoRA fine-tuning. However, these approaches neglect how input characteristics affect the predictive uncertainty estimates. To address this limitation, we propose Contextual Low-Rank Adaptation (C-LoRA) as a novel uncertainty-aware and parameter efficient fine-tuning approach, by developing new lightweight LoRA modules contextualized to each input data sample to dynamically adapt uncertainty estimates. Incorporating data-driven contexts into the parameter posteriors, C-LoRA mitigates overfitting, achieves well-calibrated uncertainties, and yields robust predictions. Extensive experiments on LLaMA2-7B models demonstrate that C-LoRA consistently outperforms the state-of-the-art uncertainty-aware LoRA methods in both uncertainty quantification and model generalization. Ablation studies further confirm the critical role of our contextual modules in capturing sample-specific uncertainties. C-LoRA sets a new standard for robust, uncertainty-aware LLM fine-tuning in few-shot regimes. Although our experiments are limited to 7B models, our method is architecture-agnostic and, in principle, applies beyond this scale; studying its scaling to larger models remains an open problem. Our code is available at https://github.com/ahra99/c_lora.
Authors: Alan Arazi, Eilam Shapira, Roi Reichart
Abstract: While deep learning has achieved remarkable success across many domains, it has historically underperformed on tabular learning tasks, which remain dominated by gradient boosting decision trees. However, recent advancements are paving the way for Tabular Foundation Models, which can leverage real-world knowledge and generalize across diverse datasets, particularly when the data contains free-text. Although incorporating language model capabilities into tabular tasks has been explored, most existing methods utilize static, target-agnostic textual representations, limiting their effectiveness. We introduce TabSTAR: a Tabular Foundation Model with Semantically Target-Aware Representations. TabSTAR is designed to enable transfer learning on tabular data with textual features, with an architecture free of dataset-specific parameters. It unfreezes a pretrained text encoder and takes as input target tokens, which provide the model with the context needed to learn task-specific embeddings. TabSTAR achieves state-of-the-art performance for both medium- and large-sized datasets across known benchmarks of classification tasks with text features, and its pretraining phase exhibits scaling laws in the number of datasets, offering a pathway for further performance improvements.
Authors: Claude Formanek, Omayma Mahjoub, Louay Ben Nessir, Sasha Abramowitz, Ruan de Kock, Wiem Khlifi, Daniel Rajaonarivonivelomanantsoa, Simon Du Toit, Arnol Fokam, Siddarth Singh, Ulrich Mbou Sob, Felix Chalumeau, Arnu Pretorius
Abstract: A key challenge in offline multi-agent reinforcement learning (MARL) is achieving effective many-agent multi-step coordination in complex environments. In this work, we propose Oryx, a novel algorithm for offline cooperative MARL to directly address this challenge. Oryx adapts the recently proposed retention-based architecture Sable and combines it with a sequential form of implicit constraint Q-learning (ICQ), to develop a novel offline autoregressive policy update scheme. This allows Oryx to solve complex coordination challenges while maintaining temporal coherence over long trajectories. We evaluate Oryx across a diverse set of benchmarks from prior works -- SMAC, RWARE, and Multi-Agent MuJoCo -- covering tasks of both discrete and continuous control, varying in scale and difficulty. Oryx achieves state-of-the-art performance on more than 80% of the 65 tested datasets, outperforming prior offline MARL methods and demonstrating robust generalisation across domains with many agents and long horizons. Finally, we introduce new datasets to push the limits of many-agent coordination in offline MARL, and demonstrate Oryx's superior ability to scale effectively in such settings.
Authors: Fu Luo, Yaoxin Wu, Zhi Zheng, Zhenkun Wang
Abstract: Recent neural combinatorial optimization (NCO) methods have shown promising problem-solving ability without requiring domain-specific expertise. Most existing NCO methods use training and testing data with a fixed constraint value and lack research on the effect of constraint tightness on the performance of NCO methods. This paper takes the capacity-constrained vehicle routing problem (CVRP) as an example to empirically analyze the NCO performance under different tightness degrees of the capacity constraint. Our analysis reveals that existing NCO methods overfit the capacity constraint, and they can only perform satisfactorily on a small range of the constraint values but poorly on other values. To tackle this drawback of existing NCO methods, we develop an efficient training scheme that explicitly considers varying degrees of constraint tightness and proposes a multi-expert module to learn a generally adaptable solving strategy. Experimental results show that the proposed method can effectively overcome the overfitting issue, demonstrating superior performances on the CVRP and CVRP with time windows (CVRPTW) with various constraint tightness degrees.
Authors: Danyal Rehman, Oscar Davis, Jiarui Lu, Jian Tang, Michael Bronstein, Yoshua Bengio, Alexander Tong, Avishek Joey Bose
Abstract: Simulation-free training frameworks have been at the forefront of the generative modelling revolution in continuous spaces, leading to large-scale diffusion and flow matching models. However, such modern generative models suffer from expensive inference, inhibiting their use in numerous scientific applications like Boltzmann Generators (BGs) for molecular conformations that require fast likelihood evaluation. In this paper, we revisit classical normalizing flows in the context of BGs that offer efficient sampling and likelihoods, but whose training via maximum likelihood is often unstable and computationally challenging. We propose Regression Training of Normalizing Flows (RegFlow), a novel and scalable regression-based training objective that bypasses the numerical instability and computational challenge of conventional maximum likelihood training in favour of a simple $\ell_2$-regression objective. Specifically, RegFlow maps prior samples under our flow to targets computed using optimal transport couplings or a pre-trained continuous normalizing flow (CNF). To enhance numerical stability, RegFlow employs effective regularization strategies such as a new forward-backward self-consistency loss that enjoys painless implementation. Empirically, we demonstrate that RegFlow unlocks a broader class of architectures that were previously intractable to train for BGs with maximum likelihood. We also show RegFlow exceeds the performance, computational cost, and stability of maximum likelihood training in equilibrium sampling in Cartesian coordinates of alanine dipeptide, tripeptide, and tetrapeptide, showcasing its potential in molecular systems.
Authors: Fuxiang Zhang, Jiacheng Xu, Chaojie Wang, Ce Cui, Yang Liu, Bo An
Abstract: Large Language Models (LLMs) have demonstrated remarkable progress in complex reasoning tasks through both post-training and test-time scaling laws. While prevalent test-time scaling approaches are often realized by using external reward models to guide the model generation process, we find that only marginal gains can be acquired when scaling a model post-trained on specific reasoning tasks. We identify that the limited improvement stems from distribution discrepancies between the specific post-trained generator and the general reward model. To address this, we propose a framework that incentivizes LLMs to self-verify their own answers. By unifying answer generation and verification within a single reinforcement learning (RL) process, we train models that can effectively assess the correctness of their own solutions. The trained model can further scale its performance at inference time by verifying its generations, without the need for external verifiers. We train our self-verification models based on Qwen2.5-Math-7B and DeepSeek-R1-Distill-Qwen-1.5B, demonstrating their capabilities across varying reasoning context lengths. Experiments on multiple mathematical reasoning benchmarks show that our models can not only improve post-training performance but also enable effective test-time scaling.
Authors: Yuanyao Chen, Rongsheng Chen, Fu Luo, Zhenkun Wang
Abstract: Neural Combinatorial Optimization (NCO) has emerged as a promising learning-based paradigm for addressing Vehicle Routing Problems (VRPs) by minimizing the need for extensive manual engineering. While existing NCO methods, trained on small-scale instances (e.g., 100 nodes), have demonstrated considerable success on problems of similar scale, their performance significantly degrades when applied to large-scale scenarios. This degradation arises from the distributional shift between training and testing data, rendering policies learned on small instances ineffective for larger problems. To overcome this limitation, we introduce a novel learning framework driven by Large Language Models (LLMs). This framework learns a projection between the training and testing distributions, which is then deployed to enhance the scalability of the NCO model. Notably, unlike prevailing techniques that necessitate joint training with the neural network, our approach operates exclusively during the inference phase, obviating the need for model retraining. Extensive experiments demonstrate that our method enables a backbone model (trained on 100-node instances) to achieve superior performance on large-scale Traveling Salesman Problem (TSP) and Capacitated Vehicle Routing Problem (CVRP) of up to 100K nodes from diverse distributions.
Authors: Yuepeng Zheng, Fu Luo, Zhenkun Wang, Yaoxin Wu, Yu Zhou
Abstract: Multi-Task Learning (MTL) in Neural Combinatorial Optimization (NCO) is a promising approach to train a unified model capable of solving multiple Vehicle Routing Problem (VRP) variants. However, existing Reinforcement Learning (RL)-based multi-task methods can only train light decoder models on small-scale problems, exhibiting limited generalization ability when solving large-scale problems. To overcome this limitation, this work introduces a novel multi-task learning method driven by knowledge distillation (MTL-KD), which enables the efficient training of heavy decoder models with strong generalization ability. The proposed MTL-KD method transfers policy knowledge from multiple distinct RL-based single-task models to a single heavy decoder model, facilitating label-free training and effectively improving the model's generalization ability across diverse tasks. In addition, we introduce a flexible inference strategy termed Random Reordering Re-Construction (R3C), which is specifically adapted for diverse VRP tasks and further boosts the performance of the multi-task model. Experimental results on 6 seen and 10 unseen VRP variants with up to 1000 nodes indicate that our proposed method consistently achieves superior performance on both uniform and real-world benchmarks, demonstrating robust generalization abilities.
Authors: Wenyu Zhu, Jianhui Wang, Bowen Gao, Yinjun Jia, Haichuan Tan, Ya-Qin Zhang, Wei-Ying Ma, Yanyan Lan
Abstract: Virtual screening (VS) is a critical component of modern drug discovery, yet most existing methods--whether physics-based or deep learning-based--are developed around holo protein structures with known ligand-bound pockets. Consequently, their performance degrades significantly on apo or predicted structures such as those from AlphaFold2, which are more representative of real-world early-stage drug discovery, where pocket information is often missing. In this paper, we introduce an alignment-and-aggregation framework to enable accurate virtual screening under structural uncertainty. Our method comprises two core components: (1) a tri-modal contrastive learning module that aligns representations of the ligand, the holo pocket, and cavities detected from structures, thereby enhancing robustness to pocket localization error; and (2) a cross-attention based adapter for dynamically aggregating candidate binding sites, enabling the model to learn from activity data even without precise pocket annotations. We evaluated our method on a newly curated benchmark of apo structures, where it significantly outperforms state-of-the-art methods in blind apo setting, improving the early enrichment factor (EF1%) from 11.75 to 37.19. Notably, it also maintains strong performance on holo structures. These results demonstrate the promise of our approach in advancing first-in-class drug discovery, particularly in scenarios lacking experimentally resolved protein-ligand complexes. Our implementation is publicly available at https://github.com/Wiley-Z/AANet.
Authors: Shakir Yousefi, Andreas Plesner, Till Aczel, Roger Wattenhofer
Abstract: Modern neural networks demonstrate state-of-the-art performance on numerous existing benchmarks; however, their high computational requirements and energy consumption prompt researchers to seek more efficient solutions for real-world deployment. Logic gate networks (LGNs) learns a large network of logic gates for efficient image classification. However, learning a network that can solve a simple problem like CIFAR-10 can take days to weeks to train. Even then, almost half of the network remains unused, causing a discretization gap. This discretization gap hinders real-world deployment of LGNs, as the performance drop between training and inference negatively impacts accuracy. We inject Gumbel noise with a straight-through estimator during training to significantly speed up training, improve neuron utilization, and decrease the discretization gap. We theoretically show that this results from implicit Hessian regularization, which improves the convergence properties of LGNs. We train networks $4.5 \times$ faster in wall-clock time, reduce the discretization gap by $98\%$, and reduce the number of unused gates by $100\%$.
Authors: Youqi Wu, Jingwei Zhang, Farzan Farnia
Abstract: State-of-the-art embeddings often capture distinct yet complementary discriminative features: For instance, one image embedding model may excel at distinguishing fine-grained textures, while another focuses on object-level structure. Motivated by this observation, we propose a principled approach to fuse such complementary representations through kernel multiplication. Multiplying the kernel similarity functions of two embeddings allows their discriminative structures to interact, producing a fused representation whose kernel encodes the union of the clusters identified by each parent embedding. This formulation also provides a natural way to construct joint kernels for paired multi-modal data (e.g., image-text tuples), where the product of modality-specific kernels inherits structure from both domains. We highlight that this kernel product is mathematically realized via the Kronecker product of the embedding feature maps, yielding our proposed KrossFuse framework for embedding fusion. To address the computational cost of the resulting high-dimensional Kronecker space, we further develop RP-KrossFuse, a scalable variant that leverages random projections for efficient approximation. As a key application, we use this framework to bridge the performance gap between cross-modal embeddings (e.g., CLIP, BLIP) and unimodal experts (e.g., DINOv2, E5). Experiments show that RP-KrossFuse effectively integrates these models, enhancing modality-specific performance while preserving cross-modal alignment. The project code is available at https://github.com/yokiwuuu/KrossFuse.
Authors: Steffen Schotth\"ofer, Timon Klein, Jonas Kusch
Abstract: Low-rank pre-training and fine-tuning have recently emerged as promising techniques for reducing the computational and storage costs of large neural networks. Training low-rank parameterizations typically relies on conventional optimizers such as heavy ball momentum methods or Adam. In this work, we identify and analyze potential difficulties that these training methods encounter when used to train low-rank parameterizations of weights. In particular, we show that classical momentum methods can struggle to converge to a local optimum due to the geometry of the underlying optimization landscape. To address this, we introduce novel training strategies derived from dynamical low-rank approximation, which explicitly account for the underlying geometric structure. Our approach leverages and combines tools from dynamical low-rank approximation and momentum-based optimization to design optimizers that respect the intrinsic geometry of the parameter space. We validate our methods through numerical experiments, demonstrating faster convergence, and stronger validation metrics at given parameter budgets.
Authors: Antonis Vasileiou, Timo Stoll, Christopher Morris
Abstract: Using message-passing graph neural networks (MPNNs) for node and link prediction is crucial in various scientific and industrial domains, which has led to the development of diverse MPNN architectures. Besides working well in practical settings, their ability to generalize beyond the training set remains poorly understood. While some studies have explored MPNNs' generalization in graph-level prediction tasks, much less attention has been given to node- and link-level predictions. Existing works often rely on unrealistic i.i.d.\@ assumptions, overlooking possible correlations between nodes or links, and assuming fixed aggregation and impractical loss functions while neglecting the influence of graph structure. In this work, we introduce a unified framework to analyze the generalization properties of MPNNs in inductive and transductive node and link prediction settings, incorporating diverse architectural parameters and loss functions and quantifying the influence of graph structure. Additionally, our proposed generalization framework can be applied beyond graphs to any classification task under the inductive or transductive setting. Our empirical study supports our theoretical insights, deepening our understanding of MPNNs' generalization capabilities in these tasks.
Authors: Yuchen Ma, Dennis Frauen, Jonas Schweisthal, Stefan Feuerriegel
Abstract: Estimating treatment effects is crucial for personalized decision-making in medicine, but this task faces unique challenges in clinical practice. At training time, models for estimating treatment effects are typically trained on well-structured medical datasets that contain detailed patient information. However, at inference time, predictions are often made using textual descriptions (e.g., descriptions with self-reported symptoms), which are incomplete representations of the original patient information. In this work, we make three contributions. (1) We show that the discrepancy between the data available during training time and inference time can lead to biased estimates of treatment effects. We formalize this issue as an inference time text confounding problem, where confounders are fully observed during training time but only partially available through text at inference time. (2) To address this problem, we propose a novel framework for estimating treatment effects that explicitly accounts for inference time text confounding. Our framework leverages large language models together with a custom doubly robust learner to mitigate biases caused by the inference time text confounding. (3) Through a series of experiments, we demonstrate the effectiveness of our framework in real-world applications.
Authors: Minhak Song, Beomhan Baek, Kwangjun Ahn, Chulhee Yun
Abstract: As both model and dataset sizes continue to scale rapidly, conventional pretraining strategies with fixed compute budgets-such as cosine learning rate schedules-are increasingly inadequate for large-scale training. Recent alternatives, including warmup-stable-decay (WSD) schedules and weight averaging, offer greater flexibility. However, WSD relies on explicit decay phases to track progress, while weight averaging addresses this limitation at the cost of additional memory. In search of a more principled and scalable alternative, we revisit the Schedule-Free (SF) method [Defazio et al., 2024], which has shown strong empirical performance across diverse settings. We show that SF-AdamW effectively navigates the "river" structure of the loss landscape without decay phases or auxiliary averaging, making it particularly suitable for continuously scaling training workloads. To understand this behavior, we conduct a theoretical and empirical analysis of SF dynamics, revealing that it implicitly performs weight averaging without memory overhead. Guided by this analysis, we propose a refined variant of SF that improves robustness to momentum and performs better under large batch sizes, addressing key limitations of the original method. Together, these results establish SF as a practical, scalable, and theoretically grounded approach for language model training.
Authors: Yu-Jie Zhang, Sheng-An Xu, Peng Zhao, Masashi Sugiyama
Abstract: We study the generalized linear bandit (GLB) problem, a contextual multi-armed bandit framework that extends the classical linear model by incorporating a non-linear link function, thereby modeling a broad class of reward distributions such as Bernoulli and Poisson. While GLBs are widely applicable to real-world scenarios, their non-linear nature introduces significant challenges in achieving both computational and statistical efficiency. Existing methods typically trade off between two objectives, either incurring high per-round costs for optimal regret guarantees or compromising statistical efficiency to enable constant-time updates. In this paper, we propose a jointly efficient algorithm that attains a nearly optimal regret bound with $\mathcal{O}(1)$ time and space complexities per round. The core of our method is a tight confidence set for the online mirror descent (OMD) estimator, which is derived through a novel analysis that leverages the notion of mix loss from online prediction. The analysis shows that our OMD estimator, even with its one-pass updates, achieves statistical efficiency comparable to maximum likelihood estimation, thereby leading to a jointly efficient optimistic method.
Authors: Azmine Toushik Wasi
Abstract: Drug-drug interactions pose a significant challenge in clinical pharmacology, with severe class imbalance among interaction types limiting the effectiveness of predictive models. Common interactions dominate datasets, while rare but critical interactions remain underrepresented, leading to poor model performance on infrequent cases. Existing methods often treat DDI prediction as a binary problem, ignoring class-specific nuances and exacerbating bias toward frequent interactions. To address this, we propose a framework combining Generative Flow Networks (GFlowNet) with Variational Graph Autoencoders (VGAE) to generate synthetic samples for rare classes, improving model balance and generate effective and novel DDI pairs. Our approach enhances predictive performance across interaction types, ensuring better clinical reliability.
Authors: Wenyou Guo, Ting Qu, Chunrong Pan, George Q. Huang
Abstract: Federated learning (FL), as a distributed collaborative machine learning (ML) framework under privacy-preserving constraints, has garnered increasing research attention in cross-organizational data collaboration scenarios. This paper proposes a class of distributed optimization algorithms based on the augmented Lagrangian technique, designed to accommodate diverse communication topologies in both centralized and decentralized FL settings. Furthermore, we develop multiple termination criteria and parameter update mechanisms to enhance computational efficiency, accompanied by rigorous theoretical guarantees of convergence. By generalizing the augmented Lagrangian relaxation through the incorporation of proximal relaxation and quadratic approximation, our framework systematically recovers a broad of classical unconstrained optimization methods, including proximal algorithm, classic gradient descent, and stochastic gradient descent, among others. Notably, the convergence properties of these methods can be naturally derived within the proposed theoretical framework. Numerical experiments demonstrate that the proposed algorithm exhibits strong performance in large-scale settings with significant statistical heterogeneity across clients.
Authors: Farnoush Rezaei Jafari, Oliver Eberle, Ashkan Khakzar, Neel Nanda
Abstract: Activation patching is a standard method in mechanistic interpretability for localizing the components of a model responsible for specific behaviors, but it is computationally expensive to apply at scale. Attribution patching offers a faster, gradient-based approximation, yet suffers from noise and reduced reliability in deep, highly non-linear networks. In this work, we introduce Relevance Patching (RelP), which replaces the local gradients in attribution patching with propagation coefficients derived from Layer-wise Relevance Propagation (LRP). LRP propagates the network's output backward through the layers, redistributing relevance to lower-level components according to local propagation rules that ensure properties such as relevance conservation or improved signal-to-noise ratio. Like attribution patching, RelP requires only two forward passes and one backward pass, maintaining computational efficiency while improving faithfulness. We validate RelP across a range of models and tasks, showing that it more accurately approximates activation patching than standard attribution patching, particularly when analyzing residual stream and MLP outputs in the Indirect Object Identification (IOI) task. For instance, for MLP outputs in GPT-2 Large, attribution patching achieves a Pearson correlation of 0.006, whereas RelP reaches 0.956, highlighting the improvement offered by RelP. Additionally, we compare the faithfulness of sparse feature circuits identified by RelP and Integrated Gradients (IG), showing that RelP achieves comparable faithfulness without the extra computational cost associated with IG.
Authors: Arth Sojitra, Mrigank Dhingra, Omer San
Abstract: Deep Operator Networks (DeepONets) have recently emerged as powerful data-driven frameworks for learning nonlinear operators, particularly suited for approximating solutions to partial differential equations. Despite their promising capabilities, the standard implementation of DeepONets, which typically employs fully connected linear layers in the trunk network, can encounter limitations in capturing complex spatial structures inherent to various PDEs. To address this limitation, we introduce Fourier-embedded trunk networks within the DeepONet architecture, leveraging random Fourier feature mappings to enrich spatial representation capabilities. Our proposed Fourier-embedded DeepONet (FEDONet) demonstrates superior performance compared to the traditional DeepONet across a comprehensive suite of PDE-driven datasets, including the two-dimensional Poisson, Burgers', Lorenz-63, Eikonal, Allen-Cahn, and the Kuramoto-Sivashinsky equation. Empirical evaluations of FEDONet consistently show significant improvements in solution reconstruction accuracy, with average relative $L^2$ performance gains ranging between 2-3$\times$ compared to the DeepONet baseline. This study highlights the effectiveness of Fourier embeddings in enhancing neural operator learning, offering a robust and broadly applicable methodology for PDE surrogate modeling.
Authors: Allen Schmaltz
Abstract: We introduce the Similarity-Distance-Magnitude (SDM) activation function, a more robust and interpretable formulation of the standard softmax activation function, 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, and enabling interpretability-by-exemplar via dense matching. We further introduce the SDM estimator, based on a data-driven partitioning of the class-wise empirical CDFs via the SDM activation, to control the class- and prediction-conditional accuracy among selective classifications. When used as the final-layer activation over pre-trained language models for selective classification, the SDM estimator is more robust to co-variate shifts and out-of-distribution inputs than existing calibration methods using softmax activations, while remaining informative over in-distribution data.
Authors: Vicky Kouni
Abstract: Unfolding networks are interpretable networks emerging from iterative algorithms, incorporate prior knowledge of data structure, and are designed to solve inverse problems like compressed sensing, which deals with recovering data from noisy, missing observations. Compressed sensing finds applications in critical domains, from medical imaging to cryptography, where adversarial robustness is crucial to prevent catastrophic failures. However, a solid theoretical understanding of the performance of unfolding networks in the presence of adversarial attacks is still in its infancy. In this paper, we study the adversarial generalization of unfolding networks when perturbed with $l_2$-norm constrained attacks, generated by the fast gradient sign method. Particularly, we choose a family of state-of-the-art overaparameterized unfolding networks and deploy a new framework to estimate their adversarial Rademacher complexity. Given this estimate, we provide adversarial generalization error bounds for the networks under study, which are tight with respect to the attack level. To our knowledge, this is the first theoretical analysis on the adversarial generalization of unfolding networks. We further present a series of experiments on real-world data, with results corroborating our derived theory, consistently for all data. Finally, we observe that the family's overparameterization can be exploited to promote adversarial robustness, shedding light on how to efficiently robustify neural networks.
Authors: Junlong Ke, Qiying Hu, Shenghai Yuan, Yuecong Xu, Jianfei Yang
Abstract: Modern signal processing (SP) pipelines, whether model-based or data-driven, often constrained by complex and fragmented workflow, rely heavily on expert knowledge and manual engineering, and struggle with adaptability and generalization under limited data. In contrast, Large Language Models (LLMs) offer strong reasoning capabilities, broad general-purpose knowledge, in-context learning, and cross-modal transfer abilities, positioning them as powerful tools for automating and generalizing SP workflows. Motivated by these potentials, we introduce SignalLLM, the first general-purpose LLM-based agent framework for general SP tasks. Unlike prior LLM-based SP approaches that are limited to narrow applications or tricky prompting, SignalLLM introduces a principled, modular architecture. It decomposes high-level SP goals into structured subtasks via in-context learning and domain-specific retrieval, followed by hierarchical planning through adaptive retrieval-augmented generation (RAG) and refinement; these subtasks are then executed through prompt-based reasoning, cross-modal reasoning, code synthesis, model invocation, or data-driven LLM-assisted modeling. Its generalizable design enables the flexible selection of problem solving strategies across different signal modalities, task types, and data conditions. We demonstrate the versatility and effectiveness of SignalLLM through five representative tasks in communication and sensing, such as radar target detection, human activity recognition, and text compression. Experimental results show superior performance over traditional and existing LLM-based methods, particularly in few-shot and zero-shot settings.
Authors: Jin Li, Shoujin Wang, Qi Zhang, Feng Liu, Tongliang Liu, Longbing Cao, Shui Yu, Fang Chen
Abstract: Uncovering cause-and-effect mechanisms from data is fundamental to scientific progress. While large language models (LLMs) show promise for enhancing causal discovery (CD) from unstructured data, their application to the increasingly prevalent multimodal setting remains a critical challenge. Even with the advent of multimodal LLMs (MLLMs), their efficacy in multimodal CD is hindered by two primary limitations: (1) difficulty in exploring intra- and inter-modal interactions for comprehensive causal variable identification; and (2) insufficiency to handle structural ambiguities with purely observational data. To address these challenges, we propose MLLM-CD, a novel framework for multimodal causal discovery from unstructured data. It consists of three key components: (1) a novel contrastive factor discovery module to identify genuine multimodal factors based on the interactions explored from contrastive sample pairs; (2) a statistical causal structure discovery module to infer causal relationships among discovered factors; and (3) an iterative multimodal counterfactual reasoning module to refine the discovery outcomes iteratively by incorporating the world knowledge and reasoning capabilities of MLLMs. Extensive experiments on both synthetic and real-world datasets demonstrate the effectiveness of the proposed MLLM-CD in revealing genuine factors and causal relationships among them from multimodal unstructured data.
Authors: Marcin Mo\.zejko, Adam Bielecki, Jurand Pr\k{a}dzy\'nski, Marcin Traskowski, Antoni Janowski, Hyun-Su Lee, Marcelo Der Torossian Torres, Micha{\l} Kmicikiewicz, Paulina Szymczak, Karol Jurasz, Micha{\l} Kucharczyk, Cesar de la Fuente-Nunez, Ewa Szczurek
Abstract: Antimicrobial peptide discovery is challenged by the astronomical size of peptide space and the relative scarcity of active peptides. Generative models provide continuous latent "maps" of peptide space, but conventionally ignore decoder-induced geometry and rely on flat Euclidean metrics, rendering exploration and optimization distorted and inefficient. Prior manifold-based remedies assume fixed intrinsic dimensionality, which critically fails in practice for peptide data. Here, we introduce PepCompass, a geometry-aware framework for peptide exploration and optimization. At its core, we define a Union of $\kappa$-Stable Riemannian Manifolds $\mathbb{M}^{\kappa}$, a family of decoder-induced manifolds that captures local geometry while ensuring computational stability. We propose two local exploration methods: Second-Order Riemannian Brownian Efficient Sampling, which provides a convergent second-order approximation to Riemannian Brownian motion, and Mutation Enumeration in Tangent Space, which reinterprets tangent directions as discrete amino-acid substitutions. Combining these yields Local Enumeration Bayesian Optimization (LE-BO), an efficient algorithm for local activity optimization. Finally, we introduce Potential-minimizing Geodesic Search (PoGS), which interpolates between prototype embeddings along property-enriched geodesics, biasing discovery toward seeds, i.e. peptides with favorable activity. In-vitro validation confirms the effectiveness of PepCompass: PoGS yields four novel seeds, and subsequent optimization with LE-BO discovers 25 highly active peptides with broad-spectrum activity, including against resistant bacterial strains. These results demonstrate that geometry-informed exploration provides a powerful new paradigm for antimicrobial peptide design.
Authors: Jinhui Bai, Andreas Christmann, Lei Shi
Abstract: In this paper, we propose a novel kernel stochastic gradient descent (SGD) algorithm for large-scale supervised learning with general losses. Compared to traditional kernel SGD, our algorithm improves efficiency and scalability through an innovative regularization strategy. By leveraging the infinite series expansion of spherical radial basis functions, this strategy projects the stochastic gradient onto a finite-dimensional hypothesis space, which is adaptively scaled according to the bias-variance trade-off, thereby enhancing generalization performance. Based on a new estimation of the spectral structure of the kernel-induced covariance operator, we develop an analytical framework that unifies optimization and generalization analyses. We prove that both the last iterate and the suffix average converge at minimax-optimal rates, and we further establish optimal strong convergence in the reproducing kernel Hilbert space. Our framework accommodates a broad class of classical loss functions, including least-squares, Huber, and logistic losses. Moreover, the proposed algorithm significantly reduces computational complexity and achieves optimal storage complexity by incorporating coordinate-wise updates from linear SGD, thereby avoiding the costly pairwise operations typical of kernel SGD and enabling efficient processing of streaming data. Finally, extensive numerical experiments demonstrate the efficiency of our approach.
Authors: Jiatong Yu, Yinghui He, Anirudh Goyal, Sanjeev Arora
Abstract: Machine unlearning seeks to selectively remove the "influence" of specific training data on a model's outputs. The ideal goal is Retrain Equivalence--behavior identical to a model trained from scratch on only the retained data. This goal was formulated for models trained on i.i.d. data batches, but modern pipelines often involve multi-stage training, with each stage having a distinct data distribution and objective. Examples include LLM fine-tuning for alignment, reasoning ability, etc. Our study shows via theory and experiments that this shift to multi-stage training introduces a fundamental barrier for machine unlearning. The theory indicates that the outcome of local unlearning--methods that only use gradients computed on the forget set--is path-dependent. That is, a model's behavior during unlearning is influenced by the order of its training stages during learning, making it impossible for path-oblivious algorithms to universally achieve Retrain Equivalence. We empirically demonstrate the same phenomenon in LLM post-training across Llama and Qwen models (1B to 14B) with gradient ascent, NPO, and SimNPO local unlearning algorithms. Models fine-tuned via different orderings of identical training stages diverge in behavior during unlearning, with the degradation in GSM8K accuracy after unlearning varying by over 20% across paths. We also observe that some learning paths consistently produce models that unlearn slowly. During unlearning, whether the probability mass gets squeezed into paraphrasing or alternative concepts is also path-dependent. These results consistently show that Retrain Equivalence is an ill-posed target for local unlearning algorithms, so long as the target models are trained in stages. In situations where access to models' training histories is hard, the current work calls for rethinking the definition and desiderata of machine unlearning.
Authors: Yehonathan Refael, Amit Aides, Aviad Barzilai, George Leifman, Genady Beryozkin, Vered Silverman, Bolous Jaber, Tomer Shekel
Abstract: Open-vocabulary object detection (OVD) models offer remarkable flexibility by detecting objects from arbitrary text queries. However, their zero-shot performance in specialized domains like Remote Sensing (RS) is often compromised by the inherent ambiguity of natural language, limiting critical downstream applications. For instance, an OVD model may struggle to distinguish between fine-grained classes such as "fishing boat" and "yacht" since their embeddings are similar and often inseparable. This can hamper specific user goals, such as monitoring illegal fishing, by producing irrelevant detections. To address this, we propose a cascaded approach that couples the broad generalization of a large pre-trained OVD model with a lightweight few-shot classifier. Our method first employs the zero-shot model to generate high-recall object proposals. These proposals are then refined for high precision by a compact classifier trained in real-time on only a handful of user-annotated examples - drastically reducing the high costs of RS imagery annotation.The core of our framework is FLAME, a one-step active learning strategy that selects the most informative samples for training. FLAME identifies, on the fly, uncertain marginal candidates near the decision boundary using density estimation, followed by clustering to ensure sample diversity. This efficient sampling technique achieves high accuracy without costly full-model fine-tuning and enables instant adaptation, within less then a minute, which is significantly faster than state-of-the-art alternatives.Our method consistently surpasses state-of-the-art performance on RS benchmarks, establishing a practical and resource-efficient framework for adapting foundation models to specific user needs.
Authors: Gereon Elvers, Gilad Landau, Oiwi Parker Jones
Abstract: Non-invasive brain-computer interfaces (BCIs) are beginning to benefit from large, public benchmarks. However, current benchmarks target relatively simple, foundational tasks like Speech Detection and Phoneme Classification, while application-ready results on tasks like Brain-to-Text remain elusive. We propose Keyword Spotting (KWS) as a practically applicable, privacy-aware intermediate task. Using the deep 52-hour, within-subject LibriBrain corpus, we provide standardized train/validation/test splits for reproducible benchmarking, and adopt an evaluation protocol tailored to extreme class imbalance. Concretely, we use area under the precision-recall curve (AUPRC) as a robust evaluation metric, complemented by false alarms per hour (FA/h) at fixed recall to capture user-facing trade-offs. To simplify deployment and further experimentation within the research community, we are releasing an updated version of the pnpl library with word-level dataloaders and Colab-ready tutorials. As an initial reference model, we present a compact 1-D Conv/ResNet baseline with focal loss and top-k pooling that is trainable on a single consumer-class GPU. The reference model achieves approximately 13x the permutation baseline AUPRC on held-out sessions, demonstrating the viability of the task. Exploratory analyses reveal: (i) predictable within-subject scaling - performance improves log-linearly with more training hours - and (ii) the existence of word-level factors (frequency and duration) that systematically modulate detectability.
Authors: Hyeongyu Kim, Geonhui Han, Dosik Hwang
Abstract: In recent advancements in Test Time Adaptation (TTA), most existing methodologies focus on updating normalization layers to adapt to the test domain. However, the reliance on normalization-based adaptation presents key challenges. First, normalization layers such as Batch Normalization (BN) are highly sensitive to small batch sizes, leading to unstable and inaccurate statistics. Moreover, normalization-based adaptation is inherently constrained by the structure of the pre-trained model, as it relies on training-time statistics that may not generalize well to unseen domains. These issues limit the effectiveness of normalization-based TTA approaches, especially under significant domain shift. In this paper, we introduce a novel paradigm based on the concept of a Buffer layer, which addresses the fundamental limitations of normalization layer updates. Unlike existing methods that modify the core parameters of the model, our approach preserves the integrity of the pre-trained backbone, inherently mitigating the risk of catastrophic forgetting during online adaptation. Through comprehensive experimentation, we demonstrate that our approach not only outperforms traditional methods in mitigating domain shift and enhancing model robustness, but also exhibits strong resilience to forgetting. Furthermore, our Buffer layer is modular and can be seamlessly integrated into nearly all existing TTA frameworks, resulting in consistent performance improvements across various architectures. These findings validate the effectiveness and versatility of the proposed solution in real-world domain adaptation scenarios. The code is available at https://github.com/hyeongyu-kim/Buffer_TTA.
Authors: Tianxiang Wang, Yingtong Ke, Dhananjay Bhaskar, Smita Krishnaswamy, Alexander Cloninger
Abstract: Single-cell technologies generate high-dimensional point clouds of cells, enabling detailed characterization of complex patient states and treatment responses. Yet each patient is represented by an irregular point cloud rather than a simple vector, making it difficult to directly quantify and compare biological differences between individuals. Nonlinear methods such as kernels and neural networks achieve predictive accuracy but act as black boxes, offering little biological interpretability. To address these limitations, we adapt the Linear Optimal Transport (LOT) framework to this setting, embedding irregular point clouds into a fixed-dimensional Euclidean space while preserving distributional structure. This embedding provides a principled linear representation that preserves optimal transport geometry while enabling downstream analysis. It also forms a registration between any two patients, enabling direct comparison of their cellular distributions. Within this space, LOT enables: (i) \textbf{accurate and interpretable classification} of COVID-19 patient states, where classifier weights map back to specific markers and spatial regions driving predictions; and (ii) \textbf{synthetic data generation} for patient-derived organoids, exploiting the linearity of the LOT embedding. LOT barycenters yield averaged cellular profiles representing combined conditions or samples, supporting drug interaction testing. Together, these results establish LOT as a unified framework that bridges predictive performance, interpretability, and generative modeling. By transforming heterogeneous point clouds into structured embeddings directly traceable to the original data, LOT opens new opportunities for understanding immune variation and treatment effects in high-dimensional biological systems.
Authors: Thomas Bailie, S. Karthik Mukkavilli, Varvara Vetrova, Yun Sing Koh
Abstract: Climate events arise from intricate, multivariate dynamics governed by global-scale drivers, profoundly impacting food, energy, and infrastructure. Yet, accurate weather prediction remains elusive due to physical processes unfolding across diverse spatio-temporal scales, which fixed-resolution methods cannot capture. Hierarchical Graph Neural Networks (HGNNs) offer a multiscale representation, but nonlinear downward mappings often erase global trends, weakening the integration of physics into forecasts. We introduce HiFlowCast and its ensemble variant HiAntFlow, HGNNs that embed physics within a multiscale prediction framework. Two innovations underpin their design: a Latent-Memory-Retention mechanism that preserves global trends during downward traversal, and a Latent-to-Physics branch that integrates PDE solution fields across diverse scales. Our Flow models cut errors by over 5% at 13-day lead times and by 5-8% under 1st and 99th quantile extremes, improving reliability for rare events. Leveraging pretrained model weights, they converge within a single epoch, reducing training cost and their carbon footprint. Such efficiency is vital as the growing scale of machine learning challenges sustainability and limits research accessibility. Code and model weights are in the supplementary materials.
Authors: Omer Jauhar Khan, Sudais Khan, Hafeez Anwar, Shahzeb Khan, Shams Ul Arifeen
Abstract: Physics Informed Neural Networks (PINNs) are gaining attention for their ability to embed physical laws into deep learning models, which is particularly useful in structural engineering tasks with limited data. This paper aims to explore the use of PINNs to predict the weight of small scale spaghetti bridges, a task relevant to understanding load limits and potential failure modes in simplified structural models. Our proposed framework incorporates physics-based constraints to the prediction model for improved performance. In addition to standard PINNs, we introduce a novel architecture named Physics Informed Kolmogorov Arnold Network (PIKAN), which blends universal function approximation theory with physical insights. The structural parameters provided as input to the model are collected either manually or through computer vision methods. Our dataset includes 15 real bridges, augmented to 100 samples, and our best model achieves an $R^2$ score of 0.9603 and a mean absolute error (MAE) of 10.50 units. From applied perspective, we also provide a web based interface for parameter entry and prediction. These results show that PINNs can offer reliable estimates of structural weight, even with limited data, and may help inform early stage failure analysis in lightweight bridge designs. The complete data and code are available at https://github.com/OmerJauhar/PINNS-For-Spaghetti-Bridges.
URLs: https://github.com/OmerJauhar/PINNS-For-Spaghetti-Bridges.
Authors: Jonathan Amar, Edward Liu, Alessandra Breschi, Liangliang Zhang, Pouya Kheradpour, Sylvia Li, Lisa Soleymani Lehmann, Alessandro Giulianelli, Matt Edwards, Yugang Jia, David Nola, Raghav Mani, Pankaj Vats, Jesse Tetreault, T. J. Chen, Cory Y. McLean
Abstract: This paper introduces an innovative Electronic Health Record (EHR) foundation model that integrates Polygenic Risk Scores (PRS) as a foundational data modality, moving beyond traditional EHR-only approaches to build more holistic health profiles. Leveraging the extensive and diverse data from the All of Us (AoU) Research Program, this multimodal framework aims to learn complex relationships between clinical data and genetic predispositions. The methodology extends advancements in generative AI to the EHR foundation model space, enhancing predictive capabilities and interpretability. Evaluation on AoU data demonstrates the model's predictive value for the onset of various conditions, particularly Type 2 Diabetes (T2D), and illustrates the interplay between PRS and EHR data. The work also explores transfer learning for custom classification tasks, showcasing the architecture's versatility and efficiency. This approach is pivotal for unlocking new insights into disease prediction, proactive health management, risk stratification, and personalized treatment strategies, laying the groundwork for more personalized, equitable, and actionable real-world evidence generation in healthcare.
Authors: Akira Tamamori
Abstract: This paper presents Two-Stage LKPLO, a novel multi-stage outlier detection framework that overcomes the coexisting limitations of conventional projection-based methods: their reliance on a fixed statistical metric and their assumption of a single data structure. Our framework uniquely synthesizes three key concepts: (1) a generalized loss-based outlyingness measure (PLO) that replaces the fixed metric with flexible, adaptive loss functions like our proposed SVM-like loss; (2) a global kernel PCA stage to linearize non-linear data structures; and (3) a subsequent local clustering stage to handle multi-modal distributions. Comprehensive 5-fold cross-validation experiments on 10 benchmark datasets, with automated hyperparameter optimization, demonstrate that Two-Stage LKPLO achieves state-of-the-art performance. It significantly outperforms strong baselines on datasets with challenging structures where existing methods fail, most notably on multi-cluster data (Optdigits) and complex, high-dimensional data (Arrhythmia). Furthermore, an ablation study empirically confirms that the synergistic combination of both the kernelization and localization stages is indispensable for its superior performance. This work contributes a powerful new tool for a significant class of outlier detection problems and underscores the importance of hybrid, multi-stage architectures.
Authors: Benjamin Karic, Nina Herrmann, Jan Stenkamp, Paula Scharf, Fabian Gieseke, Angela Schwering
Abstract: The integration of the Internet of Things (IoT) and Artificial Intelligence offers significant opportunities to enhance our ability to monitor and address ecological changes. As environmental challenges become increasingly pressing, the need for effective remote monitoring solutions is more critical than ever. A major challenge in designing IoT applications for environmental monitoring - particularly those involving image data - is to create energy-efficient IoT devices capable of long-term operation in remote areas with limited power availability. Advancements in the field of Tiny Machine Learning allow the use of Convolutional Neural Networks (CNNs) on resource-constrained, battery-operated microcontrollers. Since data transfer is energy-intensive, performing inference directly on microcontrollers to reduce the message size can extend the operational lifespan of IoT nodes. This work evaluates the use of common Low Power Wide Area Networks and compressed CNNs trained on domain specific datasets on an ESP32-S3. Our experiments demonstrate, among other things, that executing CNN inference on-device and transmitting only the results reduces the overall energy consumption by a factor of up to five compared to sending raw image data. These findings advocate the development of IoT applications with reduced carbon footprint and capable of operating autonomously in environmental monitoring scenarios by incorporating EmbeddedML.
Authors: Shakeel Abdulkareem (George Mason University, College of Science, Fairfax, VA, USA, Raregen Youth Network, Translational Medical Research Department, Oakton, VA, USA), Bora Yimenicioglu (Raregen Youth Network, Translational Medical Research Department, Oakton, VA, USA), Khartik Uppalapati (Raregen Youth Network, Translational Medical Research Department, Oakton, VA, USA), Aneesh Gudipati (George Mason University, College of Science, Fairfax, VA, USA), Adan Eftekhari (Harvard University, Cambridge, MA, USA), Saleh Yassin (Harvard University, Cambridge, MA, USA)
Abstract: Rapid triage of suspected stroke needs accurate, bedside-deployable tools; EEG is promising but underused at first contact. We present an adaptive multitask EEG classifier that converts 32-channel signals to power spectral density features (Welch), uses a recurrent-convolutional network (GRU-TCN) to predict stroke type (healthy, ischemic, hemorrhagic), hemispheric lateralization, and severity, and applies a deep Q-network (DQN) to tune decision thresholds in real time. Using a patient-wise split of the UCLH Stroke EIT/EEG data set (44 recordings; about 26 acute stroke, 10 controls), the primary outcome was stroke-type performance; secondary outcomes were severity and lateralization. The baseline GRU-TCN reached 89.3% accuracy (F1 92.8%) for stroke type, about 96.9% (F1 95.9%) for severity, and about 96.7% (F1 97.4%) for lateralization. With DQN threshold adaptation, stroke-type accuracy increased to about 98.0% (F1 97.7%). We also tested robustness on an independent, low-density EEG cohort (ZJU4H) and report paired patient-level statistics. Analyses follow STARD 2015 guidance for diagnostic accuracy studies (index test: GRU-TCN+DQN; reference standard: radiology/clinical diagnosis; patient-wise evaluation). Adaptive thresholding shifts the operating point to clinically preferred sensitivity-specificity trade-offs, while integrated scalp-map and spectral visualizations support interpretability.
Authors: Tomas Hrycej, Bernhard Bermeitinger, Massimo Pavone, G\"otz-Henrik Wiegand, Siegfried Handschuh
Abstract: The key task of machine learning is to minimize the loss function that measures the model fit to the training data. The numerical methods to do this efficiently depend on the properties of the loss function. The most decisive among these properties is the convexity or non-convexity of the loss function. The fact that the loss function can have, and frequently has, non-convex regions has led to a widespread commitment to non-convex methods such as Adam. However, a local minimum implies that, in some environment around it, the function is convex. In this environment, second-order minimizing methods such as the Conjugate Gradient (CG) give a guaranteed superlinear convergence. We propose a novel framework grounded in the hypothesis that loss functions in real-world tasks swap from initial non-convexity to convexity towards the optimum. This is a property we leverage to design an innovative two-phase optimization algorithm. The presented algorithm detects the swap point by observing the gradient norm dependence on the loss. In these regions, non-convex (Adam) and convex (CG) algorithms are used, respectively. Computing experiments confirm the hypothesis that this simple convexity structure is frequent enough to be practically exploited to substantially improve convergence and accuracy.
Authors: Shpresim Sadiku, Moritz Wagner, Sebastian Pokutta
Abstract: Sparse adversarial attacks fool deep neural networks (DNNs) through minimal pixel perturbations, often regularized by the $\ell_0$ norm. Recent efforts have replaced this norm with a structural sparsity regularizer, such as the nuclear group norm, to craft group-wise sparse adversarial attacks. The resulting perturbations are thus explainable and hold significant practical relevance, shedding light on an even greater vulnerability of DNNs. However, crafting such attacks poses an optimization challenge, as it involves computing norms for groups of pixels within a non-convex objective. We address this by presenting a two-phase algorithm that generates group-wise sparse attacks within semantically meaningful areas of an image. Initially, we optimize a quasinorm adversarial loss using the $1/2-$quasinorm proximal operator tailored for non-convex programming. Subsequently, the algorithm transitions to a projected Nesterov's accelerated gradient descent with $2-$norm regularization applied to perturbation magnitudes. Rigorous evaluations on CIFAR-10 and ImageNet datasets demonstrate a remarkable increase in group-wise sparsity, e.g., $50.9\%$ on CIFAR-10 and $38.4\%$ on ImageNet (average case, targeted attack). This performance improvement is accompanied by significantly faster computation times, improved explainability, and a $100\%$ attack success rate.
Authors: Robin Str\"asser, Manuel Schaller, Karl Worthmann, Julian Berberich, Frank Allg\"ower
Abstract: The Koopman operator serves as the theoretical backbone for machine learning of dynamical control systems, where the operator is heuristically approximated by extended dynamic mode decomposition (EDMD). In this paper, we propose SafEDMD, a novel stability- and feedback-oriented EDMD-based controller design framework. Our approach leverages a reliable surrogate model generated in a data-driven fashion in order to provide closed-loop guarantees. In particular, we establish a controller design based on semi-definite programming with guaranteed stabilization of the underlying nonlinear system. As central ingredient, we derive proportional error bounds that vanish at the origin and are tailored to control tasks. We illustrate the developed method by means of several benchmark examples and highlight the advantages over state-of-the-art methods.
Authors: Yuepeng Yang, Cong Ma
Abstract: The Rasch model, a classical model in the item response theory, is widely used in psychometrics to model the relationship between individuals' latent traits and their binary responses to assessments or questionnaires. In this paper, we introduce a new likelihood-based estimator -- random pairing maximum likelihood estimator ($\mathrm{RP\text{-}MLE}$) and its bootstrapped variant multiple random pairing MLE ($\mathrm{MRP\text{-}MLE}$) which faithfully estimate the item parameters in the Rasch model. The new estimators have several appealing features compared to existing ones. First, both work for sparse observations, an increasingly important scenario in the big data era. Second, both estimators are provably minimax optimal in terms of finite sample $\ell_{\infty}$ estimation error. Lastly, both admit precise distributional characterization that allows uncertainty quantification on the item parameters, e.g., construction of confidence intervals for the item parameters. The main idea underlying $\mathrm{RP\text{-}MLE}$ and $\mathrm{MRP\text{-}MLE}$ is to randomly pair user-item responses to form item-item comparisons. This is carefully designed to reduce the problem size while retaining statistical independence. We also provide empirical evidence of the efficacy of the two new estimators using both simulated and real data.
Authors: Nikita Zozoulenko, Thomas Cass, Lukas Gonon
Abstract: The Mahalanobis distance is a classical tool used to measure the covariance-adjusted distance between points in $\bbR^d$. In this work, we extend the concept of Mahalanobis distance to separable Banach spaces by reinterpreting it as a Cameron-Martin norm associated with a probability measure. This approach leads to a basis-free, data-driven notion of anomaly distance through the so-called variance norm, which can naturally be estimated using empirical measures of a sample. Our framework generalizes the classical $\bbR^d$, functional $(L^2[0,1])^d$, and kernelized settings; importantly, it incorporates non-injective covariance operators. We prove that the variance norm is invariant under invertible bounded linear transformations of the data, extending previous results which are limited to unitary operators. In the Hilbert space setting, we connect the variance norm to the RKHS of the covariance operator, and establish consistency and convergence results for estimation using empirical measures with Tikhonov regularization. Using the variance norm, we introduce the notion of a kernelized nearest-neighbour Mahalanobis distance, and study some of its finite-sample concentration properties. In an empirical study on 12 real-world data sets, we demonstrate that the kernelized nearest-neighbour Mahalanobis distance outperforms the traditional kernelized Mahalanobis distance for multivariate time series novelty detection, using state-of-the-art time series kernels such as the signature, global alignment, and Volterra reservoir kernels.
Authors: Julio Candanedo
Abstract: Diffusion-Map-AutoEncoder (DMAE) pairs a diffusion-map encoder (using the Nystr\"om method) with linear or RBF Gaussian-Process latent mean decoders, yielding closed-form inductive mappings and strong reconstructions.
Authors: Hossein Sholehrasa, Majid Jaberi-Douraki
Abstract: Breast cancer's complexity and variability pose significant challenges in understanding its progression and guiding effective treatment. This study aims to integrate protein sequence data with expression levels to improve the molecular characterization of breast cancer subtypes and predict clinical outcomes. Using ProtGPT2, a language model specifically designed for protein sequences, we generated embeddings that capture the functional and structural properties of proteins. These embeddings were integrated with protein expression levels to form enriched biological representations, which were analyzed using machine learning methods, such as ensemble K-means for clustering and XGBoost for classification. Our approach enabled the successful clustering of patients into biologically distinct groups and accurately predicted clinical outcomes such as survival and biomarker status, achieving high performance metrics, notably an F1 score of 0.88 for survival and 0.87 for biomarker status prediction. Feature importance analysis identified KMT2C, CLASP2, and MYO1B as key proteins involved in hormone signaling, cytoskeletal remodeling, and therapy resistance in hormone receptor-positive and triple-negative breast cancer, with potential influence on breast cancer subtype behavior and progression. Furthermore, protein-protein interaction networks and correlation analyses revealed functional interdependencies among proteins that may influence the behavior and progression of breast cancer subtypes. These findings suggest that integrating protein sequence and expression data provides valuable insights into tumor biology and has significant potential to enhance personalized treatment strategies in breast cancer care.
Authors: Yongli Yan, Jieao Zhu, Tianyue Zheng, Zhuo Xu, Chao Jiang, Linglong Dai
Abstract: Channel coding is vital for reliable sixth-generation (6G) data transmission, employing diverse error correction codes for various application scenarios. Traditional decoders require dedicated hardware for each code, leading to high hardware costs. Recently, artificial intelligence (AI)-driven approaches, such as the error correction code Transformer (ECCT) and its enhanced version, the foundation error correction code Transformer (FECCT), have been proposed to reduce the hardware cost by leveraging the Transformer to decode multiple codes. However, their excessively high computational complexity of $\mathcal{O}(N^2)$ due to the self-attention mechanism in the Transformer limits scalability, where $N$ represents the sequence length. To reduce computational complexity, we propose a unified Transformer-based decoder that handles multiple linear block codes within a single framework. Specifically, a standardized unit is employed to align code length and code rate across different code types, while a redesigned low-rank unified attention module, with computational complexity of $\mathcal{O}(N)$, is shared across various heads in the Transformer. Additionally, a sparse mask, derived from the parity-check matrix's sparsity, is introduced to enhance the decoder's ability to capture inherent constraints between information and parity-check bits, improving decoding accuracy and further reducing computational complexity by $86\%$. Extensive experimental results demonstrate that the proposed unified Transformer-based decoder outperforms existing methods and provides a high-performance, low-complexity solution for next-generation wireless communication systems.
Authors: Zhengyu Hu, Jieyu Zhang, Zhihan Xiong, Alexander Ratner, Kaize Ding, Ranjay Krishna
Abstract: Despite the remarkable success of Large Language Models (LLMs), evaluating their outputs' quality regarding preference remains a critical challenge. While existing works usually leverage a strong LLM as the judge for comparing LLMs' response pairwisely, such a single-evaluator approach is vulnerable to cyclic preference, i.e., output A is better than B, B than C, but C is better than A, causing contradictory evaluation results. To address this, we introduce PGED (Preference Graph Ensemble and Denoise), a novel approach that leverages multiple model-based evaluators to construct preference graphs, and then ensembles and denoises these graphs for acyclic, non-contradictory evaluation results. We provide theoretical guarantees for our framework, demonstrating its efficacy in recovering the ground truth preference structure. Extensive experiments on ten benchmarks demonstrate PGED 's superiority in three applications: 1) model ranking for evaluation, 2) response selection for test-time scaling, and 3) data selection for model fine-tuning. Notably, PGED combines small LLM evaluators (e.g., Llama3-8B, Mistral-7B, Qwen2-7B) to outperform strong ones (e.g., Qwen2-72B), showcasing its effectiveness in enhancing evaluation reliability and improving model performance.
Authors: Mathis Koroglu, Hugo Caselles-Dupr\'e, Guillaume Jeanneret Sanmiguel, Matthieu Cord
Abstract: We consider the problem of text-to-video generation tasks with precise control for various applications such as camera movement control and video-to-video editing. Most methods tacking this problem rely on providing user-defined controls, such as binary masks or camera movement embeddings. In our approach we propose OnlyFlow, an approach leveraging the optical flow firstly extracted from an input video to condition the motion of generated videos. Using a text prompt and an input video, OnlyFlow allows the user to generate videos that respect the motion of the input video as well as the text prompt. This is implemented through an optical flow estimation model applied on the input video, which is then fed to a trainable optical flow encoder. The output feature maps are then injected into the text-to-video backbone model. We perform quantitative, qualitative and user preference studies to show that OnlyFlow positively compares to state-of-the-art methods on a wide range of tasks, even though OnlyFlow was not specifically trained for such tasks. OnlyFlow thus constitutes a versatile, lightweight yet efficient method for controlling motion in text-to-video generation. Models and code will be made available on GitHub and HuggingFace.
Authors: Zehao Li, Zhouchen Lin, Yijie Peng
Abstract: We study parameter inference in simulation-based stochastic models where the analytical form of the likelihood is unknown. The main difficulty is that score evaluation as a ratio of noisy Monte Carlo estimators induces bias and instability, which we overcome with a ratio-free nested multi-time-scale (NMTS) stochastic approximation (SA) method that simultaneously tracks the score and drives the parameter update. We provide a comprehensive theoretical analysis of the proposed NMTS algorithm for solving likelihood-free inference problems, including strong convergence, asymptotic normality, and convergence rates. We show that our algorithm can eliminate the original asymptotic bias $O\big(\sqrt{\frac{1}{N}}\big)$ and accelerate the convergence rate from $O\big(\beta_k+\sqrt{\frac{1}{N}}\big)$ to $O\big(\frac{\beta_k}{\alpha_k}+\sqrt{\frac{\alpha_k}{N}}\big)$, where $N$ is the fixed batch size, $\alpha_k$ and $\beta_k$ are decreasing step sizes with $\alpha_k$, $\beta_k$, $\beta_k/\alpha_k\rightarrow 0$. With proper choice of $\alpha_k$ and $\beta_k$, our convergence rates can match the optimal rate in the multi-time-scale SA literature. Numerical experiments demonstrate that our algorithm can improve the estimation accuracy by one to two orders of magnitude at the same computational cost, making it efficient for parameter estimation in stochastic systems.
Authors: Ting Sun, Penghan Wang, Fan Lai
Abstract: Large language models (LLMs) have facilitated a wide range of applications with distinct service-level objectives (SLOs), from latency-sensitive online tasks like interactive chatbots to throughput-oriented offline workloads like data synthesis. The existing deployment model, which dedicates machines to each workload, simplifies SLO management but often leads to poor resource utilization. This paper introduces HyGen, an interference-aware LLM serving system that enables efficient co-location of online and offline workloads while preserving SLOs. HyGen incorporates two key innovations: (1) performance control mechanisms, including a latency predictor to estimate batch execution time and an SLO-aware profiler to quantify latency interference, and (2) SLO-aware offline scheduling policies that maximize serving throughput and prevent starvation. Our evaluation on production workloads shows that HyGen achieves up to 3.9-5.8x throughput gains over online and hybrid serving baselines, while ensuring latency SLOs. The code of HyGen is publicly available at https://github.com/UIUC-MLSys/HyGen.
Authors: Ivica Nikolic, Teodora Baluta, Prateek Saxena
Abstract: Large language models are increasingly customized through fine-tuning and other adaptations, creating challenges in enforcing licensing terms and managing downstream impacts. Tracking model origins is crucial both for protecting intellectual property and for identifying derived models when biases or vulnerabilities are discovered in foundation models. We address this challenge by developing a framework for testing model provenance: Whether one model is derived from another. Our approach is based on the key observation that real-world model derivations preserve significant similarities in model outputs that can be detected through statistical analysis. Using only black-box access to models, we employ multiple hypothesis testing to compare model similarities against a baseline established by unrelated models. On two comprehensive real-world benchmarks spanning models from 30M to 4B parameters and comprising over 600 models, our tester achieves 90-95% precision and 80-90% recall in identifying derived models. These results demonstrate the viability of systematic provenance verification in production environments even when only API access is available.
Authors: Nikita Tsoy, Ivan Kirev, Negin Rahimiyazdi, Nikola Konstantinov
Abstract: Performativity, the phenomenon where outcomes are influenced by predictions, is particularly prevalent in social contexts where individuals strategically respond to a deployed model. In order to preserve the high accuracy of machine learning models under distribution shifts caused by performativity, Perdomo et al. (2020) introduced the concept of performative risk minimization (PRM). While this framework ensures model accuracy, it overlooks the impact of the PRM on the underlying distributions and the predictions of the model. In this paper, we initiate the analysis of the impact of PRM, by studying performativity for a sequential performative risk minimization problem with binary random variables and linear performative shifts. We formulate two natural measures of impact. In the case of full information, where the distribution dynamics are known, we derive explicit formulas for the PRM solution and our impact measures. In the case of partial information, we provide performative-aware statistical estimators, as well as simulations. Our analysis contrasts PRM to alternatives that do not model data shift and indicates that PRM can have amplified side effects compared to such methods.
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: Xuandong Zhao, Will Cai, Tianneng Shi, David Huang, Licong Lin, Song Mei, Dawn Song
Abstract: Existing training-time safety alignment techniques for large language models (LLMs) remain vulnerable to jailbreak attacks. Direct preference optimization (DPO), a widely deployed alignment method, exhibits limitations in both experimental and theoretical contexts as its loss function proves suboptimal for refusal learning. Through gradient-based analysis, we identify these shortcomings and propose an improved safety alignment that disentangles DPO objectives into two components: (1) robust refusal training, which encourages refusal even when partial unsafe generations are produced, and (2) targeted unlearning of harmful knowledge. This approach significantly increases LLM robustness against a wide range of jailbreak attacks, including prefilling, suffix, and multi-turn attacks across both in-distribution and out-of-distribution scenarios. Furthermore, we introduce a method to emphasize critical refusal tokens by incorporating a reward-based token-level weighting mechanism for refusal learning, which further improves the robustness against adversarial exploits. Our research also suggests that robustness to jailbreak attacks is correlated with token distribution shifts in the training process and internal representations of refusal and harmful tokens, offering valuable directions for future research in LLM safety alignment. The code is available at https://github.com/wicai24/DOOR-Alignment
Authors: Joohwi Lee, Kaito Miyamoto
Abstract: In the rapidly advancing field of materials informatics, nonlinear machine learning models have demonstrated exceptional predictive capabilities for material properties. However, their black-box nature limits interpretability, and they may incorporate features that do not contribute to, or even deteriorate, model performance. This study employs explainable ML (XML) techniques, including permutation feature importance and the SHapley Additive exPlanation, applied to a pristine support vector regression model designed to predict band gaps at the GW level using 18 input features. Guided by XML-derived individual feature importance, a simple framework is proposed to construct reduced-feature predictive models. Model evaluations indicate that an XML-guided compact model, consisting of the top five features, achieves comparable accuracy to the pristine model on in-domain datasets (0.254 vs. 0.247 eV) while demonstrating superior generalization with lower prediction errors on out-of-domain data (0.461 vs. 0.341 eV). Additionally, the study underscores the necessity for eliminating strongly correlated features (correlation coefficient greater than 0.8) to prevent misinterpretation and overestimation of feature importance before applying XML. This study highlights XML's effectiveness in developing simplified yet highly accurate machine learning models by clarifying feature roles, thereby reducing computational costs for feature acquisition and enhancing model trustworthiness for materials discovery.
Authors: Disheng Liu, Yiran Qiao, Wuche Liu, Yiren Lu, Yunlai Zhou, Tuo Liang, Yu Yin, Jing Ma
Abstract: True intelligence hinges on the ability to uncover and leverage hidden causal relations. Despite significant progress in AI and computer vision (CV), there remains a lack of benchmarks for assessing models' abilities to infer latent causality from complex visual data. In this paper, we introduce \textsc{\textbf{Causal3D}}, a novel and comprehensive benchmark that integrates structured data (tables) with corresponding visual representations (images) to evaluate causal reasoning. Designed within a systematic framework, Causal3D comprises 19 3D-scene datasets capturing diverse causal relations, views, and backgrounds, enabling evaluations across scenes of varying complexity. We assess multiple state-of-the-art methods, including classical causal discovery, causal representation learning, and large/vision-language models (LLMs/VLMs). Our experiments show that as causal structures grow more complex without prior knowledge, performance declines significantly, highlighting the challenges even advanced methods face in complex causal scenarios. Causal3D serves as a vital resource for advancing causal reasoning in CV and fostering trustworthy AI in critical domains.
Authors: Hao Zhu, Phil Cuvin, Xinkai Yu, Charlotte Ka Yee Yan, Jason Zhang, Diyi Yang
Abstract: Agents are predominantly evaluated and optimized via task success metrics, which are coarse, rely on manual design from experts, and fail to reward intermediate emergent behaviors. We propose **AutoLibra**, a framework for agent evaluation, that transforms open-ended human feedback *e.g.* "If you find that the button is disabled, don't click it again", or "This agent has too much autonomy to decide what to do on its own" into metrics for evaluating fine-grained behaviors in agent trajectories. AutoLibra accomplishes this by grounding feedback to an agent's behavior, clustering similar positive and negative behaviors, and creating concrete metrics with clear definitions and concrete examples, which can be used for prompting LLM-as-a-Judge as evaluators. We further propose two meta metrics to evaluate the alignment of a set of (induced) metrics with open feedback: "coverage" and "redundancy". Through optimizing these meta-metrics, we experimentally demonstrate AutoLibra's ability to induce more concrete agent evaluation metrics than the ones proposed in previous agent evaluation benchmarks and discover new metrics to analyze agents. We also present two applications of AutoLibra in agent improvement: First, we show that AutoLibra serve human prompt engineers for diagonalize agent failures and improve prompts iterative. Moreover, we find that AutoLibra can induce metrics for automatic optimization for agents, which makes agents improve through self-regulation. Our results suggest that AutoLibra is a powerful task-agnostic tool for evaluating and improving language agents.
Authors: David Abel, Michael Bowling, Andr\'e Barreto, Will Dabney, Shi Dong, Steven Hansen, Anna Harutyunyan, Khimya Khetarpal, Clare Lyle, Razvan Pascanu, Georgios Piliouras, Doina Precup, Jonathan Richens, Mark Rowland, Tom Schaul, Satinder Singh
Abstract: Agents are minimally entities that are influenced by their past observations and act to influence future observations. This latter capacity is captured by empowerment, which has served as a vital framing concept across artificial intelligence and cognitive science. This former capacity, however, is equally foundational: In what ways, and to what extent, can an agent be influenced by what it observes? In this paper, we ground this concept in a universal agent-centric measure that we refer to as plasticity, and reveal a fundamental connection to empowerment. Following a set of desiderata on a suitable definition, we define plasticity using a new information-theoretic quantity we call the generalized directed information. We show that this new quantity strictly generalizes the directed information introduced by Massey (1990) while preserving all of its desirable properties. Under this definition, we find that plasticity is well thought of as the mirror of empowerment: The two concepts are defined using the same measure, with only the direction of influence reversed. Our main result establishes a tension between the plasticity and empowerment of an agent, suggesting that agent design needs to be mindful of both characteristics. We explore the implications of these findings, and suggest that plasticity, empowerment, and their relationship are essential to understanding agency
Authors: Raja Gond, Nipun Kwatra, Ramachandran Ramjee
Abstract: Distributed inference of large language models (LLMs) can introduce overheads of up to 20% even over GPUs connected via high-speed interconnects such as NVLink. Multiple techniques have been proposed to mitigate these overheads by decomposing computations into finer-grained tasks and overlapping communication with sub-tasks as they complete. However, fine-grained decomposition of a large computation into many smaller computations on GPUs results in overheads. Furthermore, the communication itself uses many streaming multiprocessors (SMs), adding to the overhead. We present TokenWeave to address these challenges. TokenWeave proposes a Token-Splitting technique that divides the tokens in the inference batch into two approximately equal subsets in a wave-aware manner. The communication of one subset is then overlapped with the computation of the other. In addition, TokenWeave optimizes the order of the layer normalization computation with respect to communication operations and implements a novel fused AllReduce--RMSNorm kernel that carefully leverages Multimem instruction support available on Hopper and Blackwell NVIDIA GPUs. These optimizations allow TokenWeave to perform communication and RMSNorm using only 2-8 SMs. Moreover, our kernel enables the memory-bound RMSNorm to be overlapped with the other batch's computation, providing additional gains. Our evaluations demonstrate up to 1.29x speedup in latency and 1.26x higher throughput across multiple models and workloads. In several settings, TokenWeave results in better performance compared to an equivalent model with all communication removed.
Authors: Jose Fuentes, Ines Ortega-Fernandez, Nora M. Villanueva, Marta Sestelo
Abstract: User and Entity Behaviour Analytics (UEBA) is a broad branch of data analytics that attempts to build a normal behavioural profile in order to detect anomalous events. Among the techniques used to detect anomalies, Deep Autoencoders constitute one of the most promising deep learning models on UEBA tasks, allowing explainable detection of security incidents that could lead to the leak of personal data, hijacking of systems, or access to sensitive business information. In this study, we introduce the first implementation of an explainable UEBA-based anomaly detection framework that leverages Deep Autoencoders in combination with Doc2Vec to process both numerical and textual features. Additionally, based on the theoretical foundations of neural networks, we offer a novel proof demonstrating the equivalence of two widely used definitions for fully-connected neural networks. The experimental results demonstrate the proposed framework capability to detect real and synthetic anomalies effectively generated from real attack data, showing that the models provide not only correct identification of anomalies but also explainable results that enable the reconstruction of the possible origin of the anomaly. Our findings suggest that the proposed UEBA framework can be seamlessly integrated into enterprise environments, complementing existing security systems for explainable threat detection.
Authors: Hao Mark Chen, Guanxi Lu, Yasuyuki Okoshi, Zhiwen Mo, Masato Motomura, Hongxiang Fan
Abstract: Test-time scaling (TTS) has proven effective in enhancing the reasoning capabilities of large language models (LLMs). Verification plays a key role in TTS, simultaneously influencing (1) reasoning performance and (2) compute efficiency, due to the quality and computational cost of verification. In this work, we challenge the conventional paradigms of verification, and make the first attempt toward systematically investigating the impact of verification granularity-that is, how frequently the verifier is invoked during generation, beyond verifying only the final output or individual generation steps. To this end, we introduce Variable Granularity Search (VG-Search), a unified algorithm that generalizes beam search and Best-of-N sampling via a tunable granularity parameter g. Extensive experiments with VG-Search under varying compute budgets, generator-verifier configurations, and task attributes reveal that dynamically selecting g can improve the compute efficiency and scaling behavior. Building on these findings, we propose adaptive VG-Search strategies that achieve accuracy gains of up to 3.1\% over Beam Search and 3.6\% over Best-of-N, while reducing FLOPs by over 52\%. We will open-source the code to support future research.
Authors: Wenquan Lu, Jiaqi Zhang, Hugues Van Assel, Randall Balestriero
Abstract: Self-Supervised Learning (SSL) has become a powerful solution to extract rich representations from unlabeled data. Yet, SSL research is mostly focused on clean, curated and high-quality datasets. As a result, applying SSL on noisy data remains a challenge, despite being crucial to applications such as astrophysics, medical imaging, geophysics or finance. In this work, we present a fully self-supervised framework that enables noise-robust representation learning without requiring a denoiser at inference or downstream fine-tuning. Our method first trains an SSL denoiser on noisy data, then uses it to construct a denoised-to-noisy data curriculum (i.e., training first on denoised, then noisy samples) for pretraining a SSL backbone (e.g., DINOv2), combined with a teacher-guided regularization that anchors noisy embeddings to their denoised counterparts. This process encourages the model to internalize noise robustness. Notably, the denoiser can be discarded after pretraining, simplifying deployment. On ImageNet-1k with ViT-B under extreme Gaussian noise ($\sigma=255$, SNR = 0.72 dB), our method improves linear probing accuracy by 4.8% over DINOv2, demonstrating that denoiser-free robustness can emerge from noise-aware pretraining. The code is available at https://github.com/wenquanlu/noisy_dinov2.
Authors: Yinghao Zhu, Ziyi He, Haoran Hu, Xiaochen Zheng, Xichen Zhang, Zixiang Wang, Junyi Gao, Liantao Ma, Lequan Yu
Abstract: The rapid advancement of Large Language Models (LLMs) has stimulated interest in multi-agent collaboration for addressing complex medical tasks. However, the practical advantages of multi-agent collaboration approaches remain insufficiently understood. Existing evaluations often lack generalizability, failing to cover diverse tasks reflective of real-world clinical practice, and frequently omit rigorous comparisons against both single-LLM-based and established conventional methods. To address this critical gap, we introduce MedAgentBoard, a comprehensive benchmark for the systematic evaluation of multi-agent collaboration, single-LLM, and conventional approaches. MedAgentBoard encompasses four diverse medical task categories: (1) medical (visual) question answering, (2) lay summary generation, (3) structured Electronic Health Record (EHR) predictive modeling, and (4) clinical workflow automation, across text, medical images, and structured EHR data. Our extensive experiments reveal a nuanced landscape: while multi-agent collaboration demonstrates benefits in specific scenarios, such as enhancing task completeness in clinical workflow automation, it does not consistently outperform advanced single LLMs (e.g., in textual medical QA) or, critically, specialized conventional methods that generally maintain better performance in tasks like medical VQA and EHR-based prediction. MedAgentBoard offers a vital resource and actionable insights, emphasizing the necessity of a task-specific, evidence-based approach to selecting and developing AI solutions in medicine. It underscores that the inherent complexity and overhead of multi-agent collaboration must be carefully weighed against tangible performance gains. All code, datasets, detailed prompts, and experimental results are open-sourced at https://medagentboard.netlify.app/.
Authors: Xiaoyin Chen, Jiarui Lu, Minsu Kim, Dinghuai Zhang, Jian Tang, Alexandre Pich\'e, Nicolas Gontier, Yoshua Bengio, Ehsan Kamalloo
Abstract: Reinforcement learning (RL) has proven effective for fine-tuning large language models (LLMs), significantly enhancing their reasoning abilities in domains such as mathematics and code generation. A crucial factor influencing RL fine-tuning success is the training curriculum: the order in which training problems are presented. While random curricula serve as common baselines, they remain suboptimal; manually designed curricula often rely heavily on heuristics, and online filtering methods can be computationally prohibitive. To address these limitations, we propose Self-Evolving Curriculum (SEC), an automatic curriculum learning method that learns a curriculum policy concurrently with the RL fine-tuning process. Our approach formulates curriculum selection as a non-stationary Multi-Armed Bandit problem, treating each problem category (e.g., difficulty level or problem type) as an individual arm. We leverage the absolute advantage from policy gradient methods as a proxy measure for immediate learning gain. At each training step, the curriculum policy selects categories to maximize this reward signal and is updated using the TD(0) method. Across three distinct reasoning domains: planning, inductive reasoning, and mathematics, our experiments demonstrate that SEC significantly improves models' reasoning capabilities, enabling better generalization to harder, out-of-distribution test problems. Additionally, our approach achieves better skill balance when fine-tuning simultaneously on multiple reasoning domains. These findings highlight SEC as a promising strategy for RL fine-tuning of LLMs.
Authors: Florian Kalinke, Shakeel Gavioli-Akilagun
Abstract: This article studies the problem of online non-parametric change point detection in multivariate data streams. We approach the problem through the lens of kernel-based two-sample testing and introduce a sequential testing procedure based on random Fourier features, running with logarithmic time complexity per observation and with overall logarithmic space complexity. The algorithm has two advantages compared to the state of the art. First, our approach is genuinely online, and no access to training data known to be from the pre-change distribution is necessary. Second, the algorithm does not require the user to specify a window parameter over which local tests are to be calculated. We prove strong theoretical guarantees on the algorithm's performance, including information-theoretic bounds demonstrating that the detection delay is optimal in the minimax sense. Numerical studies on real and synthetic data show that our algorithm is competitive with respect to the state of the art.
Authors: Jigang Fan, Quanlin Wu, Shengjie Luo, Liwei Wang
Abstract: The detection of ligand binding sites for proteins is a fundamental step in Structure-Based Drug Design. Despite notable advances in recent years, existing methods, datasets, and evaluation metrics are confronted with several key challenges: (1) current datasets and methods are centered on individual protein-ligand complexes and neglect that diverse binding sites may exist across multiple complexes of the same protein, introducing significant statistical bias; (2) ligand binding site detection is typically modeled as a discontinuous workflow, employing binary segmentation and subsequent clustering algorithms; (3) traditional evaluation metrics do not adequately reflect the actual performance of different binding site prediction methods. To address these issues, we first introduce UniSite-DS, the first UniProt (Unique Protein)-centric ligand binding site dataset, which contains 4.81 times more multi-site data and 2.08 times more overall data compared to the previously most widely used datasets. We then propose UniSite, the first end-to-end ligand binding site detection framework supervised by set prediction loss with bijective matching. In addition, we introduce Average Precision based on Intersection over Union (IoU) as a more accurate evaluation metric for ligand binding site prediction. Extensive experiments on UniSite-DS and several representative benchmark datasets demonstrate that IoU-based Average Precision provides a more accurate reflection of prediction quality, and that UniSite outperforms current state-of-the-art methods in ligand binding site detection. The dataset and codes will be made publicly available at https://github.com/quanlin-wu/unisite.
Authors: Fuat Can Beylunioglu, Mehrdad Pirnia, P. Robert Duimering
Abstract: Neural Networks (NN) with ReLU activation functions are used to model multiparametric quadratic optimization problems (mp-QP) in diverse engineering applications. Researchers have suggested leveraging the piecewise affine property of deep NN models to solve mp-QP with linear constraints, which also exhibit piecewise affine behaviour. However, traditional deep NN applications to mp-QP fall short of providing optimal and feasible predictions, even when trained on large datasets. This study proposes a partially-supervised NN (PSNN) architecture that directly represents the mathematical structure of the global solution function. In contrast to generic NN training approaches, the proposed PSNN method derives a large proportion of model weights directly from the mathematical properties of the optimization problem, producing more accurate solutions despite significantly smaller training data sets. Many energy management problems are formulated as QP, so we apply the proposed approach to energy systems (specifically DC optimal power flow) to demonstrate proof of concept. Model performance in terms of solution accuracy and speed of predictions was compared against a commercial solver and a generic Deep NN model based on classical training. Results show KKT sufficient conditions for PSNN consistently outperform generic NN architectures with classical training using far less data, including when tested on extreme, out-of-training distribution test data. Given its speed advantages over traditional solvers, the PSNN model can quickly produce optimal and feasible solutions within a second for millions of input parameters sampled from a distribution of stochastic demands and renewable generator dispatches, which can be used for simulations and long term planning.
Authors: Mohammad Jalali, Haoyu Lei, Amin Gohari, Farzan Farnia
Abstract: Diffusion models have demonstrated remarkable success in high-fidelity image synthesis and prompt-guided generative modeling. However, ensuring adequate diversity in generated samples of prompt-guided diffusion models remains a challenge, particularly when the prompts span a broad semantic spectrum and the diversity of generated data needs to be evaluated in a prompt-aware fashion across semantically similar prompts. Recent methods have introduced guidance via diversity measures to encourage more varied generations. In this work, we extend the diversity measure-based approaches by proposing the Scalable Prompt-Aware R\'eny Kernel Entropy Diversity Guidance (SPARKE) method for prompt-aware diversity guidance. SPARKE utilizes conditional entropy for diversity guidance, which dynamically conditions diversity measurement on similar prompts and enables prompt-aware diversity control. While the entropy-based guidance approach enhances prompt-aware diversity, its reliance on the matrix-based entropy scores poses computational challenges in large-scale generation settings. To address this, we focus on the special case of Conditional latent RKE Score Guidance, reducing entropy computation and gradient-based optimization complexity from the $O(n^3)$ of general entropy measures to $O(n)$. The reduced computational complexity allows for diversity-guided sampling over potentially thousands of generation rounds on different prompts. We numerically test the SPARKE method on several text-to-image diffusion models, demonstrating that the proposed method improves the prompt-aware diversity of the generated data without incurring significant computational costs. We release our code on the project page: https://mjalali.github.io/SPARKE
Authors: Hongzhen Huang, Kunming Zhang, Hanlong Liao, Kui Wu, Guoming Tang
Abstract: The rapid advancement of AI, particularly large language models (LLMs), has raised significant concerns about the energy use and carbon emissions associated with model training and inference. However, existing tools for measuring and reporting such impacts are often fragmented, lacking systematic metric integration and offering limited support for correlation analysis among them. This paper presents AIMeter, a comprehensive software toolkit for the measurement, analysis, and visualization of energy use, power draw, hardware performance, and carbon emissions across AI workloads. By seamlessly integrating with existing AI frameworks, AIMeter offers standardized reports and exports fine-grained time-series data to support benchmarking and reproducibility in a lightweight manner. It further enables in-depth correlation analysis between hardware metrics and model performance and thus facilitates bottleneck identification and performance enhancement. By addressing critical limitations in existing tools, AIMeter encourages the research community to weigh environmental impact alongside raw performance of AI workloads and advances the shift toward more sustainable "Green AI" practices. The code is available at https://github.com/SusCom-Lab/AIMeter.
Authors: Byunghee Lee, Hye Yeon Sin, Joonsung Kang
Abstract: This study introduces an integrated framework for predictive causal inference designed to overcome limitations inherent in conventional single model approaches. Specifically, we combine a Hidden Markov Model (HMM) for spatial health state estimation with a Multi Task and Multi Graph Convolutional Network (MTGCN) for capturing temporal outcome trajectories. The framework asymmetrically treats temporal and spatial information regarding them as endogenous variables in the outcome regression, and exogenous variables in the propensity score model, thereby expanding the standard doubly robust treatment effect estimation to jointly enhance bias correction and predictive accuracy. To demonstrate its utility, we focus on clinical domains such as cancer, dementia, and Parkinson disease, where treatment effects are challenging to observe directly. Simulation studies are conducted to emulate latent disease dynamics and evaluate the model performance under varying conditions. Overall, the proposed framework advances predictive causal inference by structurally adapting to spatiotemporal complexities common in biomedical data.
Authors: Dirk van der Hoeven, Julia Olkhovskaia, Tim van Erven
Abstract: We consider the fundamental problem of estimating a discrete distribution on a domain of size~$K$ with high probability in Kullback-Leibler divergence. We provide upper and lower bounds on the minimax estimation rate, which show that the optimal rate is between $\big(K + \ln(K)\ln(1/\delta)\big) /n$ and $\big(K\ln\ln(K) + \ln(K)\ln(1/\delta)\big) /n$ at error probability $\delta$ and sample size $n$, which pins down the rate up to the doubly logarithmic factor $\ln \ln K$ that multiplies $K$. Our upper bound uses techniques from online learning to construct a novel estimator via online-to-batch conversion. Perhaps surprisingly, the tail behavior of the minimax rate is worse than for the squared total variation and squared Hellinger distance, for which it is $\big(K + \ln(1/\delta)\big) /n$, i.e.\ without the $\ln K$ multiplying $\ln (1/\delta)$. As a consequence, we cannot obtain a fully tight lower bound from the usual reduction to these smaller distances. Moreover, we show that this lower bound cannot be achieved by the standard lower bound approach based on a reduction to hypothesis testing, and instead we need to introduce a new reduction to what we call weak hypothesis testing. We investigate the source of the gap with other divergences further in refined results, which show that the total variation rate is achievable for Kullback-Leibler divergence after all (in fact by he maximum likelihood estimator) if we rule out outcome probabilities smaller than $O(\ln(K/\delta) / n)$, which is a vanishing set as $n$ increases for fixed $K$ and~$\delta$. This explains why minimax Kullback-Leibler estimation is more difficult than asymptotic estimation.
Authors: Fabio Magarelli, Geraldine B. Boylan, Saeed Montazeri, Feargal O'Sullivan, Dominic Lightbody, Minoo Ashoori, Tamara Skoric, John M. O'Toole
Abstract: Machine learning (ML) has the potential to support and improve expert performance in monitoring the brain function of at-risk newborns. Developing accurate and reliable ML models depends on access to high-quality, annotated data, a resource in short supply. ML competitions address this need by providing researchers access to expertly annotated datasets, fostering shared learning through direct model comparisons, and leveraging the benefits of crowdsourcing diverse expertise. We compiled a retrospective dataset containing 353 hours of EEG from 102 individual newborns from a multi-centre study. The data was fully anonymised and divided into training, testing, and held-out validation datasets. EEGs were graded for the severity of abnormal background patterns. Next, we created a web-based competition platform and hosted a machine learning competition to develop ML models for classifying the severity of EEG background patterns in newborns. After the competition closed, the top 4 performing models were evaluated offline on a separate held-out validation dataset. Although a feature-based model ranked first on the testing dataset, deep learning models generalised better on the validation sets. All methods had a significant decline in validation performance compared to the testing performance. This highlights the challenges for model generalisation on unseen data, emphasising the need for held-out validation datasets in ML studies with neonatal EEG. The study underscores the importance of training ML models on large and diverse datasets to ensure robust generalisation. The competition's outcome demonstrates the potential for open-access data and collaborative ML development to foster a collaborative research environment and expedite the development of clinical decision-support tools for neonatal neuromonitoring.
Authors: Jan Philipp Schneider, Pratik Singh Bisht, Ilya Chugunov, Andreas Kolb, Michael Moeller, Felix Heide
Abstract: Learning editable high-resolution scene representations for dynamic scenes is an open problem with applications across the domains from autonomous driving to creative editing - the most successful approaches today make a trade-off between editability and supporting scene complexity: neural atlases represent dynamic scenes as two deforming image layers, foreground and background, which are editable in 2D, but break down when multiple objects occlude and interact. In contrast, scene graph models make use of annotated data such as masks and bounding boxes from autonomous-driving datasets to capture complex 3D spatial relationships, but their implicit volumetric node representations are challenging to edit view-consistently. We propose Neural Atlas Graphs (NAGs), a hybrid high-resolution scene representation, where every graph node is a view-dependent neural atlas, facilitating both 2D appearance editing and 3D ordering and positioning of scene elements. Fit at test-time, NAGs achieve state-of-the-art quantitative results on the Waymo Open Dataset - by 5 dB PSNR increase compared to existing methods - and make environmental editing possible in high resolution and visual quality - creating counterfactual driving scenarios with new backgrounds and edited vehicle appearance. We find that the method also generalizes beyond driving scenes and compares favorably - by more than 7 dB in PSNR - to recent matting and video editing baselines on the DAVIS video dataset with a diverse set of human and animal-centric scenes. Project Page: https://princeton-computational-imaging.github.io/nag/
URLs: https://princeton-computational-imaging.github.io/nag/
Authors: Debarpan Bhattacharya, Apoorva Kulkarni, Sriram Ganapathy
Abstract: The accurate trust assessment of multimodal large language models (MLLMs) generated predictions, which can enable selective prediction and improve user confidence, is challenging due to the diverse multi-modal input paradigms. We propose Functionally Equivalent Sampling for Trust Assessment (FESTA), a multimodal input sampling technique for MLLMs, that generates an uncertainty measure based on the equivalent and complementary input samplings. The proposed task-preserving sampling approach for uncertainty quantification expands the input space to probe the consistency (through equivalent samples) and sensitivity (through complementary samples) of the model. FESTA uses only input-output access of the model (black-box), and does not require ground truth (unsupervised). The experiments are conducted with various off-the-shelf multi-modal LLMs, on both visual and audio reasoning tasks. The proposed FESTA uncertainty estimate achieves significant improvement (33.3% relative improvement for vision-LLMs and 29.6% relative improvement for audio-LLMs) in selective prediction performance, based on area-under-receiver-operating-characteristic curve (AUROC) metric in detecting mispredictions. The code implementation is open-sourced.
Authors: Yuanrong Wang, Yingpeng Du
Abstract: Recommender systems (RS) greatly influence users' consumption decisions, making them attractive targets for malicious shilling attacks that inject fake user profiles to manipulate recommendations. Existing shilling methods can generate effective and stealthy fake profiles when training data only contain rating matrix, but they lack comprehensive solutions for scenarios where side features are present and utilized by the recommender. To address this gap, we extend the Leg-UP framework by enhancing the generator architecture to incorporate side features, enabling the generation of side-feature-aware fake user profiles. Experiments on benchmarks show that our method achieves strong attack performance while maintaining stealthiness.
Authors: Enhao Huang, Zhiyu Zhang, Tianxiang Xu, Chunshu Xia, Kaichun Hu, Yuchen Yang, Tongtong Pan, Dong Dong, Zhan Qin
Abstract: Complex-valued signals encode both amplitude and phase, yet most deep models treat attention as real-valued correlation, overlooking interference effects. We introduce the Holographic Transformer, a physics-inspired architecture that incorporates wave interference principles into self-attention. Holographic attention modulates interactions by relative phase and coherently superimposes values, ensuring consistency between amplitude and phase. A dual-headed decoder simultaneously reconstructs the input and predicts task outputs, preventing phase collapse when losses prioritize magnitude over phase. We demonstrate that holographic attention implements a discrete interference operator and maintains phase consistency under linear mixing. Experiments on PolSAR image classification and wireless channel prediction show strong performance, achieving high classification accuracy and F1 scores, low regression error, and increased robustness to phase perturbations. These results highlight that enforcing physical consistency in attention leads to generalizable improvements in complex-valued learning and provides a unified, physics-based framework for coherent signal modeling. The code is available at https://github.com/EonHao/Holographic-Transformers.
Authors: Zhilin Wang, Jiaqi Zeng, Olivier Delalleau, Ellie Evans, Daniel Egert, Hoo-Chang Shin, Felipe Soares, Yi Dong, Oleksii Kuchaiev
Abstract: Reinforcement Learning with Human Feedback (RLHF) and Reinforcement Learning with Verifiable Rewards (RLVR) are the main RL paradigms used in LLM post-training, each offering distinct advantages. However, RLHF struggles with interpretability and reward hacking because it relies on human judgments that usually lack explicit criteria, whereas RLVR is limited in scope by its focus on correctness-based verifiers. We propose Reinforcement Learning with Binary Flexible Feedback (RLBFF), which combines the versatility of human-driven preferences with the precision of rule-based verification, enabling reward models to capture nuanced aspects of response quality beyond mere correctness. RLBFF extracts principles that can be answered in a binary fashion (e.g. accuracy of information: yes, or code readability: no) from natural language feedback. Such principles can then be used to ground Reward Model training as an entailment task (response satisfies or does not satisfy an arbitrary principle). We show that Reward Models trained in this manner can outperform Bradley-Terry models when matched for data and achieve top performance on RM-Bench (86.2%) and JudgeBench (81.4%, #1 on leaderboard as of September 24, 2025). Additionally, users can specify principles of interest at inference time to customize the focus of our reward models, in contrast to Bradley-Terry models. Finally, we present a fully open source recipe (including data) to align Qwen3-32B using RLBFF and our Reward Model, to match or exceed the performance of o3-mini and DeepSeek R1 on general alignment benchmarks of MT-Bench, WildBench, and Arena Hard v2 (at <5% of the inference cost). Models: https://huggingface.co/collections/nvidia/reward-models-10-2025
URLs: https://huggingface.co/collections/nvidia/reward-models-10-2025
Authors: Emmanuel de Salis, Massimo De Santis, Davide Piras, Sambit K. Giri, Michele Bianco, Nicolas Cerardi, Philipp Denzel, Merve Selcuk-Simsek, Kelley M. Hess, M. Carmen Toribio, Franz Kirsten, Hatem Ghorbel
Abstract: Hydrogen is the most abundant element in our Universe. The first generation of stars and galaxies produced photons that ionized hydrogen gas, driving a cosmological event known as the Epoch of Reionization (EoR). The upcoming Square Kilometre Array Observatory (SKAO) will map the distribution of neutral hydrogen during this era, aiding in the study of the properties of these first-generation objects. Extracting astrophysical information will be challenging, as SKAO will produce a tremendous amount of data where the hydrogen signal will be contaminated with undesired foreground contamination and instrumental systematics. To address this, we develop the latest deep learning techniques to extract information from the 2D power spectra of the hydrogen signal expected from SKAO. We apply a series of neural network models to these measurements and quantify their ability to predict the history of cosmic hydrogen reionization, which is connected to the increasing number and efficiency of early photon sources. We show that the study of the early Universe benefits from modern deep learning technology. In particular, we demonstrate that dedicated machine learning algorithms can achieve more than a $0.95$ $R^2$ score on average in recovering the reionization history. This enables accurate and precise cosmological and astrophysical inference of structure formation in the early Universe.
Authors: Ke Wang, Zishuo Zhao, Xinyuan Song, Bill Shi, Libin Xia, Chris Tong, Lynn Ai, Felix Qu, Eric Yang
Abstract: Decentralized inference provides a scalable and resilient paradigm for serving large language models (LLMs), enabling distributed resource utilization and reducing reliance on centralized providers. However, in a permissionless environment without trusted nodes, ensuring the correctness of model outputs remains a core challenge. We introduce VeriLLM, a publicly verifiable protocol for decentralized LLM inference that achieves security under a one-honest-verifier assumption while maintaining practical efficiency. VeriLLM combines lightweight empirical rerunning with cryptographic commitments, allowing verifiers to validate results at approximately 1% of the underlying inference cost. To prevent verification bottlenecks, we design an isomorphic inference-verification architecture that multiplexes both inference and verification roles across the same GPU workers. This design (i) improves GPU utilization and overall throughput, (ii) enlarges the effective validator set, enhancing robustness and liveness, and (iii) enforces task indistinguishability to prevent node-specific optimizations or selective behavior. Through theoretical analysis and system-level evaluation, we show that VeriLLM achieves reliable public verifiability with minimal overhead, offering a practical foundation for trustworthy and scalable decentralized LLM inference.
Authors: Dustin Wright, Sarah Masud, Jared Moore, Srishti Yadav, Maria Antoniak, Chan Young Park, Isabelle Augenstein
Abstract: Large language models (LLMs) tend to generate lexically, semantically, and stylistically homogenous texts. This poses a risk of knowledge collapse, where homogenous LLMs mediate a shrinking in the range of accessible information over time. Existing works on homogenization are limited by a focus on closed-ended multiple-choice setups or fuzzy semantic features, and do not look at trends across time and cultural contexts. To overcome this, we present a new methodology to measure epistemic diversity, i.e., variation in real-world claims in LLM outputs, which we use to perform a broad empirical study of LLM knowledge collapse. We test 27 LLMs, 155 topics covering 12 countries, and 200 prompt variations sourced from real user chats. For the topics in our study, we show that while newer models tend to generate more diverse claims, nearly all models are less epistemically diverse than a basic web search. We find that model size has a negative impact on epistemic diversity, while retrieval-augmented generation (RAG) has a positive impact, though the improvement from RAG varies by the cultural context. Finally, compared to a traditional knowledge source (Wikipedia), we find that country-specific claims reflect the English language more than the local one, highlighting a gap in epistemic representation
Authors: Xuanming Cui, Jianpeng Cheng, Hong-you Chen, Satya Narayan Shukla, Abhijeet Awasthi, Xichen Pan, Chaitanya Ahuja, Shlok Kumar Mishra, Yonghuan Yang, Jun Xiao, Qi Guo, Ser-Nam Lim, Aashu Singh, Xiangjun Fan
Abstract: There is a growing interest in Universal Multimodal Embeddings (UME), where models are required to generate task-specific representations. While recent studies show that Multimodal Large Language Models (MLLMs) perform well on such tasks, they treat MLLMs solely as encoders, overlooking their generative capacity. However, such an encoding paradigm becomes less effective as instructions become more complex and require compositional reasoning. Inspired by the proven effectiveness of chain-of-thought reasoning, we propose a general Think-Then-Embed (TTE) framework for UME, composed of a reasoner and an embedder. The reasoner MLLM first generates reasoning traces that explain complex queries, followed by an embedder that produces representations conditioned on both the original query and the intermediate reasoning. This explicit reasoning step enables more nuanced understanding of complex multimodal instructions. Our contributions are threefold. First, by leveraging a powerful MLLM reasoner, we achieve state-of-the-art performance on the MMEB-V2 benchmark, surpassing proprietary models trained on massive in-house datasets. Second, to reduce the dependency on large MLLM reasoners, we finetune a smaller MLLM reasoner using high-quality embedding-centric reasoning traces, achieving the best performance among open-source models with a 7% absolute gain over recently proposed models. Third, we investigate strategies for integrating the reasoner and embedder into a unified model for improved efficiency without sacrificing performance.
Authors: Raffaele Mura, Giorgio Piras, Kamil\.e Luko\v{s}i\=ut\.e, Maura Pintor, Amin Karbasi, Battista Biggio
Abstract: Jailbreaks are adversarial attacks designed to bypass the built-in safety mechanisms of large language models. Automated jailbreaks typically optimize an adversarial suffix or adapt long prompt templates by forcing the model to generate the initial part of a restricted or harmful response. In this work, we show that existing jailbreak attacks that leverage such mechanisms to unlock the model response can be detected by a straightforward perplexity-based filtering on the input prompt. To overcome this issue, we propose LatentBreak, a white-box jailbreak attack that generates natural adversarial prompts with low perplexity capable of evading such defenses. LatentBreak substitutes words in the input prompt with semantically-equivalent ones, preserving the initial intent of the prompt, instead of adding high-perplexity adversarial suffixes or long templates. These words are chosen by minimizing the distance in the latent space between the representation of the adversarial prompt and that of harmless requests. Our extensive evaluation shows that LatentBreak leads to shorter and low-perplexity prompts, thus outperforming competing jailbreak algorithms against perplexity-based filters on multiple safety-aligned models.
Authors: Ali Atiah Alzahrani
Abstract: We tackle high-dimensional, path-dependent valuation and control and introduce a deep BSDE/2BSDE solver that couples truncated log-signatures with a neural rough differential equation (RDE) backbone. The architecture aligns stochastic analysis with sequence-to-path learning: a CVaR-tilted terminal objective targets left-tail risk, while an optional second-order (2BSDE) head supplies curvature estimates for risk-sensitive control. Under matched compute and parameter budgets, the method improves accuracy, tail fidelity, and training stability across Asian and barrier option pricing and portfolio control: at d=200 it achieves CVaR(0.99)=9.80% versus 12.00-13.10% for strong baselines, attains the lowest HJB residual (0.011), and yields the lowest RMSEs for Z and Gamma. Ablations over truncation depth, local windows, and tilt parameters confirm complementary gains from the sequence-to-path representation and the 2BSDE head. Taken together, the results highlight a bidirectional dialogue between stochastic analysis and modern deep learning: stochastic tools inform representations and objectives, while sequence-to-path models expand the class of solvable financial models at scale.
Authors: Gabriel Fiastre, Antoine Yang, Cordelia Schmid
Abstract: Dense Video Object Captioning (DVOC) is the task of jointly detecting, tracking, and captioning object trajectories in a video, requiring the ability to understand spatio-temporal details and describe them in natural language. Due to the complexity of the task and the high cost associated with manual annotation, previous approaches resort to disjoint training strategies, potentially leading to suboptimal performance. To circumvent this issue, we propose to generate captions about spatio-temporally localized entities leveraging a state-of-the-art VLM. By extending the LVIS and LV-VIS datasets with our synthetic captions (LVISCap and LV-VISCap), we train MaskCaptioner, an end-to-end model capable of jointly detecting, segmenting, tracking and captioning object trajectories. Moreover, with pretraining on LVISCap and LV-VISCap, MaskCaptioner achieves state-of-the-art DVOC results on three existing benchmarks, VidSTG, VLN and BenSMOT. The datasets and code are available at https://www.gabriel.fiastre.fr/maskcaptioner/.
Authors: Fernando Vallecillos-Ruiz, Max Hort, Leon Moonen
Abstract: Today's pursuit of a single Large Language Model (LMM) for all software engineering tasks is resource-intensive and overlooks the potential benefits of complementarity, where different models contribute unique strengths. However, the degree to which coding LLMs complement each other and the best strategy for maximizing an ensemble's potential are unclear, leaving practitioners without a clear path to move beyond single-model systems. To address this gap, we empirically compare ten individual LLMs from five families, and three ensembles of these LLMs across three software engineering benchmarks covering code generation and program repair. We assess the complementarity between models and the performance gap between the best individual model and the ensembles. Next, we evaluate various selection heuristics to identify correct solutions from an ensemble's candidate pool. We find that the theoretical upperbound for an ensemble's performance can be 83% above the best single model. Our results show that consensus-based strategies for selecting solutions fall into a "popularity trap," amplifying common but incorrect outputs. In contrast, a diversity-based strategy realizes up to 95% of this theoretical potential, and proves effective even in small two-model ensembles, enabling a cost-efficient way to enhance performance by leveraging multiple LLMs.
Authors: Jing Wang, Jiajun Liang, Jie Liu, Henglin Liu, Gongye Liu, Jun Zheng, Wanyuan Pang, Ao Ma, Zhenyu Xie, Xintao Wang, Meng Wang, Pengfei Wan, Xiaodan Liang
Abstract: Recently, GRPO-based reinforcement learning has shown remarkable progress in optimizing flow-matching models, effectively improving their alignment with task-specific rewards. Within these frameworks, the policy update relies on importance-ratio clipping to constrain overconfident positive and negative gradients. However, in practice, we observe a systematic shift in the importance-ratio distribution-its mean falls below 1 and its variance differs substantially across timesteps. This left-shifted and inconsistent distribution prevents positive-advantage samples from entering the clipped region, causing the mechanism to fail in constraining overconfident positive updates. As a result, the policy model inevitably enters an implicit over-optimization stage-while the proxy reward continues to increase, essential metrics such as image quality and text-prompt alignment deteriorate sharply, ultimately making the learned policy impractical for real-world use. To address this issue, we introduce GRPO-Guard, a simple yet effective enhancement to existing GRPO frameworks. Our method incorporates ratio normalization, which restores a balanced and step-consistent importance ratio, ensuring that PPO clipping properly constrains harmful updates across denoising timesteps. In addition, a gradient reweighting strategy equalizes policy gradients over noise conditions, preventing excessive updates from particular timestep regions. Together, these designs act as a regulated clipping mechanism, stabilizing optimization and substantially mitigating implicit over-optimization without relying on heavy KL regularization. Extensive experiments on multiple diffusion backbones (e.g., SD3.5M, Flux.1-dev) and diverse proxy tasks demonstrate that GRPO-Guard significantly reduces over-optimization while maintaining or even improving generation quality.
Authors: Alessandro Sestini, Joakim Bergdahl, Jean-Philippe Barrette-LaPierre, Florian Fuchs, Brady Chen, Michael Jones, Linus Gissl\'en
Abstract: While several high profile video games have served as testbeds for Deep Reinforcement Learning (DRL), this technique has rarely been employed by the game industry for crafting authentic AI behaviors. Previous research focuses on training super-human agents with large models, which is impractical for game studios with limited resources aiming for human-like agents. This paper proposes a sample-efficient DRL method tailored for training and fine-tuning agents in industrial settings such as the video game industry. Our method improves sample efficiency of value-based DRL by leveraging pre-collected data and increasing network plasticity. We evaluate our method training a goalkeeper agent in EA SPORTS FC 25, one of the best-selling football simulations today. Our agent outperforms the game's built-in AI by 10% in ball saving rate. Ablation studies show that our method trains agents 50% faster compared to standard DRL methods. Finally, qualitative evaluation from domain experts indicates that our approach creates more human-like gameplay compared to hand-crafted agents. As a testament to the impact of the approach, the method has been adopted for use in the most recent release of the series.
Authors: Masahiro Kato
Abstract: We develop a direct debiased machine learning framework comprising Neyman targeted estimation and generalized Riesz regression. Our framework unifies Riesz regression for automatic debiased machine learning, covariate balancing, targeted maximum likelihood estimation (TMLE), and density-ratio estimation. In many problems involving causal effects or structural models, the parameters of interest depend on regression functions. Plugging regression functions estimated by machine learning methods into the identifying equations can yield poor performance because of first-stage bias. To reduce such bias, debiased machine learning employs Neyman orthogonal estimating equations. Debiased machine learning typically requires estimation of the Riesz representer and the regression function. For this problem, we develop a direct debiased machine learning framework with an end-to-end algorithm. We formulate estimation of the nuisance parameters, the regression function and the Riesz representer, as minimizing the discrepancy between Neyman orthogonal scores computed with known and unknown nuisance parameters, which we refer to as Neyman targeted estimation. Neyman targeted estimation includes Riesz representer estimation, and we measure discrepancies using the Bregman divergence. The Bregman divergence encompasses various loss functions as special cases, where the squared loss yields Riesz regression and the Kullback-Leibler divergence yields entropy balancing. We refer to this Riesz representer estimation as generalized Riesz regression. Neyman targeted estimation also yields TMLE as a special case for regression function estimation. Furthermore, for specific pairs of models and Riesz representer estimation methods, we can automatically obtain the covariate balancing property without explicitly solving the covariate balancing objective.
Authors: Jason M. Pittman, Anton Phillips Jr., Yesenia Medina-Santos, Brielle C. Stark
Abstract: In aphasia research, Speech-Language Pathologists (SLPs) devote extensive time to manually coding speech samples using Correct Information Units (CIUs), a measure of how informative an individual sample of speech is. Developing automated systems to recognize aphasic language is limited by data scarcity. For example, only about 600 transcripts are available in AphasiaBank yet billions of tokens are used to train large language models (LLMs). In the broader field of machine learning (ML), researchers increasingly turn to synthetic data when such are sparse. Therefore, this study constructs and validates two methods to generate synthetic transcripts of the AphasiaBank Cat Rescue picture description task. One method leverages a procedural programming approach while the second uses Mistral 7b Instruct and Llama 3.1 8b Instruct LLMs. The methods generate transcripts across four severity levels (Mild, Moderate, Severe, Very Severe) through word dropping, filler insertion, and paraphasia substitution. Overall, we found, compared to human-elicited transcripts, Mistral 7b Instruct best captures key aspects of linguistic degradation observed in aphasia, showing realistic directional changes in NDW, word count, and word length amongst the synthetic generation methods. Based on the results, future work should plan to create a larger dataset, fine-tune models for better aphasic representation, and have SLPs assess the realism and usefulness of the synthetic transcripts.
Authors: Will Wolf
Abstract: Card games are widely used to study sequential decision-making under uncertainty, with real-world analogues in negotiation, finance, and cybersecurity. These games typically fall into three categories based on the flow of control: strictly sequential (players alternate single actions), deterministic response (some actions trigger a fixed outcome), and unbounded reciprocal response (alternating counterplays are permitted). A less-explored but strategically rich structure is the bounded one-sided response, where a player's action briefly transfers control to the opponent, who must satisfy a fixed condition through one or more moves before the turn resolves. We term games featuring this mechanism Bounded One-Sided Response Games (BORGs). We introduce a modified version of Monopoly Deal as a benchmark environment that isolates this dynamic, where a Rent action forces the opponent to choose payment assets. The gold-standard algorithm, Counterfactual Regret Minimization (CFR), converges on effective strategies without novel algorithmic extensions. A lightweight full-stack research platform unifies the environment, a parallelized CFR runtime, and a human-playable web interface. The trained CFR agent and source code are available at https://monopolydeal.ai.
URLs: https://monopolydeal.ai.
Authors: Yahong Yang, Sun Lee, Jeff Calder, Wenrui Hao
Abstract: We derive an energy-based continuum limit for $\varepsilon$-graphs endowed with a general connectivity functional. We prove that the discrete energy and its continuum counterpart differ by at most $O(\varepsilon)$; the prefactor involves only the $W^{1,1}$-norm of the connectivity density as $\varepsilon\to0$, so the error bound remains valid even when that density has strong local fluctuations. As an application, we introduce a neural-network procedure that reconstructs the connectivity density from edge-weight data and then embeds the resulting continuum model into a brain-dynamics framework. In this setting, the usual constant diffusion coefficient is replaced by the spatially varying coefficient produced by the learned density, yielding dynamics that differ significantly from those obtained with conventional constant-diffusion models.
Authors: Runxi Huang, Mingxuan Yu, Mingyu Tsoi, Xiaomin Ouyang
Abstract: Real-time multimodal inference on resource-constrained edge devices is essential for applications such as autonomous driving, human-computer interaction, and mobile health. However, prior work often overlooks the tight coupling between sensing dynamics and model execution, as well as the complex inter-modality dependencies. In this paper, we propose MMEdge, an new on-device multi-modal inference framework based on pipelined sensing and encoding. Instead of waiting for complete sensor inputs, MMEdge decomposes the entire inference process into a sequence of fine-grained sensing and encoding units, allowing computation to proceed incrementally as data arrive. MMEdge also introduces a lightweight but effective temporal aggregation module that captures rich temporal dynamics across different pipelined units to maintain accuracy performance. Such pipelined design also opens up opportunities for fine-grained cross-modal optimization and early decision-making during inference. To further enhance system performance under resource variability and input data complexity, MMEdge incorporates an adaptive multimodal configuration optimizer that dynamically selects optimal sensing and model configurations for each modality under latency constraints, and a cross-modal speculative skipping mechanism that bypasses future units of slower modalities when early predictions reach sufficient confidence. We evaluate MMEdge using two public multimodal datasets and deploy it on a real-world unmanned aerial vehicle (UAV)-based multimodal testbed. The results show that MMEdge significantly reduces end-to-end latency while maintaining high task accuracy across various system and data dynamics.