cs.LG · 2026-06-07 · No. 16

Machine Learning, 2026-06-07.

59 new papers in cs.LG. Titles, authors, abstracts. Links to arXiv. Want this in your inbox every morning? Subscribe →

01 — The papers

59 entries
  1. 01

    TailLoR: Protecting Principal Components in Parameter-Efficient Continual Learning

    Marius Dragoi, Ioana Pintilie, Alexandra Dragomir, Antonio Barbalau, Florin Brad

    cs.LG

    Parameter-efficient finetuning methods based on spectral decomposition have enabled progress in Continual Learning. In this paper we introduce TailLoR, which utilizes the singular bases U and V of the pre-trained weights as a fixed reference frame to learn a low-rank update applied to the singular value matrix. A soft spectral penalty discourages updates aligned with dominant singular directions, reducing interference while routing...

    arxiv.org/abs/2606.06494 · PDF

  2. 02

    Regret Minimization with Adaptive Opponents in Repeated Games

    Mingyang Liu, Asuman Ozdaglar, Tiancheng Yu, Kaiqing Zhang

    cs.LG · cs.AI · cs.GT

    In this paper, we study regret minimization in repeated games with \emph{adaptive} opponents who can respond based on histories of play. The standard metric of \emph{external regret} in online learning is known to fail to capture such adaptivity. To account for players' counterfactual reasoning, we introduce {\tt Repeated Policy Regret (RP-Regret)}, a game-theoretic metric that measures the difference between the \emph{realized} and the...

    arxiv.org/abs/2606.06486 · PDF

  3. 03

    Pretraining Recurrent Networks without Recurrence

    Akarsh Kumar, Phillip Isola

    cs.LG · cs.AI

    Training recurrent neural networks (RNNs) requires assigning credit across long sequences of computations. Standard backpropagation through time (BPTT) addresses this problem poorly: it is sequential in time, limiting parallelism, and suffers from vanishing or exploding gradients, making long-range associations difficult to learn. We propose Supervised Memory Training (SMT), a method for training nonlinear RNNs that sidesteps recurrent credit...

    arxiv.org/abs/2606.06479 · PDF

  4. 04

    RREDCoT: Segment-Level Reward Redistribution for Reasoning Models

    Mykyta Ielanskyi, Kajetan Schweighofer, Lukas Aichberger, Sepp Hochreiter

    cs.LG · cs.AI

    Recent advancements in reasoning language models have been driven by Reinforcement Learning (RL) fine-tuning. Most often, these rely on the Group Relative Policy Optimization (GRPO) algorithm or modifications thereof to steer the models to produce Chain-of-Thought (CoT) traces. The final answer can only be verified, and the reward assigned, after the CoT trace is complete, making it a delayed reward problem. GRPO and its modifications...

    arxiv.org/abs/2606.06475 · PDF

  5. 05

    PC Layer: Polynomial Weight Preconditioning for Improving LLM Pre-Training

    Senmiao Wang, Tiantian Fang, Haoran Zhang, Yushun Zhang, Kunxiang Zhao, Alex Schwing, Ruoyu Sun

    cs.LG · cs.AI

    We propose a preconditioning (PC) layer, a weight parameterization via polynomial preconditioner that ensures stable weight conditioning throughout LLM training. The PC module reshapes the singular-value spectrum of weight matrices via low-degree polynomial preconditioning. After training, the preconditioned weights can be merged back into the original architecture, incurring no inference overhead. We demonstrate the advantage of the proposed...

    arxiv.org/abs/2606.06470 · PDF

  6. 06

    Event Detection for Parameter-to-KPI Dependency Learning for AI-RAN

    Christie Djidjev, Nicholas Kaminski

    cs.LG

    Next-generation wireless networks are expected to rely on multiple concurrent AI-driven control functions that optimize different network objectives simultaneously, particularly in AI-integrated and open radio access network architectures such as AI Radio Access Network (AI-RAN) and Open Radio Access Network (O-RAN). When these functions interact, they can interfere with one another in ways that are difficult to detect from raw network data...

    arxiv.org/abs/2606.06459 · PDF

  7. 07

    In-Context Multiple Instance Learning

    Alexander Möllers, Marvin Sextro, Julius Hense, Gabriel Dernbach, Klaus-Robert Müller

    cs.LG · cs.AI · cs.CV

    Multiple Instance Learning (MIL) addresses problems where supervision is available at the level of bags of instances and has been successfully applied in fields ranging from computational pathology to satellite imagery. Nevertheless, existing algorithms struggle in the low-label regime that characterizes many real-world applications. Flexible models overfit and rigid ones fail to adapt to the task at hand. We show that pretraining an...

    arxiv.org/abs/2606.06458 · PDF

  8. 08

    Causal Atlases from Entropic Inference: Bayesian Networks beyond Optimal DAGs

    Hazhir Aliahmadi, Irina Babayan, Greg van Anders

    cs.LG · stat.ML

    Data-driven causal relationship identification is pertinent to advancing understanding of complex systems both within and beyond science. Bayesian networks offer a probabilistic method for modelling generic causal relationships via directed acyclic graphs (DAGs). However, typical techniques for constructing Bayesian networks rely on optimization, which can be ill-suited for learning causal relationships because the underlying data may admit...

    arxiv.org/abs/2606.06440 · PDF

  9. 09

    Double Preconditioning (DoPr): Optimization for Test-Time Performance, not Validation Loss

    Thomas T. Zhang, Alok Shah, Yifei Zhang, Vincent Zhang, Nikolai Matni, Max Simchowitz

    cs.LG · cs.AI · eess.SY

    Many modern applications of deep learning involve training a neural network via a one-step prediction loss (e.g., $L^2$ regression, cross-entropy), but deploy the network by rolling out along its own predictions. Key examples include autoregressive language modeling, flow-based generative modeling, and robot policy learning. It is well-documented that these settings induce a phenomenon we call test-time feedback (TTF): the mismatch between...

    arxiv.org/abs/2606.06418 · PDF

  10. 10

    The Post-GCN Decade Revisited: Curvature-Stratified Evaluation of Relational Learning

    Shuo Wang, Xiangyu Wang, Quanxin Wang, Bailin Wu, Bokui Wang, Shunyang Huang, Boyan Deng, Haonan Liu, Ruiyi Fang,...

    cs.LG

    Current evaluation practices in relational learning rely heavily on flat leaderboards that average performance across heterogeneous datasets, implicitly assuming a uniform underlying structure. We show that this assumption introduces systematic bias: it obscures geometry-dependent performance variations and can lead to misleading conclusions about model generalization. In this work, we identify intrinsic geometry as a key latent factor...

    arxiv.org/abs/2606.06397 · PDF

  11. 11

    Proper Scoring Rules for Right-Censored Survival Data

    Jef Jonkers, Glenn Van Wallendael, Luc Duchateau, Sofie Van Hoecke

    cs.LG

    Proper scoring rules provide a rigorous theoretical basis for the training and evaluation of probabilistic forecasts. However, in the presence of right censoring, the event time is only partially observed, rendering conventional scoring rules inapplicable in their standard form. We propose a framework for proper scoring of right-censored survival outcomes based on a simple idea: first, map the predictive distribution through the censoring...

    arxiv.org/abs/2606.06393 · PDF

  12. 12

    Learned Response-Field Inertia Operator for HEC-RAS 2D Water-Surface Elevation Prediction

    Edward Holmberg, Elias Ioup, Md Meftahul Ferdaus, Mahdi Abdelguerfi, Julian Simeonov

    cs.LG

    This article presents a cross-dataset evaluation of learned native-cell surrogate models for solver-consistent water-surface elevation (WSE) prediction in HEC-RAS 2D. To avoid raster remapping error and information-access confounding, surrogates are evaluated directly on the original nonuniform computational cells under an explicit policy that separates static project inputs, current hydraulic state, project-input forcing, calibration-derived...

    arxiv.org/abs/2606.06385 · PDF

  13. 13

    End-to-End Subgraph Detection with GraphDETR

    Dexiong Chen, Till Hendrik Schulz, Karsten Borgwardt

    cs.LG · stat.ML

    Subgraph detection seeks to identify whether and where instances of query patterns occur within a larger graph. This problem is fundamental across scientific domains and is closely related to subgraph isomorphism, which is NP-complete, limiting combinatorial approaches to small patterns or moderately sized graphs. We introduce GraphDETR, a deep learning framework that formulates subgraph detection as a set prediction problem, analogous to...

    arxiv.org/abs/2606.06364 · PDF

  14. 14

    Maximising the Set-Piece Return: Optimising Football Corner Tactics with Graph Reinforcement Learning

    Sean Groom, Michael Groom, Francisco Belo, Axl Rice, Liam Anderson, Victor-Alexandru Darvariu, Shuo Wang

    cs.LG

    Machine learning is increasingly employed for the evaluation of football tactics. However, existing approaches focus on characterising historical actions or analyst-specified counterfactual scenarios. In this work, we seek to go beyond the imitation of historically observed patterns towards discovering new generalisable player configurations and strategies. To tackle this, we focus on optimising corner kick routines, and formulate a...

    arxiv.org/abs/2606.06353 · PDF

  15. 15

    Performance Evaluation of GraphCast for Medium-Range Weather Forecasting over Brazil

    Wolfgang R. Rowell, Lucas S. Kupssinskü

    cs.LG

    The paradigm of global weather forecasting is rapidly shifting with the emergence of Machine Learning Weather Prediction models (MLWP). While these data-driven architectures demonstrate remarkable global skill, regional benchmarks in the Global South remain scarce, leaving their efficacy in complex, highly convective environments largely unverified. This study evaluates the performance of GraphCast operational against the deterministic ECMWF...

    arxiv.org/abs/2606.06348 · PDF

  16. 16

    Equivariant Neural Belief Propagation

    Zehua Cheng, Wei Dai, Jiahao Sun

    cs.LG · cs.SC

    Probabilistic inference over spatially embedded variables requires beliefs that respect $SE(3)$ symmetry, yet existing equivariant networks produce only scalars and vectors -- not the rank-2 precision tensors needed for anisotropic uncertainty, and single-component messages collapse multi-modal energy landscapes to physically meaningless averages. We introduce Equivariant Neural Belief Propagation (ENBP), a factor-graph framework whose...

    arxiv.org/abs/2606.06344 · PDF

  17. 17

    Bridging Domain Expertise and Generalization for Performance Estimation

    Shuxuan Li, Zhilin Zhao, Quyu Kong, Wei-Shi Zheng

    cs.LG · cs.AI

    Performance estimation under distribution shift aims to predict how a model behaves on an unlabeled test set whose distribution differs from the training data, a scenario that requires reliable indicators that can faithfully reflect model behavior without ground-truth labels. Existing approaches rely solely on the outputs of the given model whose biases are amplified once the distribution shifts, weakening the correlation with the true...

    arxiv.org/abs/2606.06335 · PDF

  18. 18

    Quantifying the Privacy of Counterfactuals by Leveraging Membership Inference Attacks Against Synthetic Data

    Maryam Babaei, Yingke Wang, Hadrien Lautraite, Heber H. Arcolezi, Ulrich Aivodji, Sebastien Gambs

    cs.LG

    Counterfactuals are typically used in high-stakes decision areas to explain a machine learning model by showing how changes to the user profiles result in the desired outcome. However, explaining the model's decisions through counterfactuals can also be exploited by an adversary to conduct privacy attacks against the model or its training data. Drawing on the analogy that counterfactuals provide realistic substitutes for real training data,...

    arxiv.org/abs/2606.06334 · PDF

  19. 19

    Subspace-Aware Sparse Autoencoders for Effective Mechanistic Interpretability

    Seyed Arshan Dalili, Mehrdad Mahdavi

    cs.LG · cs.AI

    Sparse Autoencoders (SAEs) are widely used for mechanistic interpretability in large language models, yet their formulation assigns each latent feature a single decoder direction, implicitly assuming features to be one-dimensional. We show that this assumption mismatches with the multi-dimensional structure of model features, provably inducing feature splitting through two distinct mechanisms. Geometrically, reconstructing a feature of...

    arxiv.org/abs/2606.06333 · PDF

  20. 20

    Efficient Mean Curvature Computation on High-Dimensional Data Manifolds

    Alexandre L. M. Levada

    cs.LG · cs.CG · cs.CV · stat.ML

    Estimating local mean curvature at each point of a high-dimensional dataset is a key ingredient of geometry-aware machine learning algorithms, such as the Mean Curvature Boundary Points (MCBP) method. The naive implementation of this computation, based on a local shape operator approximated from k-nearest neighbor patches, involves an explicit construction of a matrix $H$ whose trace form yields an $O(m^4)$ cost per point, rendering the...

    arxiv.org/abs/2606.06329 · PDF

  21. 21

    PAMF: Prior-Aware Multimodal Fusion for Incomplete Time Series Data

    Ziwen Kan, Wugeng Zheng, Tianlong Chen, Song Wang

    cs.LG · cs.AI

    In healthcare, multimodal time series tasks often operate on incomplete observations in practice, for example when ECG segments are lost because electrodes detach or an entire respiratory channel is unavailable during overnight monitoring. Such missingness typically appears in two structurally distinct patterns: within-modality missing, where values are absent within an otherwise observed modality, and modality-level missing, where an entire...

    arxiv.org/abs/2606.06328 · PDF

  22. 22

    Learning What to Forget: Improving LLM Unlearning via Learned Token-Level Importance

    Gizem Yüce, Giorgos Nikolaou, Nicolas Flammarion

    cs.LG · cs.AI · cs.CL

    Machine unlearning aims to remove targeted knowledge from a trained model while preserving its general capabilities. For autoregressive language models, not all tokens in a forget sample are equally relevant to forgetting. Existing approaches either ignore this heterogeneity or rely on auxiliary models, heuristics, or external annotations to estimate each token's relevance for forgetting. We instead characterize it through the interaction...

    arxiv.org/abs/2606.06320 · PDF

  23. 23

    Plug-and-Play Guidance for Discrete Diffusion Models via Gradient-Informed Logit Correction

    Hongkun Dou, Zike Chen, Fengji Li, Hongjue Li, Yue Deng

    cs.LG · cs.AI

    Controllable generation with discrete diffusion models is often hindered by high computational overhead or the need for retraining. In this paper, we present \underline{\textbf{G}}radient-\underline{\textbf{I}}nformed \underline{\textbf{L}}ogit \underline{\textbf{C}}orrection (\textbf{GILC}), a plug-and-play framework that efficiently estimates guidance signals by repurposing the pretrained denoising network as a variational proxy. To...

    arxiv.org/abs/2606.06303 · PDF

  24. 24

    Tangram: Unlocking Non-Uniform KV Cache for Efficient Multi-turn LLM Serving

    Hyungmin Kim, Minsoo Kim, Hongseok Kim, Jungwook Choi

    cs.LG · cs.SE

    Multi-turn Large Language Model (LLM) serving is critical for consistent user experiences, yet the linear growth of the Key-Value (KV) cache imposes significant pressure on GPU memory and bandwidth. Non-uniform KV compression effectively preserves more information by considering the individual importance of each KV cache. However, such KV cache heterogeneity introduces various systemic challenges - including memory fragmentation, scheduling...

    arxiv.org/abs/2606.06302 · PDF

  25. 25

    Reactive Flux Matching: Mechanism Discovery and Adaptive Sampling of Rare Events

    Rishal Aggarwal, David Ryan Koes, Nicholas M. Boffi, Eric Vanden-Eijnden

    cs.LG · physics.bio-ph · physics.chem-ph

    Path sampling methods generate ensembles of reactive trajectories connecting metastable states, but extracting mechanistic insight from these data remains nontrivial. We introduce Flux Matching, a framework that learns two complementary objects directly from reactive trajectory data: a current velocity $u(z)$, whose streamlines trace the dominant reaction pathways, and a scalar potential $h(z)$, obtained from a weighted Helmholtz-Hodge...

    arxiv.org/abs/2606.06295 · PDF

  26. 26

    PAC-Bayesian Adversarially Robust Generalization for Message Passing Graph Neural Networks: A Sensitivity Analysis

    Ziling Liang, Xinping Yi, Qingsong Wen, Shi Jin

    cs.LG · stat.ML

    Whilst the vulnerability of graph neural networks (GNNs) to adversarial attacks poses a critical threat to graph representation learning, the understanding of the robust generalization behavior remains a fundamental challenge in the adversarial setting. Recently, PAC-Bayesian margin-based generalization analysis substantially advances this line of research by providing a flexible and data-dependent analytical framework. However, existing...

    arxiv.org/abs/2606.06293 · PDF

  27. 27

    Your GFlowNet Secretly Learns an Optimal Transport Plan

    Ian Maksimov, Nikita Morozov, Denis Belomestny, Sergey Samsonov

    cs.LG · cs.AI

    Generative Flow Networks (GFlowNets) are a framework for sampling structured objects via stochastic trajectories in a directed graph. In this work, we establish a theoretical connection between non-acyclic GFlowNets and optimal transport (OT). We show that fixing the initial flow distribution in a minimum-flow GFlowNet reduces its objective to a Kantorovich OT problem with graph-induced shortest path costs. At the optimum, the learned...

    arxiv.org/abs/2606.06272 · PDF

  28. 28

    Generative Criticality in Large Language Model Temperature Scaling

    Huajian Ruan, Jinyang Li, Xingyu Guo, Lingxiao Wang

    cs.LG · cond-mat.stat-mech · hep-lat

    We propose a statistical-field framework for text generated by large language models (LLMs), treating token embeddings as continuous spin variables on a one-dimensional chain. Defining a susceptibility from the connected two-point correlator and an order parameter from the ensemble-averaged embedding field, we vary the \texttt{softmax} temperature $T$ and observe a sharp susceptibility peak near a characteristic $T_c$ with power-law-like...

    arxiv.org/abs/2606.06238 · PDF

  29. 29

    Tracing the Oracle: Improving Diffusion Timestep Scheduling for 3D CT Reconstruction

    Yujia Wu, Zhaoqiang Liu

    cs.LG

    Pretrained diffusion models demonstrate impressive potential in solving highly ill-posed 3D computed tomography (CT) inverse problems, while the inference process suffers from significant computational overhead. Furthermore, existing uniform timestep schedules fail to capture the non-uniform evolution of the reverse conditional diffusion stochastic differential equation, thereby introducing substantial truncation errors. To overcome this...

    arxiv.org/abs/2606.06236 · PDF

  30. 30

    Design a Reliable LLM-Integrated Interface for Mortality Forecasting

    Thi Kim Ngan Nguyen

    cs.LG · cs.AI

    Mortality forecasting plays an important role in actuarial and policy decision-making, but its implementation remains technically complex and inaccessible to non-expert users. This project proposes a reliable large language model (LLM)-integrated interface that improves usability while maintaining statistical power. The LLM is designed as a constrained orchestration layer that translates natural-language inputs into structured configurations...

    arxiv.org/abs/2606.06235 · PDF

  31. 31

    Non-Negative Matrix Factorization for Event Data

    Raphaël Romero

    cs.LG

    Continuous-time event data, in which entities emit instantaneous events over time, arises naturally across many domains such as neuroscience, seismology, and social networks. Non-negative matrix factorization (NMF) is a natural tool to uncover interpretable structure in such data, but it has so far only been applied after binning or smoothing the entity-level counting measures. This preprocessing step comes with the risk of erasing...

    arxiv.org/abs/2606.06205 · PDF

  32. 32

    A Machine Learning-Based Framework for Discovering Huntington's Disease Stages: Integrating Graph Representation Learning and clustering to Uncover Progression Dynamics in Longitudinal Enroll-HD Dataset

    Lubna M. Abu Zohair, Marta Vallejo, MD Azher Uddin, John R. Woodward, Hind Zantout

    cs.LG

    Huntington's disease (HD) is a progressive brain disorder that gradually affects movement, cognitive function, and behavior. Identifying the stage of the disease accurately and consistently is important for understanding its course, grouping patients, personalized care, and discovering treatment. Existing clinical staging frameworks rely primarily on predefined clinical measurement thresholds and clinical expert decisions, yet these discrete...

    arxiv.org/abs/2606.06196 · PDF

  33. 33

    Learning to Route LLMs from Implicit Cost-Performance Preferences via Meta-Learning

    Jiahao Zeng, Ming Tang, Ningning Ding

    cs.LG · cs.AI · cs.CL

    Large language models (LLMs) present a trade-off between performance and cost, where more powerful models incur greater expense. LLM routing aims to mitigate expenses while maintaining performance by sending queries to the most suitable model. However, existing methods cannot perform well for different user cost-performance preferences. To address this gap, we introduce a novel perceptive LLM routing paradigm for personalized and user-centric...

    arxiv.org/abs/2606.06178 · PDF

  34. 34

    Learning to model pediatric asthma exacerbation from multiple risk factors: a case study in coastal Virginia

    Jonathan Colen, Eric Werner, Maryam Golbazi, Heather Richter, Diana McSpadden, Amy Quinn, Jocel Santos, Mary Jane...

    cs.LG · stat.AP

    Childhood asthma is a common illness exacerbated by air pollution as well as meteorological and neighborhood-level socioeconomic factors. Modeling asthma exacerbation (AE) in large spatiotemporal datasets requires disentangling impacts from multiple contributors. In this case study, we compared three techniques that balance predictive power with interpretability to predict AE in Hampton Roads, a coastal Virginia region comprising 7 cities and...

    arxiv.org/abs/2606.06174 · PDF

  35. 35

    On the training of physics-informed neural operators for solving parametric partial differential equations

    Nanxi Chen, Chuanjie Cui, Airong Chen, Sifan Wang, Rujin Ma

    cs.LG · physics.comp-ph

    Physics-informed neural operators (PINOs) aim to learn solution operators for partial differential equations by using the governing physics as supervision, rather than relying solely on paired input-output simulation data. By incorporating physical constraints into the training objective, PINOs combine the cross-instance generalization of neural operators with the data efficiency of physics-informed learning. Despite this promise, how to...

    arxiv.org/abs/2606.06164 · PDF

  36. 36

    Trust-Aware Predictive Emissions Monitoring for Gas Turbine Fleets with Limited Labelled Data

    Rebecca Potts, Aiden Durrant, Rick Hackney, Georgios Leontidis

    cs.LG

    Machine learning-based predictive emissions monitoring systems offer a practical alternative to direct emissions measurement, but their deployment across gas turbine fleets is challenging when emissions labels are available for only a small subset of assets. In this work, a trust-aware probabilistic framework is proposed for fleet-level gas turbine NOx prediction under limited labelled supervision. The framework combines a multi-head...

    arxiv.org/abs/2606.06156 · PDF

  37. 37

    Tight list replicability bounds via a novel sphere covering theorem

    Ari Blondal, Hamed Hatami, Pooya Hatami, Chavdar Lalov, Sivan Tretiak

    cs.LG

    In recent years, list replicability has emerged as a framework for formalizing reproducibility in learning theory. A central question is how the required list size relates to the accuracy parameter and natural complexity measures of the hypothesis class. To achieve sharp bounds on list replicability, we prove a novel topological sphere covering theorem, derived from the Borsuk-Ulam theorem. Specifically, if the $d$-sphere is covered by open...

    arxiv.org/abs/2606.06148 · PDF

  38. 38

    Adaptive state-action abstractions via rate-distortion

    Fernando E. Rosas

    cs.LG · stat.ML

    When learning to walk, infants seem to address a coarse version of the problem first - stay upright, reach the caregiver - and refine it only when further practice at that resolution stops paying off. Reinforcement learning offers multiple techniques for building simple versions of complex tasks, but lacks general principles for how to dynamically adjust the granularity of these abstractions during learning. This paper proposes one such...

    arxiv.org/abs/2606.06123 · PDF

  39. 39

    A Sliced-Wasserstein Framework on Correlation Matrices for EEG Decoding

    Chen Hu, Rui Wang, Jiale Zhou, Jingjun Yi, Shaocheng Jin, Yidong Song, Yefeng Zheng

    cs.LG

    Electroencephalography (EEG) offers noninvasive, millisecond resolution recordings of neuronal activity and is widely used in neuroscience and healthcare. Many EEG decoding pipelines rely on covariance descriptors for their robustness to noise, but such representations are sensitive to channel-wise scaling. Recent studies have therefore advocated full-rank correlation matrices as a scale-invariant alternative for EEG decoding. In this paper,...

    arxiv.org/abs/2606.06104 · PDF

  40. 40

    OrderGrad: Optimizing Beyond the Mean with Order-Statistic Policy Gradient Estimation

    Paavo Parmas, Yongmin Kim, Kohsei Matsutani, Shota Takashiro, Soichiro Nishimori, Takeshi Kojima, Yusuke Iwasawa,...

    cs.LG · cs.AI · cs.CL

    Policy-gradient methods usually optimize expected return, but many real world applications care about distributional properties of returns: tail risk, outlier robustness, or best-of-K discovery. We introduce OrderGrad, a family of likelihood-ratio and reparameterization gradient estimators for order-statistic objectives. OrderGrad optimizes finite-sample L-statistics, i.e., weighted averages of sorted rewards or costs, recovering objectives...

    arxiv.org/abs/2606.06096 · PDF

  41. 41

    On Advantage Estimates for Max@K Policy Gradients

    Shota Takashiro, Soichiro Nishimori, Paavo Parmas, Yongmin Kim, Kohsei Matsutani, Gouki Minegishi, Yusuke Iwasawa,...

    cs.LG · cs.AI · cs.CL

    Reinforcement learning with verifiable rewards is widely used for post-training reasoning models, but sparse outcome rewards make exploration difficult. A complementary approach is to optimize inference-time objectives such as pass@K and max@K directly, yet existing policy-gradient estimators for these objectives use different signals, baselines, and normalizations, making their relationships unclear. We study this issue through baseline...

    arxiv.org/abs/2606.06080 · PDF

  42. 42

    MDP-GRPO: Stabilized Group Relative Policy Optimization for Multi-Constraint Instruction Following

    Mohammad Mahdi Salmani-Zarchi, Zahra Rahimi, Heshaam Faili, Mohammad Javad Dousti

    cs.LG · cs.AI · cs.CL

    Reinforcement learning with verifiable rewards is ideal for multi-constraint instruction following, yet standard group-relative policy optimization (GRPO) becomes unstable under discrete, low-dispersion rewards, where within-group reward distributions are frequently homogeneous. We identify and formalize three pathologies of z-score group normalization in this regime: low-variance amplification, mean-centering blindness, and zero-variance...

    arxiv.org/abs/2606.06058 · PDF

  43. 43

    Online KL-Regularized Reinforcement Learning with Function Approximation under Misspecification

    Haoyang Hong, Zichen Wang, Quanquan Gu, Huazheng Wang

    cs.LG

    We study KL-regularized contextual bandits and episodic reinforcement learning (RL) under general function approximation with model misspecification. Existing guarantees rely on realizability and therefore do not extend to misspecified models, where classical regret bounds may fail. This work introduces KL misspecification formulations for contextual bandits and episodic RL and analyzes regression-based algorithms with Gibbs policy updates....

    arxiv.org/abs/2606.06053 · PDF

  44. 44

    When Good Enough Is Optimal: Multiplication-Only Matrix Inversion Approximation for Quantized Gated DeltaNet

    Luoming Zhang, Yuwei Ren, Kui Zhang, Tian Liu, Lingjuan Ge, Denghao Li, Matthew Harper Langston, Yin Huang, Weiliang...

    cs.LG · cs.AI

    Matrix inversion in chunk-wise parallel linear attention is a major bottleneck for long-context modeling, particularly on NPUs, where forward-substitution-based methods exhibit limited parallelism and poor hardware utilization. We propose a fast, Matrix Multiplication (MatMul)-based algorithm tailored for strictly lower-triangular matrices arising in chunk-wise linear attention. Motivated by the rapid growth of Neumann-series terms and the...

    arxiv.org/abs/2606.06034 · PDF

  45. 45

    Catastrophic Forgetting as Accessibility Collapse: A Three-Level Framework for Knowledge Persistence in Continual Learning

    Ayushman Trivedi, Bhavika Melwani

    cs.LG

    Catastrophic forgetting is commonly interpreted as the irreversible erasure of previously acquired knowledge during sequential learning. In this work, we investigate an alternative perspective: that forgetting may arise not from complete destruction of task representations but from a loss of accessibility to preserved information. We introduce a three-level framework separating knowledge storage, representation, and accessibility, and...

    arxiv.org/abs/2606.06032 · PDF

  46. 46

    OPRD: On-Policy Representation Distillation

    Shenzhi Yang, Guangcheng Zhu, Bowen Song, Haobo Wang, Mingxuan Xia, Xing Zheng, Yingfan Ma, Zhongqi Chen, Weiqiang...

    cs.LG · cs.AI

    On-policy distillation (OPD) supervises the student only in output space by matching next-token probabilities. This output-only paradigm has two limits: (1) sampling variance from Monte Carlo KL estimates over large vocabularies (e.g., Qwen's ~150k tokens) persists throughout training, and (2) it treats the teacher as a black-box, discarding all intermediate hidden states after the LM head. We propose On-Policy Representation Distillation...

    arxiv.org/abs/2606.06021 · PDF

  47. 47

    Adaptive Oscillatory-State Alignment for Time Series Forecasting

    Zhangyao Song, Ziqiong Li, Xiangfei Qiu, Chao Zha, Yinfei Xu, Tao Guo

    cs.LG · cs.DB

    Long-term time series forecasting benefits from inductive biases that expose recurring temporal structure. Existing periodic forecasting methods typically model recurrence through predefined periods, global spectral components, or fixed learnable templates. However, real-world temporal dynamics are rarely rigidly periodic: oscillatory behavior often evolves through amplitude modulation, phase drift, and local frequency variation. Under these...

    arxiv.org/abs/2606.06010 · PDF

  48. 48

    Diffusion Models for Adaptive Sequential Data Generation

    Haoyang Cao, Minshuo Chen, Yinbin Han, Renyuan Xu

    cs.LG

    Generating realistic synthetic sequential data is critical in real-world applications across operations research, finance, healthcare, energy systems, and scientific computing, where time-indexed observations are used for prediction, simulation, risk assessment, and data-driven decision-making. While diffusion models have achieved remarkable success in generating static data, their direct extensions to sequential settings often fail to...

    arxiv.org/abs/2606.06007 · PDF

  49. 49

    HoT-SSM:Higher-order Temporal Knowledge Graph Reasoning with State Space Models for Health Care

    Thummaluru Siddartha Reddy, Vempalli Naga Sai Saketh, Yash Punjabi, Mahesh Chandran

    cs.LG · eess.SP

    Medical knowledge graphs (MKGs) infused with clinical knowledge have been increasingly used to model electronic health records (EHRs) to support interpretable predictions in healthcare domain. However, existing MKG-based approaches are limited in capturing pairwise relations between clinical concepts (e.g., conditions, procedures, and medications), and restricts their ability to model higher-order interactions among co-occurring or...

    arxiv.org/abs/2606.05994 · PDF

  50. 50

    Compress-Distill: Reasoning Trace Compression for Efficient Knowledge Distillation

    Maxime Griot, Paul Steven Scotti, Tanishq Mathew Abraham

    cs.LG · cs.CL

    Reasoning models produce long chain-of-thought traces that are costly to distill and encourage verbose student outputs. We study post-hoc compression of such traces before knowledge distillation. Two teachers, Qwen3.5-397B-A17B and gpt-oss-120B, generate about 283k correct traces each; two instruction-tuned models then compress them to 8.6-21.0% of their original character length. Across a 48-run main grid plus seven Qwen-teacher truncation...

    arxiv.org/abs/2606.05988 · PDF

  51. 51

    LLM Explainability with Counterfactual Chains and Causal Graphs

    Nirit Nussbaum-Hoffer, Nitay Calderon, Liat Ein-Dor, Roi Reichart

    cs.LG

    Causal graphs provide a high-level language for making mechanisms transparent. Recent work uses Large Language Models (LLMs) to recover causal graphs of external-world processes. Instead, in this paper, we use causal graphs to model LLM inference itself, providing stakeholders with a transparent view of how the model perceives and organizes high-level concepts to produce a prediction. We propose a four-phase method for constructing such...

    arxiv.org/abs/2606.05972 · PDF

  52. 52

    Steering Vectors are an Adversarial Attack Surface

    Abzal Aidakhmetov, Donato Crisostomi, Tommaso Mencattini, Adrian Robert Minut, Iacopo Masi, Emanuele Rodolà

    cs.LG

    Activation steering has become a popular way to control Large Language Model (LLM) behavior without fine-tuning. Since the technique is plug-and-play, users share datasets and precomputed vectors to steer model activations. However, we show that a \emph{stealth data poisoning attack} silently compromises this pipeline. By substituting $4{-}6\%$ of tokens in the steering dataset, an attacker can silently align the resulting vector with an...

    arxiv.org/abs/2606.05958 · PDF

  53. 53

    Dead Directions: Geometric Singular Learning

    Tejas Pradeep Shirodkar

    cs.LG · stat.ML

    Singular learning theory and information geometry have studied the same parameter spaces in mostly separate vocabularies: the former computes Bayesian invariants in resolved coordinates, the latter works in original coordinates under a non-degeneracy assumption that overparameterised models routinely violate. We bridge them through one primitive, the dead direction: a unit vector along which the Fisher metric degenerates, equivalently a...

    arxiv.org/abs/2606.05957 · PDF

  54. 54

    Short paper: Models in the dark -- Rectification and erasure under GDPR in ML supply chains

    Henrik Graßhoff, Malte Hansen, Meiko Jensen, Sara Ramezanian

    cs.LG

    The rights to rectification and erasure, as established under the General Data Protection Regulation (GDPR), are central to protecting individuals' privacy. However, their effective enforcement in machine learning (ML) systems remains challenging. Existing work has largely addressed these rights from either a legal or a technical perspective in isolation and disregards the fact that models are produced in complex supply chains involving...

    arxiv.org/abs/2606.05946 · PDF

  55. 55

    Addressing Imbalance in Multi-Label Data via Label-Specific Distance-based Oversampling

    Bin Liu, Jun Wu, Haoyu Peng, Ao Zhou, Jin Wang, QiaoSong Chen, Grigorios Tsoumakas

    cs.LG

    The complex imbalanced label distribution poses a crucial challenge to multi-label classification, as most classifiers are biased towards the majority class and high-frequent labels. Oversampling is an efficient and flexible solution that augments instances to provide a more balanced training dataset for multi-label classifiers. Most existing oversampling methods create synthetic instances in a heuristic way that essentially relies on...

    arxiv.org/abs/2606.05927 · PDF

  56. 56

    High-Dimensional Theory of LoRA Fine-Tuning in a Solvable Attention Model

    O. Duranthon, F. Boncoraglio, L. Zdeborová

    cs.LG · cond-mat.dis-nn

    We develop a high-dimensional statistical theory of low-rank adaptation (LoRA) in attention models, capturing the interplay between pre-training and fine-tuning. We introduce a solvable framework in which a single-head attention layer is first pre-trained on a data-abundant task and subsequently adapted via a rank-one LoRA update on limited data. In the high-dimensional limit, both stages admit a sharp asymptotic characterization in terms of...

    arxiv.org/abs/2606.05899 · PDF

  57. 57

    When Denser Credit Is Not Enough: Evidence-Calibrated Policy Optimization for Long-Horizon LLM Agent Training

    Yuanfan Li, Qi Zhou, Wenjing Duan, Lu Chen

    cs.LG

    Long-horizon LLM agents require reinforcement learning methods that can assign credit to intermediate decisions under sparse and delayed rewards. Recent group-based methods such as GiGPO improve over GRPO by constructing step-level advantages at repeated anchor states. However, we show that such dense credit can be statistically unreliable: under limited rollouts, rare but lucky actions may receive overly large advantages, producing divergent...

    arxiv.org/abs/2606.05885 · PDF

  58. 58

    TS-ICL: A Flexible Time-Indexed Foundation Model for Time Series via In-Context Learning

    Etienne Le Naour, Tahar Nabil, Adrien Petralia

    cs.LG

    Foundation models mark a profound paradigm shift in time series modeling, with task-specific models being superseded by general-purpose zero-shot models. Yet, current approaches primarily focus on forecasting, while real-world time series are often irregularly and partially observed, requiring models that can jointly forecast, impute missing values, and handle degraded sampling conditions. To address these challenges, we introduce TS-ICL, a...

    arxiv.org/abs/2606.05878 · PDF

  59. 59

    Deciphering Two Training Clocks in Grokking via Deep Linear Network Theory with Conditional ReLU Reduction

    Hu Tan, Kuo Gai, Shihua Zhang

    cs.LG · cs.AI

    Grokking suggests that fitting the training data and learning a simple underlying rule may occur on different time scales. We formalize this phenomenon by separating the fast decay of the classification loss from the slower simplification of the learned representation, and we call the resulting pair of stopping times two training clocks. For deep linear networks, we show that a post-margin gap-growth or one-step tail-contraction condition...

    arxiv.org/abs/2606.05863 · PDF

This edition is part of The Daily Abstract — cs.LG archive. Subscribe to receive these in your inbox each morning, automatically translated to Spanish, with reply-to-PDF: arxivdaily.ignorelist.com.

Colophon Set in Georgia, with system sans for interface chrome and a monospaced stack for code and paper identifiers. Sole accent: amber #D99C5E. Built and served on an always-free VM. The masthead is set 14% letterspaced because newspapers do that and it works.