cs.LG · 2026-10-28 · No. 156

Machine Learning, 2026-10-28.

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

01 — The papers

74 entries
  1. 01

    When Should We Intervene at Test Time?

    Jiahao Zhang, Xinling Wen, Liangliang Li, Yu Chen

    cs.LG

    Test-time adaptation helps a deployed model respond to changing inputs, but the same intervention can rescue one batch and damage the next. Updating normalization statistics, for example, can improve corrupted images yet fail when a batch contains mostly one class. Average accuracy hides these failures and cannot tell a model when to use an intervention. Target labels can guide this decision, but they are scarce and some choices are already...

    arxiv.org/abs/2610.37658 · PDF

  2. 02

    Heads of Security: Learned Attention Gates Improve Prompt Injection Robustness

    Igor Shilov, Xiaoxue Yang, Alexander Panfilov, Evgenii Kortukov, Matthieu Meeus, Peter Romov, Wojciech Samek, Jonas...

    cs.LG · cs.CL · cs.CR

    We introduce Heads of Security (HoS), a prompt injection defense that enforces instruction-data separation via a small set of learned per-head attention gates. HoS learns how each head should redistribute attention between trusted and untrusted tokens, enabling models to reason over untrusted content without following injected instructions. We evaluate HoS on Llama-3.3-70B-Instruct and GPT-OSS-20B. We compare it against SecAlign++, a...

    arxiv.org/abs/2610.37657 · PDF

  3. 03

    Attention as Lattice Gauge Theory: Relations Interact Only Through Wilson Loops

    Aakash Sarkar, Ananya Mukherjee, Motoshige Sato

    cs.LG · cs.AI · physics.comp-ph

    Relations compose through the entity they share - Alice is Bob's mother and Bob is Carol's father, so Alice is Carol's grandmother - and the order matters, since a father's mother is not a mother's father; verifying a composed relation means closing the loop. We show that if each entity's coordinate frame may be rotated independently (per-entity frame freedom), three things follow. In the pure-gauge sector, relations that share no cycle...

    arxiv.org/abs/2610.37656 · PDF

  4. 04

    Scaling Laws for Tool-Use Agents via Controllable Environment Synthesis

    Andrew Szot, Samira Abnar, Omar Attia, Abbas Kazerouni, Michael Kirchhof, Cheng-Yu Hsieh, Raviteja Vemulapalli,...

    cs.LG

    We explore how to build performant large language model (LLM) tool-use agents that complete complex, multi-step tasks. Specifically, we study how data composition in synthetically generated tool-use environments affects the training of such agents. To do so, we propose the SynthAPI pipeline, which explicitly controls the diversity of generated environments along the axes of tools, tasks, and the databases on which the tools operate. By...

    arxiv.org/abs/2610.37648 · PDF

  5. 05

    MuonGS: Pretraining LLMs with Relaxed Weight-Norm Control

    Athanasios Glentis, Dawei Li, Mingyi Hong

    cs.LG · cs.AI · math.OC

    Weight decay and explicit weight-norm constraints improve optimizer performance in language model pretraining, yet a unified understanding of why such control helps and how much is needed remains incomplete. We analyze norm control in Muon over the entire training trajectory using the \emph{accumulated effective learning rate}, $C_T=\sum_{t<T}η_t/\|W_t\|_F$, where $η_t$ is the nominal learning rate and $η_t/\|W_t\|_F$ is the effective...

    arxiv.org/abs/2610.37647 · PDF

  6. 06

    EnvSched: Environment-Aware Scheduling for Rollout-Efficient Agent Reinforcement Learning

    Sihan Yang, Weichen Bai, Ruichuan An, Yiming Cheng, Zhiyuan Feng, Tianyi Jiang, Renye Yan, Ning Ding, Yifan Zhang,...

    cs.LG · cs.CL

    As executable environments for language agents scale, generating and verifying their rollouts becomes increasingly expensive. Under finite rollout budgets, this raises an important question: can data management at the environment level make agentic reinforcement learning (RL) more efficient by organizing, evaluating, and scheduling task instances together with the environments that determine their tools, states, transitions, and verification?...

    arxiv.org/abs/2610.37609 · PDF

  7. 07

    GAS-PFN: Graph-Aware Spatiotemporal Prior-Data Fitted Networks for Zero-Shot Traffic Forecasting

    Joseph Natter, Long Vu, Yifan Zhang, Bobin Deng, Kun Suo, Rui Wu

    cs.LG · cs.AI · cs.NE

    Traffic forecasting requires modeling complex temporal dynamics and the spatial dependencies imposed by road topology. Supervised graph-based models achieve strong performance, but are often trained for a particular dataset and sensor graph, making deployment to new or evolving networks costly. Prior-Data Fitted Networks (PFNs) offer a zero-shot alternative by learning in-context inference from synthetic task priors, which provide an explicit...

    arxiv.org/abs/2610.37598 · PDF

  8. 08

    Understanding the limits of latent world models

    Rodrigo Gonzalez Laiz, Steffen Schneider

    cs.LG · cs.AI

    World models are widely proposed as core components for enabling agents to reason over long temporal and spatial horizons. Yet current evaluations often emphasize task performance while providing limited insight into whether learned world models acquire predictive latent dynamics that reflect the underlying structure of the environment. Here we introduce the concept of faithful latent world models as an equivariance condition between the...

    arxiv.org/abs/2610.37588 · PDF

  9. 09

    LCM-Promo: An LLM-Based Uplift Model for Promotion Allocation

    Nesta Midavaine, Floris Fok, Paulo Mendes, Joao Silva, Önder Akaçık, Dimas Lima, Zülküf Genç

    cs.LG · cs.IR

    We present LCM-Promo, an LLM-based uplift model for promotion allocation. LCM-Promo is a task-specific Large Commerce Model (LCM): a finetuned LLM trained on transactional commerce data to predict user intent and promotion response. It learns monotonic promotion-response curves from production logs by interpreting single-token outputs as conversion-threshold probabilities, while conditioning on session context such as day and delivery shift....

    arxiv.org/abs/2610.37570 · PDF

  10. 10

    Gym-Anything-RT: Scaling Real-Time Environments for Vision-Language Agents

    Pranjal Aggarwal, Sean Welleck

    cs.LG · cs.AI

    AI agents have made substantial progress in domains where they have a non-trivial amount of time to deliberate between actions, such as mathematics, software engineering, and performing tasks on desktop applications. It is less clear how such models fare in settings that demand real-time interaction, such as playing a video game or operating a physical device. In these settings, the environment state may evolve as the agent is reasoning, or...

    arxiv.org/abs/2610.37567 · PDF

  11. 11

    Pneuma: A Large-Scale Real-World Time Series Benchmark for Commercial Aircraft Predictive Maintenance

    Jiafeng Lin, Yutian Zhang, Huan Zhang, Yuxuan Wang, Jinyan Feng, Yuhui Liu, Zhongyi Pei, Jianmin Wang

    cs.LG

    Predictive maintenance is vital for commercial aviation, yet progress remains hindered by the lack of public, large-scale real-world datasets. As a result, most methods are developed and evaluated on proprietary data, limiting reproducibility. We introduce Pneuma, the first publicly available large-scale benchmark for predictive maintenance using onboard time series from in-service commercial aircraft. Built around bleed air systems, Pneuma...

    arxiv.org/abs/2610.37563 · PDF

  12. 12

    When uncertainty quantification fails: structural vulnerabilities and corrections for selective prediction under clinical class imbalance

    Walter Endrizzi, Flavio Ragni, Raffaele Marchesi, Stefano Bovo, Monica Moroni, Giuseppe Jurman, Venet Osmani

    cs.LG · cs.AI

    Background. Selective prediction systems improve the safety of clinical artificial intelligence by deferring ambiguous cases to human experts while automating high-confidence decisions. However, this safety mechanism relies entirely on the premise that uncertainty scores accurately reflect clinical ambiguity. In prognostic tasks characterized by severe class imbalance, where the clinical outcome of interest is a minority, this assumption...

    arxiv.org/abs/2610.37561 · PDF

  13. 13

    Collective Dynamics of Self-Reinforcement: How Cross-Task Coupling Drives Policy Collapse in Unsupervised RLVR

    Peize Ding

    cs.LG

    Label-free reinforcement learning with verifiable rewards (RLVR) rewards a language model for agreeing with its own majority vote. It improves reasoning at first but eventually collapses: the self-assessed reward goes to one while accuracy falls, often until the model gives the same answer to every problem. The collapse appears only when training uses many problems. We explain this dataset-size dependence as a collective instability. In a...

    arxiv.org/abs/2610.37553 · PDF

  14. 14

    GTFormer: A Syntax-Free Graph Network Alternative to Transformers for 8192$\times$ Length Extrapolation in Linear Time

    Elisabeth Fink

    cs.LG

    Self-attention mechanisms have established the Transformer as the dominant backbone across natural language processing, source code intelligence, and automated theorem proving. However, all-to-all pairwise token comparison enforces a quadratic $\mathcal{O}(N^2)$ compute and memory barrier alongside severe softmax attention dilution as sequence length $N$ scales beyond the training horizon. Inspired by recent advances in algebraic graph...

    arxiv.org/abs/2610.37524 · PDF

  15. 15

    The Structure Gap: When More Data Cannot Replace Structured Reasoning

    Kheer Sagar Patel, Mojtaba Soltanalian

    cs.LG · cs.AI

    A persistent belief in modern AI is that sufficiently large learners trained on sufficiently much data can make explicit domain knowledge optional. This paper examines a limitation of that belief. We define structure-encoded problems: tasks whose input-output map is determined, or sharply constrained, by known structure such as a solver, symmetry group, graph, physical law, measurement operator, grammar, combinatorial search space, or...

    arxiv.org/abs/2610.37523 · PDF

  16. 16

    Orthogonal and Parametrized Multi-Attribute Steering of LLMs

    Remy Ogasawara, Cullen Anderson, Narmeen Fatimah Oozeer, Foad Namjoo, Amir Abdullah, Jens Kristian Refsgaard Schou,...

    cs.LG · cs.AI

    Activation steering offers an effective and efficient way to modify an LLM's behavior at inference time. However, it remains unclear how to apply steering such that the signal is consistently strong and the steering strength has a universally interpretable meaning. We address this first by identifying the most steerable interval for each steering vector and rescaling the steering range accordingly. Building on this, we introduce a more...

    arxiv.org/abs/2610.37517 · PDF

  17. 17

    Follow the Gradient: Tracing Safety Degradation Back to Training Data

    Saket Reddy, Amro Abbas, Anshuman Suri, Ari Morcos, David Schwab, Pratyush Maini

    cs.LG · cs.AI · cs.CL

    Fine-tuning Large Language Models (LLMs) on instruction data can quietly erode their safety, even when no training example looks harmful. Because this safety-degrading data looks benign, standard content filters struggle to catch it. We tackle this problem by developing SimpleLoRIF, a scalable gradient embedding method that identifies fine-tuning examples whose gradients align with unsafe behavior. By simplifying a low-rank attribution method...

    arxiv.org/abs/2610.37481 · PDF

  18. 18

    An Input Skip Can Reparameterize a VAE Without Improving the Generator

    Kevin Lee

    cs.LG · physics.ao-ph · stat.ML

    An input skip can reduce reconstruction error while leaving a model's distribution of new samples unchanged. For a Gaussian variational autoencoder (VAE) with constant input skip $0\leqα<1$, a model without a skip can be written in skip form by multiplying its decoder output by $b=1-α$ and its noise variance by $b^2$. The distribution of observations and latent variables and its exact generative scores stay the same, while reconstruction mean...

    arxiv.org/abs/2610.37456 · PDF

  19. 19

    Interpretable Meta-Policies for Neural Decision Making

    Raban Emunds, Jannis Blüml, Quentin Delfosse, Kristian Kersting

    cs.LG

    Deep Reinforcement Learning remains difficult to interpret and align due to the black-box nature of neural networks. While (neuro-)symbolic approaches improve transparency by incorporating symbolic reasoning, applying them directly to low-level actions results in overly complex, unreadable policies. We introduce NEXUS, a hierarchical RL framework that resolves this tension by combining neural skills with symbolic meta-policies. By shifting...

    arxiv.org/abs/2610.37451 · PDF

  20. 20

    CRISP: Co-designing Pruning and Scheduling for Efficient MoE Inference on Edge Servers

    Congwei Zhang, Youming Tao, Yifei Zou, Yu Yang, Ruirui Zhang, Xiaoyu Zhang, Falko Dressler, Xiuzhen Cheng, Dongxiao Yu

    cs.LG

    Mixture-of-Experts (MoE) inference on memory-constrained edge platforms is often dominated by expert-weight movement rather than computation. Although each token activates only a small subset of experts, routing is input-dependent, so all experts must remain accessible and memory movement dominates latency. Existing approaches such as uniform pruning and expert dropping overlook two properties of MoE execution: expert combinations are highly...

    arxiv.org/abs/2610.37442 · PDF

  21. 21

    When Visual Evidence Is Insufficient: How Instructions, Post-Training, and Internal Activations Shape Refusal in Multimodal Language Models

    Ananth K Suresh, Aveen Dayal, Gouthaman KV, Aditya Pola, Vineeth N Balasubramanian

    cs.LG · cs.CV

    For reliable deployment, multimodal large language models (MLLMs) should recognize when an image provides insufficient evidence to answer a question and withhold unsupported answers. Yet refusal alone is not necessarily evidence of such recognition: models may refuse because of unrelated safety constraints or without identifying the missing evidence. We study how refusal on visually unanswerable questions is shaped by three sources of...

    arxiv.org/abs/2610.37430 · PDF

  22. 22

    Bayes-Adaptive MDPs for Rollout Allocation to Enhance GRPO Training

    Faeze Moradi Kalarde, Armin Toroghi, Nam Ho Koh, Doo Hwan Hwang, Seungjin Kim, Kee-Eung Kim, Scott Sanner

    cs.LG · cs.AI

    Group Relative Policy Optimization (GRPO) has demonstrated strong performance in LLM post-training. However, it allocates a uniform rollout budget to every training question, ignoring variations in question difficulty. As a result, highly difficult questions may receive too few rollouts to generate informative positive reward signals, limiting their contribution to training, while moderately difficult questions often receive more rollouts...

    arxiv.org/abs/2610.37429 · PDF

  23. 23

    Displacement Moments and Path-Terminal Separation in Brownian Interpolant Sampling

    Zhong Han

    cs.LG · math.PR

    Brownian stochastic interpolants connect paired endpoint samples, yet their sampling error can depend on displacement moments even when neither endpoint has a first moment. For arbitrary Borel couplings, we prove early-stopped, target-to-output KL guarantees for standard Euler sampling from the true source. A fourth displacement moment gives an explicit $O(N^{-1})$ bound; a second moment suffices for fixed-law convergence as drift error...

    arxiv.org/abs/2610.37413 · PDF

  24. 24

    LinMix: Data Mixing as Linear Domain Attribution

    Hubert Baniecki, Przemyslaw Biecek, Martin Pawelczyk

    cs.LG · cs.CL · stat.ML

    Language model performance depends strongly on the mixture of domains in the pretraining data corpus. Popular data mixing methods select these domain weights by training many small proxy models on random mixes and then fitting a regression model to predict performance from the mix. We recognize this practice as a domain attribution problem, where the regression measures each domain's contribution to the target metric. Through this lens, the...

    arxiv.org/abs/2610.37390 · PDF

  25. 25

    How Many and How Much? Contrastive Learning with Negative Samples for Knowledge Graph Embedding

    Sooho Moon, Yunyong Ko

    cs.LG

    Knowledge graph embedding (KGE), a fundamental technique for knowledge-centric Web and KG-augmented AI applications, is commonly trained through contrastive learning with negative samples to learn discriminative representations. Despite its empirical success, two crucial questions regarding negative samples remain underexplored: (Q1) how many negative samples should be used for each positive sample and (Q2) how much should each negative...

    arxiv.org/abs/2610.37389 · PDF

  26. 26

    Para-Gaussian Approximations of Random Deep Neural Networks

    Mihai Nica, Janosch Ortmann

    cs.LG · math.PR

    It is known that randomly initialized neural networks have Gaussian behaviour in the $n\to \infty$ limit. Unfortunately, this approximation can be poor for finite $n$ and gets worse as one goes deeper into the network. In this work, we develop a more accurate approximation that is \emph{slightly} non-Gaussian, which we call the para-Gaussian distribution and is constructed from the Hermite polynomials. We show how this para-Gaussian...

    arxiv.org/abs/2610.37388 · PDF

  27. 27

    Channel Selection for Mixture-of-Experts Decoding Needs No Calibration Data

    Geunhyoung Kim, Sung Ju Hwang

    cs.LG · cs.AI

    Mixture-of-Experts (MoE) models are bottlenecked by loading routed-expert weights from memory during interactive decoding at small batch sizes. Because most intermediate channels in a routed expert contribute little to any single token, channel selectors predict which channels matter and read only their weights, typically keeping channels above a threshold calibrated on offline data. We show that channel selection needs no calibration data....

    arxiv.org/abs/2610.37386 · PDF

  28. 28

    Euclidean Crowding-Driven Adaptive Euclidean-Hyperbolic Graph Learning

    Hao Chen, Jian Sun, Chuan Shi, Hongying Zhang

    cs.LG

    Hyperbolic graph representation learning has emerged as a powerful paradigm for modeling hierarchical data. Unlike Euclidean space, which exhibits polynomial volume growth, hyperbolic space grows exponentially with radius, naturally accommodating tree-like structures without the geometric distortion known as Euclidean crowding. Despite strong empirical performance, the theoretical mechanisms underlying hyperbolic embeddings remain poorly...

    arxiv.org/abs/2610.37373 · PDF

  29. 29

    TRIDENT-Seq: Profile-Matched Prefix-Future Contrast for Label-Efficient Payment Risk Modeling

    Weiyu Han, Haoyu Jiang, Jie Sheng, Ruobing Zuo, Huilin He, Ze Zhao, Xiao Tan, Nairong Zhang, Shuai Chen, Tianyi Zhang

    cs.LG

    Ant International accumulates large volumes of transaction history across users and payment businesses. Confirmed fraud and chargeback outcomes, by contrast, are scarce, delayed, and specific to each business. A single encoder pretrained on unlabeled histories could serve all businesses, but only if pretraining captures the behavior that separates risk. Coarse activity differences can make contrastive pretraining on transaction logs too easy:...

    arxiv.org/abs/2610.37365 · PDF

  30. 30

    EmbeddingGemma 2: Unified and Efficient Multimodal Representations

    Henrique Schechter Vera, Sahil Dua, Jovan Andonov, Sara Smoot, Sebastian Russo, Ivan Llanos, Max Matveev, Jae Hyeon...

    cs.LG · cs.CL · cs.CV · cs.SD

    We present EmbeddingGemma 2, an open multimodal embedding model which natively represents text, images, audio, video, and combinations thereof in a unified space. Our approach introduces improvements such as simplified per-layer embeddings and a multi-stage curriculum with modality and objective progression which ensure strong quality across unimodal, cross-modal, and mixed-modal evaluations. Notably, in comprehensive embedding benchmarks...

    arxiv.org/abs/2610.37364 · PDF

  31. 31

    Normative Deviation Modeling for EEG-based Depression Detection

    Wan Jiang, Jingqi Cheng, Hui Zhang, Xu Zhao, Xiaojing Chen, Shengeng Tang, Jingjing Wu, Richang Hong

    cs.LG

    Web-enabled digital health services are expanding access to physiological monitoring and mental health screening. However, electroencephalography (EEG)-based major depressive disorder (MDD) detection remains challenging across heterogeneous data sources, where individual variability and acquisition differences may affect disease-related representations. In this work, we investigate the use of healthy normative references for EEG-based MDD...

    arxiv.org/abs/2610.37338 · PDF

  32. 32

    CoS-VQ: Coordinated Shared-Scale Vector Quantization for Large Language Models

    Zukang Xu, Zhixiong Zhao, Xing Hu, Zhe Jiang, Xin Si, Dawei Yang

    cs.LG · cs.AI

    Large language models (LLMs) require substantial memory, motivating low-bit weight quantization for efficient deployment. Existing vector quantization (VQ) methods often improve accuracy by increasing the effective coding dimension, which can complicate independent lookup and local codeword updates. We propose CoS-VQ (Coordinated Shared-Scale Vector Quantization), which preserves low-dimensional codewords while jointly optimizing hard indices...

    arxiv.org/abs/2610.37336 · PDF

  33. 33

    WCC: A Learning-Based Framework for Deciding When to Apply Cooperative Coevolution in Large-Scale Optimization

    Wenjie Qiu, Hepei Li, Jianfeng Wang, Jun Zhang, Yue-Jiao Gong

    cs.LG · cs.AI

    Cooperative Coevolution (CC) and Non-Decomposition Algorithms (NDAs) are two primary approaches for large-scale global optimization (LSGO). Although both have achieved success, they exhibit markedly different performance across problem types and each faces approach-specific limitations, leaving the question of "when to use CC" unresolved. We address this gap with WCC, a novel learning-based framework that dynamically decides when to deploy CC...

    arxiv.org/abs/2610.37329 · PDF

  34. 34

    Structured Representation Probing for Risk Anticipation in End-to-End Autonomous Driving

    Jiawei Liang, Tianrui Lou, Le Yang, Xianghao Jiao, Wei Wang, Xiaochun Cao

    cs.LG

    End-to-end driving policies form structured internal representations of traffic, ego behavior, and road constraints. These representations offer a monitoring interface for anticipating failures, complementing risk signals from trajectories, controls, and manually designed features. We introduce the Structured Representation Probe (SRP), a lightweight monitor that reads actor, ego, and map representations from a frozen policy. SRP learns a...

    arxiv.org/abs/2610.37299 · PDF

  35. 35

    Bayesian Inference of Expensive Simulators: Survey and Comparison of Design Choices

    Šimon Soldát, Václav Šmídl

    cs.LG

    We present a survey of approaches to Bayesian inference with simulator forward models, focusing on settings with prohibitively expensive simulators. In this domain, Bayesian optimization-based methods have emerged as the most promising approach. We unify existing methods into a general framework, enabling consistent reasoning about shared algorithmic structure across the fragmented literature. Within this framework, we identify the primary...

    arxiv.org/abs/2610.37284 · PDF

  36. 36

    Where Should the Evaluation Budget Go in LLM-Driven Agent Scaffold Evolution?

    Sichun Luo, Yi Huang, Mukai Li, Haochen Luo, Guanzhi Deng, Haibo Wang, Jian Xu, Zefa Hu, Junlan Feng, Qi Liu

    cs.LG · cs.AI · cs.CL · cs.NE

    Large language model (LLM)-driven agent scaffold evolution improves agent workflows through iterative candidate generation and evaluation. Evaluation both guides subsequent search and determines the final scaffold, but evaluating every candidate on the full task set incurs substantial execution costs. We investigate how evaluation budgets can be allocated across these two roles. Our experiments show that sparse feedback can sustain productive...

    arxiv.org/abs/2610.37276 · PDF

  37. 37

    Your Policy Is Secretly a Harness: The Decision Geometry of Agent Interfaces

    Jian Mu, Ruibo Shen, Yao Shu

    cs.LG · cs.CL

    Language model agents are post-trained and deployed through a harness, the prompts, tools, and control logic that present a task to the model and execute its outputs. Harnesses are often revised and judged by how an already trained model performs under them, which assumes that what the model learned under one harness carries over to the next. We ask what a policy learns from its harness and what survives when the harness changes. Under...

    arxiv.org/abs/2610.37255 · PDF

  38. 38

    When Is a Relation Worth Trusting? Manipulation Costs, Exposure, and the Trust Trap in Multi-Relational Graph Learning

    Vincess Dongmo, Astride Melvin Fokam Ninyim

    cs.LG · cs.AI · cs.CR

    Multi-relational graph detectors learn from clean data how much to trust each relation, but an adversary pays a different price to manipulate each relation and each action; taking such costs as given, we ask when using a relation is worse than ignoring it. Our first main result proves that degree gating strictly beats every selection or reweighting of a relation when part of the population is isolated on it and the attacker cannot afford to...

    arxiv.org/abs/2610.37239 · PDF

  39. 39

    Sharpness-Aware Muon: Matrix-Aware Optimization with Momentum-Guided Perturbations

    Zhiwei Zhai, Wenjing Yan, Ying-Jun Angela Zhang

    cs.LG

    Muon is a matrix-aware optimizer that combines momentum with orthogonalization to produce structured update directions with strong empirical behavior. However, it only shapes the geometry of the update itself and does not explicitly control neighborhood worst-case loss or bias training toward flat regions, which are often linked to stronger generalization. To address this limitation, we introduce the sharpness-aware formulation of Muon. We...

    arxiv.org/abs/2610.37225 · PDF

  40. 40

    SimplexLoRA: Dynamic Rank Updating via Alternating Minimization with Simplex Projection

    Igor Shalygin, Grigorii Davydenko, Andrey Veprikov, Aleksey Antonov, Nikolay Bushkov, Stanislav Moiseev, Aleksandr Beznosikov

    cs.LG

    The success of large language models in real-world applications relies on parameter-efficient fine-tuning methods such as Low-Rank Adaptation (LoRA), which inserts compact, trainable adapters into selected layers. In this paper, we propose SimplexLoRA, a LoRA variant that adaptively adjusts adapter ranks during fine-tuning based on their learned importance. Unlike standard LoRA, which employs a fixed rank across all layers, SimplexLoRA...

    arxiv.org/abs/2610.37206 · PDF

  41. 41

    Action-Value Calibration via Continuation Sharing for Long-Horizon Agentic Reinforcement Learning

    Yingjian Chen, Lang Zhou, Zixin Zhao, Pingfan Li, Weishi Zheng, Zhilin Zhao

    cs.LG · cs.AI

    Reinforcement learning for long-horizon agents relies on feedback from sequences of interactions, making credit assignment to individual actions difficult. Existing methods such as GiGPO compare actions at shared states to compute relative advantages, but estimate each action's value from a single sampled continuation. The resulting local credit remains sensitive to decisions made later in the trajectory. To address this problem, we propose...

    arxiv.org/abs/2610.37195 · PDF

  42. 42

    MLE-World: Predicting Machine Learning Experiment Outcomes with a Learned World Model

    Brian Ko, Han Wang, Hejia Geng, Li Erran Li, Sherry Yang

    cs.LG · cs.AI

    AI research agents can propose, execute, and improve programs that solve machine learning tasks. However, executing each candidate program requires training and evaluating ML models, limiting how many candidates an agent can explore at each step. We introduce MLE-World, a world model finetuned from Qwen3.5-4B that predicts a program's terminal output (error traces, training logs, and validation scores) from its code and data description...

    arxiv.org/abs/2610.37175 · PDF

  43. 43

    Physics-Guided Flow Matching: A Generative AI Framework for Global Total Electron Content Forecasting

    Si-Tong Yang, Feng Ye, Xian Lu, Dong Lin, Benjamin C. Martinez

    cs.LG · physics.space-ph · stat.ML

    Accurate forecasting of global total electron content (TEC) is important for space weather monitoring and the reliable operation of satellite communication and navigation systems. We propose a physics-guided generative artificial intelligence (AI) framework for day-ahead global TEC forecasting. The framework builds on flow matching, which learns to gradually transform samples from an initial distribution into new samples that share the...

    arxiv.org/abs/2610.37159 · PDF

  44. 44

    Variational Optimal Control for Diffusion Guidance

    Ruoyu Wang, Xinyi Ye, Yunshuo Shi, Liangyu Yuan, Beier Zhu, Chi Zhang

    cs.LG

    We present a variational control framework that connects sampling-time guidance and reward-based model adaptation through a shared Jacobian pullback. Starting from terminal reward maximization with a path-space KL penalty, the Gibbs variational principle and Doob transform yield a correction to the generative dynamics. A local denoising approximation exposes the pullback that transports output preferences to the controlled variable. This...

    arxiv.org/abs/2610.37146 · PDF

  45. 45

    Beyond Parameter Protection: Fisher-Shaped EWC for Replay-Free Continual Learning

    Liying Wang, Simone Magistri, Joost van de Weijer

    cs.LG · cs.AI

    In replay-free continual learning, Elastic Weight Consolidation (EWC) uses Fisher information matrix (FIM) to protect acquired knowledge by penalizing changes to historically important parameters. However, a local importance estimate at the retained solution does not in general characterize sensitivity after finite parameter displacement. This motivates controlling Fisher sensitivity over a neighborhood of potential future updates. We propose...

    arxiv.org/abs/2610.37145 · PDF

  46. 46

    Generative Replay for Equilibrium Propagation: Toward Edge-Friendly Continual Learning in Hopfield Networks

    Victor Cook, Gilles Sassatelli, Thierry Gil, Mohamed Watfa, Marina Reyboz

    cs.LG · cs.AI

    Equilibrium Propagation (EqProp) offers a biologically grounded and hardware-aligned alternative to backpropagation for training Hopfield Networks. However, its behavior in continual learning settings remains largely unexplored. This work confronts catastrophic forgetting in EqProp-trained Hopfield Networks from the inside out, eliminating the need for external data buffers or architectural overhead that standard Experience Replay...

    arxiv.org/abs/2610.37139 · PDF

  47. 47

    Mechanistic Failure Modes of Weight Pruning in Transformer Language Models

    Nabeelah Faumi, Deshan Sumanathilaka

    cs.LG · cs.CL

    Pruned language models are judged almost entirely by accuracy and perplexity, which assumes that preserved output behaviour implies preserved internal computation. We test this assumption with a factorial pruning study on GPT-2, crossing two importance criteria (magnitude, random), two component types (attention heads, MLP neurons) and three sparsity levels (3%, 50%, 70%). Using the fully characterised indirect object identification (IOI)...

    arxiv.org/abs/2610.37126 · PDF

  48. 48

    Self-Regularized Flow Language Models

    Theodoros Kouzelis, Spyros Gidaris, Nikos Komodakis

    cs.LG · cs.CL

    We introduce SERF (Self-Regularized Flow Language Models), a simple self-supervised framework for improving continuous diffusion language models by exploiting the denoising trajectory itself. While continuous flow-based language models offer a promising alternative to autoregressive generation, their training objectives typically rely on supervision from clean target tokens, leaving the denoising trajectory underexploited as a learning...

    arxiv.org/abs/2610.37115 · PDF

  49. 49

    On the Sparsity of Direct Preference Optimization: Weight Disentanglement in the NTK Regime

    Kensuke Sasaki, Issei Sato

    cs.LG

    Direct Preference Optimization (DPO) has been widely adopted for aligning language models with human preferences. Recent empirical work has revealed that DPO induces remarkably sparse parameter updates, yet no theoretical explanation for this phenomenon has been offered. We analyze DPO's optimization dynamics under a neural tangent kernel linearization and show that the logistic structure of the DPO loss gives rise to an implicit...

    arxiv.org/abs/2610.37085 · PDF

  50. 50

    EnergyCal: Taming Loss Spikes in Muon-Family Optimizers for Sparse MoE Training

    Masato Fujitake

    cs.LG · cs.AI · cs.CL

    Sparse mixture-of-experts (MoE) language-model training can exhibit large loss spikes with Muon-family optimizers. We introduce EnergyCal, a lightweight rule that attenuates routed-expert updates according to the normalized squared Frobenius norm of their finite-step Newton-Schulz directions. EnergyCal leaves the spectral transform unchanged and adds no persistent optimizer state. We evaluate it on a 4B-parameter MoE model. Against a...

    arxiv.org/abs/2610.37062 · PDF

  51. 51

    Exploring the Limits of the Delta Learning Hypothesis: Single-Pair Generator Biases in Direct Preference Optimization

    Giuseppe Curci, Stefano Simonazzi, Andrea Molinari, Andrea Zugarini

    cs.LG · cs.AI · cs.CL

    Direct Preference Optimization (DPO) typically relies on preference pairs generated from diverse model pools and ranked by an external LLM judge. The recently proposed delta-learning hypothesis suggests that comparable supervision can instead be obtained from a single pair of models by treating the larger model's response as preferred. We investigate the limits of this approach across model families and scales. We show, both theoretically and...

    arxiv.org/abs/2610.37058 · PDF

  52. 52

    Crop Yield Prediction Using Regression Models: A Comparative Evaluation of Model Performance

    Saptak Das, Arijit Sarkar, Soumik Neogi, Suborno Naskar

    cs.LG · cs.AI

    Reliable crop-yield prediction can support agricultural planning and resource allocation. This study evaluates five regression models for predicting crop yield from environmental, crop-related, and agricultural management variables. The dataset initially contained 1,000,000 observations and ten variables, with 231 observations having negative target values removed during preprocessing. Categorical variables were represented using one-hot...

    arxiv.org/abs/2610.37022 · PDF

  53. 53

    V-LUMEN: Visual Lookup Memory for Embedding Scaling in Vision-Language Models

    Miso Choi, Daekeun Kim, Eunji Kim, Jungbeom Lee

    cs.LG · cs.AI · cs.CL · cs.CV

    Embedding-level scaling expands the capacity of language models through lookup-based memory rather than by increasing model depth, width, or dense computation. Extending this idea to vision-language models (VLMs) is challenging because visual features are continuous, spatially structured, and query-dependent, unlike discrete text tokens that naturally support n-gram lookup. We introduce V-LUMEN, a visual lookup memory for VLMs. V-LUMEN...

    arxiv.org/abs/2610.37019 · PDF

  54. 54

    Compote: Collapse-Resistant Multiple Hypothesis Learning For Sequential Recommendation

    Louis Castres Saint Martin, Clement Calauzenes, Alain Rakotomamonjy, Patrick Gallinari

    cs.LG · cs.IR

    Next-item prediction is inherently ambiguous: a single interaction history often admits multiple plausible continuations across distinct catalogue regions. Standard sequential recommenders compress these alternatives into a single prediction, conflating user intents. We introduce COllapse-resistant Multiple hyPOThesis lEarning (COMPOTE), a multi-hypothesis transformer trained with an annealed objective that transitions from uniform...

    arxiv.org/abs/2610.37005 · PDF

  55. 55

    Meta Models: Models that Model Learning

    Peihang Li, Ziming Liu

    cs.LG

    Every neural-network training run produces not only a model, but also evidence about how models learn. Can these accumulated experiments enable us to predict and compare new training runs before executing them? We study meta-models: models that predict the learning behavior of other models. From this perspective, final-performance prediction, learning-curve prediction, and hyperparameter selection are different queries to the same training...

    arxiv.org/abs/2610.37001 · PDF

  56. 56

    CaLLM: Continual Adaptation of LLMs via Non-Parametric Adapter Routing

    Bilge Celik, Anna Vettoruzzo, Joaquin Vanschoren

    cs.LG

    Continual adaptation of large language models (LLMs) to evolving task distributions while mitigating catastrophic forgetting remains an open problem. Existing strategies often struggle to scale or lack a robust mechanism for zero-shot task-to-expert assignment. We introduce CaLLM, a framework that resolves these tensions by combining metric-based meta-learning with modular parameter isolation via lightweight adapters. First, CaLLM projects...

    arxiv.org/abs/2610.36994 · PDF

  57. 57

    Advancement or Acceleration? How Data Curation Improves LLM Training

    Jiaxing Wang, Zicheng Zhang, Deping Xiang, Jin Xu, Yong Li, Chao Liu, Pengzhang Liu, Qixia Jiang

    cs.LG · cs.AI

    While data curation is widely credited with improving large language model training, a basic question remains unresolved: does curation expand what a model can ultimately learn, or merely reach the same solution faster? In this paper, we formalize this distinction between advancement (improving the attainable performance) and acceleration (reaching the same performance with less data). Through an analytic characterization and real-world LLM...

    arxiv.org/abs/2610.36983 · PDF

  58. 58

    GNN-Guided Network Reduction for Centralized Urban Traffic Routing

    Nikolai Antonov, Lukáš Chrpa, Andrii Nyporko, Jan Hůla

    cs.LG

    We consider the centralized traffic routing (CTR) problem, where a central planner assigns routes to origin-destination (OD) vehicle flows to minimize congestion-dependent segment cost. Prior work formulates CTR efficiently as a mixed-integer linear program (MILP), but scaling this formulation to the full road network is computationally prohibitive within an operational time budget. Bounded Routes and Smart Routes speed up optimization using...

    arxiv.org/abs/2610.36981 · PDF

  59. 59

    On the VC Dimension of Autoregressive Linear Threshold Functions

    Ilan Doron-Arad, Idan Mehalel, Elchanan Mossel

    cs.LG

    Linear threshold functions (LTFs) are a canonical class of Boolean predictors. In the autoregressive setting, the same LTF reads the most recent $d$ bits, generates the next bit, and returns the bit produced after $T$ steps. While the one-step class has VC dimension $d$, for general $T$ the counting upper bound is $O(d^2)$. We prove that the VC dimension is $Θ(d^2)$ for every $T\ge d^2$, resolving the quadratic gap for deterministic...

    arxiv.org/abs/2610.36970 · PDF

  60. 60

    Live-OPSD: On-Policy Self-Distillation for Proactive Streaming Large Video Models

    Kaixuan Du, Xin Wan, Dai Guan, Ming Chen, Xiaoxi Jiang, Guanjun Jiang

    cs.LG

    Streaming multimodal assistants must make response decisions while events are still unfolding. They need to recognize when the available evidence warrants a response, produce an appropriate continuation, and remain silent otherwise. Supervised fine-tuning learns from recorded assistant turns, but does not directly train on the response prefixes produced by the evolving policy itself. We introduce Live-OPSD, an on-policy self-distillation...

    arxiv.org/abs/2610.36959 · PDF

  61. 61

    DCKV: Learned Transform Coding of the Prompt KV Cache in Hybrid Language Models

    Egor Maximov, Aleksei Merkelov, Vasiliy Alekseev, Ivan Krivorotov, Valentin Beliaev, Nikita Arsenin, Dmitry...

    cs.LG · cs.AI

    Hybrid language models keep a key-value (KV) cache only in their few full-attention layers, but this cache is still the part of a prompt's state that grows with the prompt and that must be stored or moved to reuse it. We present DCKV, a differentiable codec for the KV cache and the first KV-cache codec that is trained end to end. Its transform, its quantizer and its entropy models, which are conditioned on the prompt tokens and on already...

    arxiv.org/abs/2610.36950 · PDF

  62. 62

    Algorithmic Monoculture in Machine Learning-based Site-Specific Fertilizer Recommendation

    Abdelghani Belgaid

    cs.LG · cs.CY · econ.TH

    Machine learning recommendations are often evaluated one model at a time, despite deployment across interconnected decision makers. For continuous NPK fertilizer prescriptions, correlated errors can aggregate across farms and generate systemic costs. A model of shared and farm-specific recommendation errors links expected social loss to model adoption shares; the Herfindahl-Hirschman Index characterizes concentration effects when common model...

    arxiv.org/abs/2610.36941 · PDF

  63. 63

    TDLM: Adaptive Temporal Integration for Diffusion Language Models

    Bohang Sun, Max Zhu, Francesco Caso, Christopher Irwin, Jialin Yu, Pietro Liò

    cs.LG

    Masked diffusion language models repeatedly compute representations for unresolved tokens, but conventional denoisers do not explicitly retain these representations for reuse across steps. We introduce Temporal Self-Conditioning Diffusion Language Model (TDLM), which lets each unresolved position attend to its own history of hidden representations within a frozen denoiser. This gives later predictions access to computation already performed....

    arxiv.org/abs/2610.36940 · PDF

  64. 64

    Gaussian Process Bandits for Constraint Black Box Lever Optimization

    Bram van den Akker, Hidde Fokkema, Anastasiia Kornilova

    cs.LG

    Optimizing discrete allocation weights across competing segments of a shared budget is a fundamental challenge in resource-constrained, sequential decision-making environments. In practice, this optimization is often hindered by delayed, noisy, and aggregate feedback, alongside strict operational constraints on how frequently weights can be adjusted. Because of the resulting combinatorial action space and extreme data sparsity, classical...

    arxiv.org/abs/2610.36935 · PDF

  65. 65

    Provenance Signatures for Low Rank Adaptations of Image Diffusion Models

    Tobias Cheung, Mohan Dantam, Marc Juarez, Rik Sarkar

    cs.LG

    Diffusion models have revolutionized the generation of realistic artificial images. Efficient techniques such as Low Rank Adaptation (LoRA) can modify these models for particular use cases, making them more diverse and widely usable. At the same time, open base models may be adapted to deliver biased, harmful or illegal images in an online service. In this paper, we develop a protocol to detect when a service may be using an adaptation of an...

    arxiv.org/abs/2610.36933 · PDF

  66. 66

    HiLo-RD: Cross-Budget Self-Distillation for Masked Diffusion Language Models

    Xiang Li, Chengting Yu, Jungang Li, Yilun Zheng, Junzhuo Li, Xingyao Zhang, Wenbo Su, Xuming Hu

    cs.LG · cs.CL

    We propose HiLo-RD (High-to-Low Budget Representation Distillation) to improve masked diffusion language models (MDLMs) under a limited generation budget. Unlike autoregressive models, MDLMs choose both what to generate and where to commit it. This creates two coupled challenges: predicting useful content before reasoning is complete, and deciding when to fix answer tokens that shape later predictions. We identify budget-induced premature...

    arxiv.org/abs/2610.36901 · PDF

  67. 67

    Hyperparameter and Neural Architecture Search via Evolutionary Optimization for Reinforcement Learning: HPNAS-RL

    Nihal Acharya Adde, Hanno Gottschalk

    cs.LG

    In this study, we propose a dual-branch evolutionary framework that simultaneously performs Neural Architecture Search (NAS) and Hyperparameter Optimization (HPO) for Reinforcement Learning (RL) in Autonomous Driving (AD) tasks. While NAS has shown strong potential in automating architecture design for Convolutional Neural Networks (CNNs), its application to RL remains limited, especially in scenarios that require joint optimization of...

    arxiv.org/abs/2610.36882 · PDF

  68. 68

    Tail Cues, Principal Corrections: Plug-and-Play Rectification for Open-Set Test-Time Adaptation

    Yingkai Yang, Chaoqi Chen, Hui Huang

    cs.LG · cs.CV

    Open-Set Test-Time Adaptation (OSTTA) operates on non-stationary target streams where covariate and semantic shifts coexist. Existing TTA methods often obtain adaptation signals by updating model parameters or maintaining mutable memories, which can be costly and vulnerable to unknown samples. We present TALOR, a lightweight, backward-free rectification module that uses the source linear head as a geometric anchor. The head-induced basis...

    arxiv.org/abs/2610.36870 · PDF

  69. 69

    When All Rollouts Fail: Policy-Compatible Joint Decoding for Hard-Problem Reasoning

    Zehao Wang, Hanzhu Chen, Runquan Gui, Zhe Wang, Lin Yang

    cs.LG · cs.AI

    Reinforcement Learning with Verifiable Rewards (RLVR) derives its learning signal from binary rewards via on-policy exploration and becomes a powerful paradigm for improving the reasoning of large language models. Yet it faces a learning cliff on problems beyond model's current ability: when every rollout fails, the samples reveal nothing about a correct solution, and group-relative estimators (e.g., GRPO, RLOO) collapse the update to zero...

    arxiv.org/abs/2610.36849 · PDF

  70. 70

    SetGRPO: Calibrating Grader-Induced Policy Optimization

    Jiale Huang, Zixu Li, Zhiwei Chen, Zhiheng Fu, Jinhe Bi, Yupeng Hu

    cs.LG · cs.AI

    Rubric-based reinforcement learning uses fine-grained grader feedback to train large language models on open-ended tasks, but grader randomness can alter GRPO updates in ways that score discrepancies alone cannot characterize. Because GRPO jointly normalizes rewards within each response group and combines the resulting training coefficients with response-level gradients, a scoring discrepancy can shift the coefficients of the entire group and...

    arxiv.org/abs/2610.36807 · PDF

  71. 71

    LLM knowledge for improved Out-of-Distribution Tabular Prediction - A Cross-Validated Regularisation Approach

    Sølve Langseth, Kristoffer Herland Hellton, Hugo L. Hammer

    cs.LG · stat.ME · stat.ML

    Prediction models degrade when the deployment population differs from the one used to train them, and training data cannot say which feature-response relationships will transfer. Expert elicitation is slow and unscalable. We ask whether coefficient priors elicited from large language models (LLMs) improve out-of-distribution (OOD) prediction as the shrinkage target of a regularised regression instead of zero, across five tabular datasets...

    arxiv.org/abs/2610.36806 · PDF

  72. 72

    API-TAIL: Detecting Model Substitution in LLM APIs through Temperature-Aligned Likelihood Tails

    Haocheng Mei, Yuchuan Tian, Ye Yuan, Chao Xu, Allen Chen, Kai Han, Yunhe Wang

    cs.LG · cs.CL

    Commercial language model APIs return only text, allowing a provider to cut costs by quietly routing a fraction of user requests to a cheaper substitute model while billing for the advertised one. Detecting such partial substitution is challenging because honest providers also alter the output distribution through undocumented decoding and serving choices, such as temperature scaling, sampling truncation, repetition and presence penalties,...

    arxiv.org/abs/2610.36801 · PDF

  73. 73

    Select-TTT: Query-Conditioned Selective Fast Weight Updates for In-Place Test-Time Training

    Yue Wu, Yang Zhang, Weinan Gan, Chongming Gao, Xinsen Guo, Xiaoyuan Li, Fuli Feng

    cs.LG · cs.AI · cs.CL

    Test-time training (TTT) enables a deployed language model to internalize incoming context on-the-fly into a small set of fast-weight parameters, rather than merely treating it as transient context for attention. In-Place TTT recently made this idea practical for large language models by repurposing the down-projection of standard MLP blocks as fast weights. However, we find that its benefit largely disappears on realistic web long-context...

    arxiv.org/abs/2610.36794 · PDF

  74. 74

    Does LLM Pre-Training Typically Occur at the Edge of Stability?

    Yuhang Cai, Haodong Wen, Haofeng Huang, Deyi Liu, Yiyuan Ma, Kaifeng Lyu

    cs.LG

    Quadratic approximations provide a common lens for studying neural network optimization, but recent evidence challenges their predictive validity. In full-batch gradient descent with learning rate (LR) $η$, Cohen et al. (2021) observed the Edge of Stability (EoS), where the largest Hessian eigenvalue concentrates near $2/η$, seemingly at odds with classical stability conditions predicted by quadratic approximations. In this work, we revisit...

    arxiv.org/abs/2610.36778 · 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.