跳到正文
研踪PaperTrail

版本更新

查看已发布论文的最新修订,在详情页对照实际观测到的版本。

部分采集异常

全站更新概况

+1084新增论文
+883其中 AI 相关
+130已有论文修订
+94新增 AI 导读

arXiv 最近完成扫描:2026-11-18 04:28 · 全站最近入库:2026-11-18 19:45各信源进度 →

统计口径与数据新鲜度

统计 2026-11-18 06:48 — 2026-11-19 06:48(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-18 04:28最近发现新增版本:2026-11-18 19:45库内论文最新更新时间:2026-11-18 08:00

每分钟刷新已保存的数据。空检查不代表采集异常;新收录可能包含历史论文。AI 导读按新增版本计数。

26,870 已收录论文22,374 AI 相关8,156 有修订112 HF 社区发现

论文修订列表

时间范围跟随排序:版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常,查看状态 →

找到 8,156 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-18 04:27(北京时间)
版本更新 · v2arXiv:2604.24737

Learning to Think from Multiple Thinkers: How Chain-of-Thought Heterogeneity Affects Efficient Learnability

Nirmit Joshi, Roey Magen, Nathan Srebro, Nikolaos Tsilivis, Gal Vardi

原文摘要

Chain-of-thought (CoT) supervision makes computationally hard learning problems tractable by exposing intermediate steps that are easy to learn via next-token prediction. However, in practice, CoT datasets draw on heterogeneous sources with different reasoning styles, and substantial effort goes into curating supervision suitable for next-token training. What role does this curation play in enabling the computational advantage of CoT supervision? In this work, we take a first step toward understanding this question from a learning-theoretic perspective. We introduce several models of $heterogeneous$ $CoT$ $supervision$ from multiple sources, each providing correct traces with the same final answer but different intermediate steps. Assuming that fitting a next-token predictor to $homogeneous$ traces from any single source is computationally tractable, we ask when this suffices for efficient learning from heterogeneous supervision. We show $cryptographic$ $hardness$ of learning from $passively$ collected traces from multiple sources, even though learning from homogeneous CoTs is easy. This hardness can stem either from $hard$-$to$-$learn$ $contribution$ $biases$ determining which sources contribute traces on different parts of the domain, or from receiving an $unfiltered$ heterogeneous mixture (though a filtered subset of homogeneous traces is sufficient for efficient learning). On the positive side, inspired by iterative and adaptive fine-tuning of language models, we give a computationally efficient learning algorithm that $actively$ and $adaptively$ collects CoTs from multiple sources with a per-source CoT budget completely independent of the target error $\varepsilon$. We argue that our results illuminate data collection and curation practices central to making CoT supervision effective for learning.

cs.LG · cs.AI · cs.CC · stat.ML

版本更新 · v2arXiv:2601.21582

Depth-Recurrent Attention Mixture (Dreamer): Giving Latent Reasoning the Attention it Deserves

Jonas Knupp, Jan Hendrik Metzen, Jeremias Bohn, Georg Groh, Kristian Kersting

原文摘要

Depth-recurrence facilitates latent reasoning by sharing parameters across depths. However, prior work lacks combined FLOP-, parameter-, and memory-matched baselines, underutilizes depth-recurrence due to partially fixed layer stacks, and ignores the bottleneck of constant hidden-sizes that restricts many-step latent reasoning. To address this, we introduce a modular framework of depth-recurrent attention mixtures (Dreamer), combining sequence attention, depth attention, and sparse expert attention. It alleviates the hidden-size bottleneck through attention along depth, decouples scaling dimensions, and allows depth-recurrent models to scale efficiently and effectively. Across language reasoning benchmarks, our models require 2 to 8x fewer training tokens for the same accuracy as FLOP-, parameter-, and memory-matched SOTA, and outperform ca. 2x larger SOTA models with the same training tokens. We further present insights into knowledge usage across depths, e.g., showing 2 to 11x larger expert selection diversity than SOTA MoEs.

cs.AI · cs.CL · cs.LG

版本更新 · v2arXiv:2611.02775

Human-Centred, Risk-Constrained Evaluation of Financial Document Recognition: Evidence, Critical Errors, and Selective Automation

Subodh Lonkar

原文摘要

Financial document recognition is often assessed using character and word error rates, although these metrics do not express whether a system preserves the financial facts on which people and organisations act. This paper develops a human-centred, risk-constrained framework for evaluating financial document recognition as evidence recovery and decision support. It defines typed financial facts linked to visible source regions and separates verbatim transcription, deterministic normalization, and business interpretation. A layered quality model covers text, localization, layout, tables, fields, fact correctness, evidence grounding, uncertainty, robustness, and operational performance. Non-compensatory acceptance gates use one-sided uncertainty bounds for critical-field errors, preventing strong aggregate scores from masking failures in amounts, identifiers, signs, periods, or entity associations. Selective automation is formulated as a constrained routing problem that accounts for residual risk, human-review capacity, and review cost. The framework further specifies annotation and adjudication procedures, leakage-resistant benchmark partitions, statistical estimands, subgroup analyses, and a reproducibility card. Formal results establish the limitations of compensatory scoring, the non-zero upper risk bound after zero observed errors, and a loss sensitive decision rule for routing uncertain cases to human review. The contribution is theoretical and methodological; no new model, dataset, or empirical performance result is claimed.

cs.CV · cs.AI

版本更新 · v3arXiv:2603.04314

MOO: A Multi-view Oriented Observations Dataset for Viewpoint Analysis in Cattle Re-Identification

William Grolleau, Achraf Chaouch, Astrid Sabourin, Guillaume Lapouge, Catherine Achard

原文摘要

Animal re-identification (ReID) faces critical challenges due to viewpoint variations, particularly in Aerial-Ground (AG-ReID) settings where models must match individuals across drastic elevation changes. However, existing datasets lack the precise angular annotations required to systematically analyze these geometric variations. To address this, we introduce the Multi-view Oriented Observation (MOO) dataset, a large-scale synthetic AG-ReID dataset of $1,000$ cattle individuals captured from $128$ uniformly sampled viewpoints ($128,000$ annotated images). Using this controlled dataset, we quantify the influence of elevation and identify a critical elevation threshold, above which models generalize significantly better to unseen views. Finally, we validate the transferability to real-world applications in both zero-shot and supervised settings, demonstrating performance gains across four real-world cattle datasets and confirming that synthetic geometric priors effectively bridge the domain gap. Collectively, this dataset and analysis lay the foundation for future model development in cross-view animal ReID. MOO is publicly available at https://github.com/TurtleSmoke/MOO.

cs.CV · cs.AI

版本更新 · v2arXiv:2610.25608

Hardness Amplification for Approximate Nash Equilibrium with Constantly Many Players

Xiaotie Deng, Ningyuan Li, Zihe Wang, Jie Zhang

原文摘要

For every fixed integer $k\ge2$ and fixed parameters $δ,\varepsilon\in(0,1)$ and $η\in[0,1)$, we give a deterministic polynomial-time reduction from finding a $δ$-approximate Nash equilibrium ($δ$-ANE) in a $k$-player game $G$ with payoffs in $[0,1]$ to finding an $(\varepsilon,η)$-weak approximate Nash equilibrium (weak ANE) in a game $H$ with $C$ players, where $C$ depends only on $k,δ,\varepsilon,η$, independently of the action-set sizes. The constructed game is symmetric with binary payoffs. The games and strategies are represented by complete payoff tables and probabilities, with rational entries encoded in binary. Consequently, the existence of deterministic polynomial-time weak-ANE algorithms at one fixed pair of parameters $\varepsilon\in(0,1),η\in[0,1)$ for every fixed number of players would give deterministic polynomial-time $δ$-ANE algorithms at every fixed approximation parameter $δ\in(0,1)$ and number of players. Under deterministic PPAD-ETH, combining our reduction with Rubinstein's hardness result (FOCS 2016) for bimatrix games at some fixed approximation parameter $γ_0\in(0,1)$ rules out a deterministic algorithm for $(\varepsilon,η)$-weak ANE with running time $N^{O((\log N)^{1-ξ})}$ for any fixed $ξ\in(0,1)$. This holds for every fixed $\varepsilon\in(0,1)$ and $η\in[0,1)$ at a suitable constant number $C$ of players, where $N$ is the binary encoding length of the constructed game's complete payoff table.

cs.GT · cs.CC

版本更新 · v4arXiv:2604.20575

A Quadratic Lower Bound for Noncommutative Circuits

Pratik Shastri, Sophus Valentin Willumsgaard

原文摘要

We prove that every fan-in 2 noncommutative arithmetic circuit computing the palindrome polynomial $\mathrm{Pal}_{n,d}(X,Y)$ of degree $2d$ in $2n$ variables has size $Ω(nd)$, which is tight up to a constant. In particular, when $d = n$ we obtain an $Ω(n^2)$ lower bound, improving the previous lower bound of $Ω(n \log n)$ by Nisan and concurrent work of $Ω(n^{1.5})$ by Raz (FOCS 2026). The proof builds on and refines a previous work of the first author, by constructing for every gate a small set that spans a subset of the coefficients of the polynomial computed by that gate. We show that the structure of the set depends only on the subsets of multiplication gates along paths going from that gate to a leaf. From this description, we are then able to get an upper bound on this set based on the amount of product gates in the circuit. The proof is surprisingly short and simple to understand, considering that improving Nisan's $Ω(n \log n)$ lower bound for noncommutative circuits has been an open problem for decades.

cs.CC

版本更新 · v2arXiv:2610.20171

FreeSBee: A Constant-Time Compilation Pipeline

Julian Ganz, David Knothe, David Oswald, Sebastian Reiter, Thorben Schneider, Simon Wegener

原文摘要

Constant-time execution is an important security property, especially for IoT devices, that prevents the leakage of sensitive information over timing side-channels. However, achieving constant-time execution is very hard when going down to the cycle level, at least on realistic devices. A lot of research is concerned with producing tools that either determine whether code is constant-time or that ensure it is compiled into a constant-time binary, with various degrees of accuracy. In this work, we present and evaluate a full workflow for hardening against timing side-channels during compilation. It starts with taint analysis to determine the code locations that depend on sensitive information. These are then the object of automated control-flow transformations inside the compiler to seal possible timing side-channels. We evaluate our workflow in detail on a realistic application running on a RISC-V core using both static timing analysis and statistical methods. Even though we do not reach cycle-accurate constant-time execution, our statistical results suggest that our hardened binary is safe against timing attacks.

cs.CR

版本更新 · v2arXiv:2609.31140

Can Linguistic Reasoning Vectors Enhance Multimodal Reasoning Ability?

Ziyi Wang, Li Li, Aolin Zhou, Yankun Shen, Chonghan Liu, Shuxia Lin, Xu Yang

原文摘要

Most Vision-Language Models (VLMs) are built by extending pretrained Large Language Models (LLMs) with visual modules and multimodal alignment. However, this multimodal scaling often degrades the language-side reasoning ability originally encoded in the base LLM. While the base LLM retains usable reasoning after scaling, the aligned VLM itself cannot reliably access this ability. Therefore, recovering the degraded reasoning capability in VLMs would benefit more from seeking help from the base LLM than from the VLM alone. Motivated by this, we propose LIFT (Language-side reasonIng Facilitation and Transfer), a lightweight vector-intervention method that transfers reasoning capability from the base LLM to the VLM without retraining the backbone. LIFT defines Reasoning Vectors as answer-token hidden-state differences between a Reasoner path with an explicit reasoning trace and a Solver path without it, and injects these vectors into language-side activations of the target VLM. LIFT further supports learnable vector adaptation while keeping the VLM backbone frozen. We evaluate LIFT on two VLMs across six reasoning benchmarks, comparing Reasoning Vectors extracted from the base LLM and from the aligned VLM under matched protocols. Results show that LLM-derived vectors consistently outperform VLM-derived vectors, confirming that the base LLM is a more effective source for recovering reasoning. LIFT partially recovers degraded reasoning through lightweight language-side interventions. Further analyses show that Reasoning Vectors influence intermediate reasoning behavior rather than merely altering final answers. The source code will be released soon.

cs.AI

版本更新 · v2arXiv:2611.05229

Santa Claus and Weighted Nash Social Welfare with Subadditive Valuations

Yuhang Chen, Yuda Feng, Weijiang Hu, Shi Li, Ruilong Zhang

原文摘要

We study max-min allocation (the Santa Claus problem) and weighted Nash social welfare for $n$ agents with subadditive valuations over $m$ indivisible items. Using demand queries, we give randomized polynomial-time approximation algorithms with factors $O(\sqrt n)$ and $O(Γ(w))$, respectively, where $Γ(w)=\exp((\ln n-H(w))/2)$ and $H(w)=-\sum_i w_i\ln w_i$ for positive weights summing to one. Both ratios are independent of $m$. The max-min result improves the $2n-1$ bound of Khot and Ponnuswami (2007). We also give simpler $O(\sqrt n)$ algorithms for additive valuations and for submodular valuations with value queries, improving the $O(\sqrt{n\log n})$ bound of Saha and Srinivasan (2018) in the additive case. For Nash welfare, $Γ(w)=1$ for equal weights and $Γ(w)\le\sqrt n$ in general. We prove a configuration-LP integrality gap of $Θ(Γ(w))$ for every prescribed weight vector, even for XOS valuations. For every fixed $D>0$, we also prove information-theoretic approximation lower bounds of $Ω_D(\sqrt{n/\ln n})$ and $Ω_D(Γ(w)/\sqrt{\ln n})$, respectively, for randomized algorithms making at most $n^D$ adaptive queries. These hold on XOS instances with $m=Θ(n)$ items, even with simultaneous access to value, demand, and XOS supporting-clause oracles. They apply both to guarantees with success probability at least $2/3$ and to guarantees in expected welfare. The weighted bound holds for every prescribed weight vector and is vacuous below one. Thus the oracle lower bounds are within an $O(\sqrt{\ln n})$ factor of our upper bounds, while the configuration-LP gaps match them up to constants. Our algorithms build on random matchings for large items and charge the aggregate loss across configurations to the at most $n$ matched items.

cs.DS

版本更新 · v4arXiv:2610.08327

MedZERO: Self-Evolving Agents for Open-Ended Medical Reasoning Through Controlled Knowledge Accumulation

Xilin Dang, Weilin Ruan, Xue Yang, Jinghao Wang, Xiaowei Hu, Jinpeng Li, Pheng-Ann Heng

原文摘要

Large language models (LLMs) have shown promise in medical question answering and clinical reasoning, yet their improvement remains constrained by static parametric knowledge and costly expert supervision. Self-evolving agents offer a promising alternative by enabling models to improve through iterative task generation and problem-solving. However, most existing self-evolving methods are designed for easily verifiable domains such as mathematics and coding, where solutions can be checked by exact answers or executable programs. Medical reasoning is fundamentally different: it is open-ended, knowledge-intensive, and often only partially verifiable. We present MedZERO, a self-evolving framework for open-ended medical reasoning. MedZERO couples an Examiner that generates frontier medical question-option pairs with a Reasoner that solves them through evidence-grounded multi-turn reasoning with external knowledge tools. To support reliable, continual improvement, MedZERO adopts controlled knowledge accumulation, which maintains temporary exploratory knowledge and curated persistent knowledge in reasoning. We evaluate MedZERO on five public medical reasoning benchmarks using 4B- and 8B-scale base models under open-ended evaluation. At both model scales, MedZERO achieves the highest average accuracy among the evaluated methods. Code is available at https://github.com/IcecreamArtist/MedZERO.

cs.AI

版本更新 · v3arXiv:2606.03357

The Unsampled Truth: Quantifying Prompt Artifacts in LM Psychometrics

Nils Schwager, Christoph Hau, Simon Münker, Achim Rettinger

原文摘要

When prompting language models for psychometric assessment, researchers assume that the responses reflect the injected persona and the meaning of the survey item. We test this premise using a diagnostic design that crosses five semantically distinct baseline personas with five semantically equivalent variants of each of four prompt components (persona wording, task instruction, item wording, option symbol). Measuring the 1-Wasserstein distance between the resulting response distributions and partitioning the variation among the five components allows for the separation of target effects from prompt artifacts. We apply the framework to 13 open-weight small language models (0.6B to 14B) on the Big Five Inventory and the Short Dark Triad. We find that in most models, the task instruction and option symbol displace response distributions further than paraphrasing the persona description or the item itself. For a substantial share of items, the artifact share of explained variation exceeds 50%; non-semantic changes of the prompt account for more response variation than the baseline personas. Our framework lets researchers quantify these prompt artifacts before interpreting psychometric output.

cs.CL · cs.AI

版本更新 · v2arXiv:2610.31166

Memory before Goal: Role of Pre-History Memory for Humanoid Loco-Manipulation

Chenyuan Chen, Ran Ding, Lechen Zhang, Zewen He, Yoshihiko Nakamura

原文摘要

Humanoid loco-manipulation in dynamic scenes is non-Markovian: cues vanish as objects change or are hidden, or leave the view as the robot turns and walks, yet whole-body vision-language-action (VLA) policies see only the present. We present WholeBodyMem, an object-guided perceptual memory that keeps the pre-history of object events seen before the robot acts on its goal. SAM3 tracks the task objects, and their events causally select key frames within an adaptive budget, which the action expert reads by cross-attention together with object tokens encoding appearance, geometry and time. We also contribute HuMeGo, a humanoid memory benchmark with six simulated and six real-world Unitree G1 tasks. Keeping the right frames rather than more frames, WholeBodyMem surpasses memoryless and uniform-memory baselines, raising its base policy's success from 29.1% to 72.6% in simulation and from 36.7% to 88.3% on the real robot, while also improving the standard SIMPLE tasks (70.2% to 86.9%).

cs.RO · cs.CV · cs.LG

版本更新 · v5arXiv:2605.06152

Grokking or Glitching? How Low-Precision Drives Slingshot Loss Spikes

Hanqing Liu, Jianjun Cao, Yuanze Li, Zijian Zhou

原文摘要

Deep neural networks exhibit periodic loss spikes during unregularized long-term training, a phenomenon known as the "Slingshot Mechanism." Existing work usually attributes this to intrinsic optimization dynamics, but its triggering mechanism remains unclear. This paper proves that this phenomenon is a result of floating-point arithmetic precision limits. As training enters a high-confidence stage, the difference between the correct-class logit and the other logits may exceed the absorption-error threshold. Then during backpropagation, the gradient of the correct class is rounded exactly to zero, while the gradients of the incorrect classes remain nonzero. This breaks the zero-sum constraint of gradients across classes and introduces a systematic drift in the parameter update of the classifier layer. We prove that this drift forms a positive feedback loop with the feature, causing the global classifier mean and the global feature mean to grow exponentially. We call this mechanism Numerical Feature Inflation (NFI). This mechanism explains the rapid norm growth before a Slingshot spike, the subsequent reappearance of gradients, and the resulting loss spike. We further show that NFI is not equivalent to an observed loss spike: in more practical tasks, partial absorption may not produce visible spikes, but it can still break the zero-sum constraint and drive rapid growth of parameter norms. Our results reinterpret Slingshot as a numerical dynamic of finite-precision training, and provide a testable explanation for abnormal parameter growth and logit divergence in late-stage training.

cs.LG · cs.CL · math.OC · stat.ML

版本更新 · v2arXiv:2609.16401

A Resolution of Friedgut's Conjecture on Influential Coalitions

Eshan Chattopadhyay, Mohit Gurumukhani

原文摘要

We prove that, for every constant $\varepsilon>0$ and every function $f:Σ^n\to\{0, 1\}$, there is a coalition of $O(n/\sqrt{\log n})$ coordinates and a target output $b\in\{0, 1\}$ such that, after the remaining coordinates are sampled uniformly and independently, the coalition can choose its values to make the output equal to $b$ with probability at least $1-\varepsilon$. The bound is independent of the alphabet size and also holds for monotone Boolean functions on $[0,1]^n$, resolving a conjecture of Friedgut (Combinatorics, Probability and Computing, 2004). Unlike the Boolean cube setting, where Kahn, Kalai, and Linial (FOCS, 1988) give a coalition bound of $O(n/\log n)$, no sublinear bound independent of the alphabet size was previously known. In collective coin flipping, our result gives the first sublinear bound on the number of bad players needed to force a fixed output with probability at least $1-\varepsilon$ in any one-round protocol with independent uniform messages, regardless of the message length. A key ingredient in our proof is an encoding that lets us relate the influence of a function on a product space to the $p$-biased influence of the encoded function. We then rely on a structure theorem of Hatami (Annals of Mathematics, 2012) for functions with small $p$-biased influence to bias the encoded function.

cs.CC · cs.DM

版本更新 · v3arXiv:2509.26092

Hybrid Dual-Batch and Cyclic Progressive Learning for Efficient Distributed Training of Convolutional Neural Networks

Kuan-Wei Lu, Ding-Yong Hong, Pangfeng Liu, Jan-Jan Wu

原文摘要

Distributed deep learning is critical for training convolutional neural network (CNN) models on large-scale image datasets with numerous parameters. Current research primarily focuses on leveraging additional hardware resources and powerful computing units to accelerate the training process. As a result, larger batch sizes are often employed to speed up training. However, training with large batch sizes can lead to lower accuracy due to poor generalization. To address this issue, we propose the dual-batch learning scheme, a distributed training method built on the parameter server framework. This approach maximizes training efficiency by utilizing the largest batch size that the hardware can support while incorporating a smaller batch size to enhance model generalization. By using two different batch sizes simultaneously, this method improves accuracy with minimal additional training time. Additionally, to mitigate the time overhead caused by dual-batch learning, we propose the cyclic progressive learning scheme. This technique repeatedly and gradually increases image resolution from low to high during training, thereby reducing training time. By combining cyclic progressive learning with dual-batch learning, our hybrid approach improves both model generalization and training efficiency for image classification tasks. Experimental results with ResNet-18 demonstrate that, compared to conventional training methods, our approach improves accuracy by 3.3% while reducing training time by 10.1% on CIFAR-100, and further achieves a 34.7% reduction in training time on ImageNet.

cs.DC · cs.LG

版本更新 · v2arXiv:2509.25592

Machine Learning Algorithms for Improving Black Box Optimization Solvers

Morteza Kimiaei, Vyacheslav Kungurtsev

原文摘要

Black-box optimization (BBO) addresses problems where objectives are accessible only through costly queries without gradients or explicit structure. Classical derivative-free methods---line search, direct search, and model-based solvers such as Bayesian optimization---form the backbone of BBO, yet often struggle in high-dimensional, noisy, or mixed-integer settings. Recent advances use machine learning (ML) and reinforcement learning (RL) to enhance BBO: ML provides expressive surrogates, adaptive updates, meta-learning-based algorithm selection, and generative models, while the RL--BBO interface includes dynamic operator configuration, robust policy search, and meta-optimization across tasks. This paper surveys these developments, covering representative algorithms such as the modular model-based optimization framework (mlrMBO), zeroth-order adaptive momentum methods (ZO-AdaMM), automated BBO (ABBO), distributed block-wise optimization (DiBB), partition-based Bayesian optimization (SPBOpt), the transformer-based optimizer (B2Opt), diffusion-model-based BBO, surrogate-assisted RL for differential evolution (Surr-RLDE), robust BBO (RBO), coordinate-ascent MORE with step size adaptation (CAS-MORE), log-barrier stochastic gradient descent (LB-SGD), policy improvement with black-box (PIBB), and offline Q-learning with Mamba backbones (Q-Mamba). We further review benchmark efforts such as the NeurIPS 2020 BBO Challenge and the MetaBox framework. A controlled numerical study further confirms that ML-based candidate screening can improve the evaluation-budget performance of direct search and differential evolution, while the benefit remains solver-dependent.

cs.LG

版本更新 · v2arXiv:2611.05003

Unifying and Beyond the Theory of Pervasive Capability Degradation in LLM Post-training SFT from Feature Interference

Hang Chen, Jiaying Zhu, Wenya Wang

原文摘要

Pervasive Capability Degradation (PCD) refers to the phenomenon in which supervised fine-tuning (SFT), while adapting a large language model (LLM) to a target task, often erodes the general capabilities acquired during pre-training. Prevailing explanations attribute PCD to macroscopic distribution shift and therefore lack microscopic manipulability, leaving existing mitigation methods without a complete theoretical foundation. In this paper, we first develop a bottom-up geometric theory of PCD grounded in feature interference under superposition: narrow-domain SFT data disrupt the almost-orthogonal feature structure formed during pre-training, and the resulting interference drives activations across decision hyperplanes in the output Laguerre-Voronoi geometry, flipping predictions on general tasks. This theory unifies existing methods into six paradigms and reveals both the source of their efficacy and their limitations. We then argue that cross-entropy under one-hot supervision is structurally limited in resolving PCD on its own, and accordingly propose the AEGIS loss, which directly constrains the drift of activation margins along all competing token directions, thereby decoupling interference suppression from task adaptation. Across 3 model families and 14 SFT tasks, AEGIS outperforms representative methods of every paradigm on target tasks and pervasive capabilities simultaneously, at only about 1.3 times the computational cost of standard SFT. Its generality, scalability, and mechanistic interpretability offer a new perspective on mitigating PCD.

cs.LG

版本更新 · v3arXiv:2605.12887

EcoGEO: Trajectory-Aware Evidence Ecosystems for Web-Enabled LLM Search Agents

Hengwei Ye, Jiasheng Mao, Zhenhan Guan, Zheng Tian

原文摘要

Web-enabled LLM agents are changing how online information influences search outcomes. Existing Generative Engine Optimization (GEO) studies mainly focus on individual webpages. However, agentic web search is not a single-document setting: an agent may issue queries, crawl pages, follow links, reformulate searches, and synthesize evidence across multiple browsing steps. Influence therefore depends not only on page content, but also on how pages are organized, connected, and encountered along the agent's browsing trajectory. We study this shift through Ecosystem Generative Engine Optimization (EcoGEO), which treats GEO as an environment-level influence problem for web-enabled LLM agents. To instantiate this perspective, we propose TRACE, a Trajectory-Aware Coordinated Evidence Ecosystem. Given a recommendation query and a fictional target product, our method builds a controlled evidence environment that coordinates an agent-facing navigation entry page with heterogeneous support pages. These pages use shared terminology, internal links, and consistent product attributes to introduce, verify, and reinforce the target product. We evaluate our method on OPR-Bench, a benchmark for open-ended product recommendation. Experiments show that it consistently outperforms page-level GEO baselines in Recommendation. Trajectory-level metrics show improved initial target access and additional internal-link browsing, while follow-up-search effects vary across backbones and comparison methods. Together with the multi-page baseline and ablations, these findings support the role of evidence organization in shaping the agent's evidence-acquisition process. Overall, our findings support an ecosystem research paradigm for GEO, where web-enabled LLM agents are studied in relation to the broader evidence environments that guide search, browsing, and answer synthesis.

cs.IR · cs.AI

版本更新 · v4arXiv:2606.08493

Querying Counterfactuals on Tissue Graphs with Supervised Disentanglement

Abdul Moeed, Stefan Schrod, Martin Rohbeck, Marc Jan Bonder, Pavlo Lutsik, Oliver Stegle, Daniel Dimitrov

原文摘要

Tissue graph counterfactuals ask what a cell would express if its spatial neighborhood were altered, a question central to predicting cell behavior in tissues. Existing methods, however, present only specific instances of such queries. In this work, we first formalize tissue graph counterfactuals as a class of spatial interventions that either rewire connections between cells (edge perturbation) or modify the expression of their neighbors (node perturbation). We then introduce Cellina, a framework that uses supervised disentanglement to decompose a cell's intrinsic state from its spatial context, using the latter as a conditioning input for in-silico perturbation predictions. Across benchmarks spanning over 2.5 million spatially-resolved cells in observational and interventional screens, Cellina outperforms both spatially-informed and non-spatial competitors in in-silico perturbations, while training as fast as non-spatial methods. Moreover, we show that Cellina reveals cancer subdomains with distinct signaling programs in an unsupervised manner and enables targeted neighbor perturbation simulations.

q-bio.GN · cs.LG · stat.ML

版本更新 · v3arXiv:2507.21931

Post-Training Large Language Models via Reinforcement Learning from Self-Feedback

Carel van Niekerk, Renato Vukovic, Benjamin Ruppik, Hsien-chin Lin, Shutong Feng, Milica Gašić

原文摘要

Large Language Models (LLMs) often produce plausible but poorly-calibrated answers, limiting their reliability on reasoning-intensive tasks. Recent research suggests that Chain-of-Thought (CoT) reasoning paths are inherent in pre-trained LLMs and can be elicited by simply altering the decoding process, where the presence of a CoT path correlates with higher answer confidence. Building on these insights, we present Reinforcement Learning from Self-Feedback (RLSF), a post-training stage that utilises the model's intrinsic confidence as a self-generated reward. By generating multiple CoT decoding beams from a frozen LLM, we compute the confidence of each final answer span and rank the resulting traces accordingly to create synthetic preferences. These preferences are subsequently utilised to fine-tune the policy through standard preference optimisation, requiring no human labels, gold answers, or externally curated rewards. RLSF simultaneously (i) refines the model's probability estimates--restoring well-behaved calibration--and (ii) strengthens step-by-step reasoning, yielding improved performance on arithmetic reasoning and multiple-choice question answering. By converting a model's own uncertainty into structured self-feedback, RLSF affirms reinforcement learning on intrinsic model behaviour as a principled and data-efficient component of the LLM post-training pipeline. Our results demonstrate that leveraging these inherent reasoning capabilities provides a robust path for enhancing model reliability without manual prompt engineering or external supervision.

cs.CL · cs.AI

版本更新 · v2arXiv:2602.18742

RoboCurate: Harnessing Diversity with Action-Verified Synthetic Data for Robot Learning

Seungku Kim, Suhyeok Jang, Byungjun Yoon, Dongyoung Kim, John Won, Jinwoo Shin

原文摘要

Synthetic data generated by video generative models has shown promise for robot learning as a scalable pipeline, but it often suffers from noisy action labels due to imperfect video generation or inaccurate action prediction. Recently, vision-language models (VLMs) have been leveraged to validate video quality, but they cannot directly evaluate the generated actions themselves. To tackle this issue, we introduce RoboCurate, a novel framework for action-verified synthetic robot data that addresses inconsistencies between generated videos and annotated actions. Specifically, RoboCurate executes the actions predicted by an inverse dynamics model (IDM) in a simulator and verifies action consistency by measuring the motion agreement between the simulator rollout and the generated video. To better leverage such a framework, we further expand the scene diversity of synthetic data beyond the available dataset through off-the-shelf generative models with structured editing axes. We then use a VLM to pre-filter physically implausible generated videos before verifying their action labels. Across diverse manipulation scenarios, RoboCurate yields substantial relative improvements in success rates compared to using only real data, achieving +90.3% on GR-1 Tabletop (300 demos), +29.1% on DexMimicGen in the pre-training setup on simulation benchmarks, and +179.9% and +42.9% in the co-finetuning setup on challenging real-world ALLEX and OpenArm humanoid benchmarks, respectively.

cs.RO · cs.AI · cs.CV

版本更新 · v2arXiv:2610.38897

RARE: Response-Aware Rational Equilibrium in Budget-Constrained Pricing Games

Bhaskar Ray Chaudhury, Jugal Garg, Xin Huang, Eklavya Sharma

原文摘要

We study a pricing game in markets with budget-constrained buyers, where each seller offers a single good and sets its price. We consider both non-rivalrous goods, such as datasets, and rivalrous goods, such as physical goods. For non-rivalrous goods, each buyer chooses a utility-maximizing bundle within her budget. For rivalrous goods, an allocation rule selects a Pareto-optimal allocation subject to the buyers' budgets and the goods' supply constraints. Pure Nash equilibria often fail to exist in this game. A key reason is that Nash equilibrium assumes that the other sellers keep their prices fixed when one seller deviates. In practice, however, sellers harmed by a price change may respond, for example, by matching the new price. We propose a new equilibrium notion called Response-Aware Rational Equilibrium (RARE). At a RARE, every deviation by a seller can be deterred by a response from the other sellers, in which no responder earns less than she did right after the deviation. Every Nash equilibrium is a RARE. Assuming that the allocation rule is continuous in prices, we show that a RARE always exists for rivalrous goods, and for non-rivalrous goods when buyers have equal budgets. Moreover, the RARE that we construct extracts the buyers' entire budget. For non-rivalrous goods with unequal budgets, we show that an approximate RARE exists, where a deviating seller can increase her revenue by a factor equal to the ratio of the average budget to the minimum budget. We also study a natural dynamic in which sellers repeatedly raise their prices to earn the revenue that they could guarantee themselves by deviating. We show that, in the settings where we prove exact existence, this dynamic reaches a RARE with additive error $δ$ within $\lfloor B/δ\rfloor$ rounds, where $B$ is the buyers' total budget. This paper's main results have been formally verified in Lean 4.

cs.GT

版本更新 · v3arXiv:2606.15920

OmniOPSD: Rationale-Privileged On-Policy Self-Distillation for Affective Computing

Zebang Cheng, Shuimu Chen, Boxue Yang, Yuanshen Guan, Jingyi Chen, Zheng Lian, Xiaojiang Peng, Fei Ma 等 10 位作者

原文摘要

Multimodal large language models (MLLMs) have advanced affective computing by enabling nuanced reasoning over visual, acoustic, and linguistic evidence. A key approach to improving their generalization is outcome-based reinforcement learning, which converts existing label-level annotations into rewards for final-answer correctness. However, its sparse sequence-level supervision provides limited fine-grained guidance for interpreting multimodal affective cues and may reward correct labels even when those cues are misinterpreted. To address this challenge, we introduce OmniOPSD, the first on-policy self-distillation framework for omni-modal affective reasoning. At its core, OmniOPSD transforms label-level supervision into evidence-rich, teacher-only rationales by leveraging reference affective judgments as semantic priors for offline multimodal evidence elicitation from frontier MLLMs. Through rationale-privileged self-distillation, it couples on-policy student rollouts with dense, evidence-informed token-level supervision from a local self-teacher, allowing the student to learn along its own trajectories rather than imitate fixed reasoning traces. This decoupling of offline evidence acquisition from local policy learning enables training without additional human annotation, frontier-model logits, or online frontier-model calls, while preserving student-only inference from the original task input. At both 3B and 7B scales, OmniOPSD outperforms supervised fine-tuning (SFT) and group relative policy optimization (GRPO) in mean macro-F1 on both in-domain and out-of-domain evaluations. Among the compared models, it ranks first by MER-UniBench mean across all three input-modality settings, reaching 84.13 with audio--video--text input, and achieves the highest MME-Emotion recognition score (Rec-S) of 42.4.

cs.CV

版本更新 · v3arXiv:2601.22947

Reconsidering Positional Supervision in Masked Diffusion Language Model Training

Mengyu Ye, Keito Kudo, Ryosuke Takahashi, Jun Suzuki

原文摘要

Masked diffusion language models (MDLMs) generate text by unmasking tokens in parallel and have recently emerged as alternatives to autoregressive language models. They can be viewed as parallel decoders trained with a position-wise cross-entropy (CE) loss, the same setup as non-autoregressive translation (NAT). In NAT, CE-trained parallel decoders have been argued to be sensitive to small positional shifts, since CE penalizes them harshly. We ask whether CE-trained MDLMs are similarly sensitive to such shifts under iterative decoding. To probe this, we apply a controlled intervention that introduces them during decoding. On LLaDA-8B-Instruct with Arena-Hard, displacing as little as 1% of generated tokens by one position substantially reduces win rates against the unintervened model, showing that MDLMs are sensitive to such small shifts under iterative parallel decoding. Motivated by this, we adapt connectionist temporal classification (CTC), an alignment-flexible objective known to mitigate it there, to MDLM supervised fine-tuning. By relaxing the strict position-wise match that CE imposes, CTC gives the loss room to absorb small positional shifts; concretely, we modified CTC objective to use a special <slack> token that absorbs positional uncertainty between target tokens and output positions, and a updated collapse map that preserves target surface forms. Across four open-ended generation benchmarks, the resulting model consistently improves over both the original model and a matched cross-entropy-trained baseline, with statistically significant gains on all four. These results identify training-side alignment flexibility as a useful design dimension for MDLM SFT, complementary to the inference-time approaches explored in prior work.

cs.CL · cs.LG

版本更新 · v3arXiv:2510.11726

scTREAT: Latent Diffusion for Single-Cell Drug-Response Prediction

Zhaokang Liang, Shuyang Zhuang, Xiaoran Jiao, Weian Mao, Hao Chen, Chunhua Shen

原文摘要

Generating drug-induced transcriptional responses requires a model to account for cellular background together with molecular identity and dose. We present scTREAT, the first diffusion framework to factorize state and drug guidance for single-cell drug-response prediction. A shared latent expression space connects baseline and response, while Guided Decomposable Attention (GD-Attn) injects a baseline representation and a dose-conditioned molecular embedding. Independent channel dropout exposes four conditioning states during training; inference combines three conditional predictions with separately specified state and drug coefficients, and dose modulates drug-guidance strength. On condition-level pseudobulk profiles derived from the Tahoe-100M single-cell atlas, scTREAT achieves the best value on 10 of 12 metrics for unseen cell line-drug combinations (UC) and 11 of 12 for unseen drugs (UD). Ablations examine condition injection, molecular representation, and guidance design. On UD, the relative gains over the second-best model reach 36.11% and 34.21% for DEG logFC-Spearman and logFC-Pearson. These results demonstrate the effectiveness of combining shared latent expression modeling with explicit state and drug conditioning for response prediction.

q-bio.QM · cs.LG

版本更新 · v3arXiv:2606.26578

EvoOptiGraph: Weakness-Driven Coevolution via Graph-Based Structural Generation for Optimization Modeling

Qingcan Kang, Mingyang Liu, Xiaojin Fu, Shixiong Kai, Tao Zhong, Mingxuan Yuan

原文摘要

Automating optimization modeling from natural language with large language models (LLMs) faces two key challenges. First, training corpora lack structural diversity. Second, data generation pipelines remain static and decoupled from model learning. To address these challenges, we propose EvoOptiGraph, a novel framework where data and model co-evolve, driven by model weaknesses. EvoOptiGraph represents each mixed-integer linear program (MILP) as an attributed bipartite graph and applies validity-preserving evolutionary operators to generate structurally diverse instances. The evolved graphs are converted into solver code and natural language via deterministic compilation and verified back-translation. Training proceeds in two stages: supervised fine-tuning (SFT) on an initial dataset, followed by reinforcement learning with verifiable rewards (RLVR), where graph-derived weakness signals guide the generation of new instances targeting the model's failures. This forms a closed loop that continuously updates the training distribution. Empirical results on six public datasets show that EvoOptiGraph significantly outperforms larger generalist models, agentic methods, and specialized baselines in accuracy, executability, and generalization. These results demonstrate that targeted data-model coevolution is an effective strategy for improving LLMs on optimization modeling tasks.

cs.AI

版本更新 · v3arXiv:2603.15834

Spectral Hierarchy of the Cosmic Web

Francisco-Shu Kitaura, Francesco Sinigaglia

原文摘要

We introduce a spectral hierarchy of cosmic-web classifications obtained by applying simple scale-weighting kernels to the density field before performing a standard eigenvalue-based web classification. This unifies and extends several widely used web definitions within a single framework: the familiar potential/tidal web (large-scale, nonlocal), a curvature-based web (more local, peak- and ridge-sensitive), and additional higher-derivative levels that progressively emphasize smaller-scale structure. Because the classification is built from second derivatives of the filtered field, successive hierarchy levels align naturally with operator families that appear in renormalised bias and effective descriptions of large-scale structure, providing an explicit bridge between cosmic-web environments and long- and short-range nonlocal bias ingredients. We quantify the information content of the hierarchy with a compact statistic: we map each cell to one of four ordered web types (void, sheet, filament, knot), construct a corresponding ``web contrast'' field, and measure its cross-correlation with halos from the AbacusSummit simulation suite on a coarse mesh with $ΔL\simeq 5.5\,h^{-1}\mathrm{Mpc}$. We find that the hierarchy retains significant tracer-relevant information from very large scales down to the mesh Nyquist limit, with the more local (curvature/higher-derivative) levels dominating toward nonlinear scales. This makes the spectral hierarchy a practical, interpretable conditioning basis for fast mock-galaxy production and field-level modelling, and a flexible tool for studying environment-dependent clustering and assembly bias.

astro-ph.CO · cs.CV

版本更新 · v2arXiv:2607.09587

CoDiMAD: Diffusion-Based Privileged Distillation for Communication-Free Multi-Robot Coordination

Jiyue Tao, Shunheng Xin, Tongsheng Shen, Dexin Zhao, Feitian Zhang

原文摘要

Decentralized multi-robot coordination under partial observability remains challenging, especially when agents act solely on local sensing without communication. Privileged policy distillation provides a promising approach by transferring knowledge from a globally informed oracle to sensor-constrained students. However, in multi-agent systems, the same local observation may correspond to multiple global configurations requiring qualitatively different cooperative actions, making the conditional action distribution inherently multi-modal. Standard deterministic distillation collapses these modes to their mean, often yielding invalid or hesitant actions. To address this issue, we propose CoDiMAD, a three-stage framework that trains a privileged oracle with MAPPO, constructs an offline dataset of local-observation-oracle-action pairs, and distills the oracle into decentralized students parameterized as conditional denoising diffusion probabilistic models. By approximating the conditional oracle-action distribution through the diffusion reverse process, CoDiMAD samples decisive actions from coherent coordination modes rather than averaging across them. Theoretical analysis characterizes the mode-averaging failure of deterministic distillation and the distributional recovery property of diffusion-based distillation. Experiments on three cooperative tasks show that CoDiMAD consistently outperforms direct local MARL and deterministic distillation baselines. The source code will be made publicly available upon acceptance.

cs.RO

版本更新 · v3arXiv:2304.14717

faulTPM: Exposing AMD fTPMs' Deepest Secrets

Hans Niklas Jacob, Christian Werling, Robert Buhren, Jean-Pierre Seifert

原文摘要

Trusted Platform Modules constitute an integral building block of modern security features. Moreover, as Windows 11 made a TPM 2.0 mandatory, they are subject to an ever-increasing academic challenge. While discrete TPMs - as found in higher-end systems - have been susceptible to attacks on their exposed communication interface, more common firmware TPMs (fTPMs) are immune to this attack vector as they do not communicate with the CPU via an exposed bus. In this paper, we analyze a new class of attacks against fTPMs: Attacking their Trusted Execution Environment can lead to a full TPM state compromise. We experimentally verify this attack by compromising the AMD Secure Processor, which constitutes the TEE for AMD's fTPMs. In contrast to previous dTPM sniffing attacks, this vulnerability exposes the complete internal TPM state of the fTPM. It allows us to extract any cryptographic material stored or sealed by the fTPM regardless of authentication mechanisms such as Platform Configuration Register validation or passphrases with anti-hammering protection. First, we demonstrate the impact of our findings by - to the best of our knowledge - enabling the first attack against Full Disk Encryption solutions backed by an fTPM. Furthermore, we lay out how any application relying solely on the security properties of the TPM - like Bitlocker's TPM- only protector - can be defeated by an attacker with 2-3 hours of physical access to the target device. Lastly, we analyze the impact of our attack on FDE solutions protected by a TPM and PIN strategy. While a naive implementation also leaves the disk completely unprotected, we find that BitLocker's FDE implementation withholds some protection depending on the complexity of the used PIN. Our results show that when an fTPM's internal state is compromised, a TPM and PIN strategy for FDE is less secure than TPM-less protection with a reasonable passphrase.

cs.CR

版本更新 · v3arXiv:2610.20424

BizAgentBench: Unveiling Hidden Challenges in Deploying Long-Horizon Agents for Real-World Business

Enhan Zhao, Yongqi Tong, Xuejun Yu, Hang Wang, Pan Wang, Yichao Wang, Di He, Jianshe Li 等 10 位作者

原文摘要

Existing long-horizon agent benchmarks often emphasize final task success while providing limited assessment of intermediate execution validity. Yet real-world business operations require agents to navigate extensive, potentially outdated or conflicting policies while satisfying critical constraints throughout execution. These requirements are particularly consequential in high-stakes domains such as payments, where intermediate violations can incur financial losses or regulatory consequences regardless of the final outcome. To bridge this evaluation gap, we introduce BizAgentBench, a multi-turn benchmark grounded in real-world procedures from a leading payments company. Its executable environment integrates 789 knowledge documents and 601 discoverable business tools, with consistent policies, states, and tool behavior. The benchmark comprises 50 tasks involving resource constraints, temporal dependencies, prior commitments, and conflicting or outdated guidance. Each task is validated through reference execution, while deterministic checks assess both goal attainment and execution validity without prescribing a single solution path. Experiments reveal substantial limitations even among frontier models: Claude Opus 5 and GPT-6 Sol achieve strict success rates of only 64.0% and 61.3%, respectively. For Qwen3.8-27B, goal attainment reaches 42.0%, but strict success drops to 32.7%, demonstrating how outcome-based evaluation can mask invalid execution. Beyond success rates, DeepSeek-V4.1-Flash averages 191.3 tool calls per run, approximately 70% more than reference executions, revealing substantial execution overhead. Overall, BizAgentBench establishes a testbed for evaluating and advancing long-horizon agents beyond task completion, toward reliable and constraint-compliant execution in complex business environments.

cs.CE