跳到正文
研踪PaperTrail

YOUR RESEARCH, IN SIGHT

让下一篇重要论文，不再错过。

覆盖计算机全学科，以 AI 为快捷入口。关注方向、作者和关键词，把发现变成持续的研究积累。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+1072新增论文
+845其中 AI 相关
+94已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-09 18:35 · 全站最近入库：2026-11-09 13:22各信源进度 →

统计口径与数据新鲜度

统计 2026-11-08 19:18 — 2026-11-09 19:18（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：2026-11-09 18:35最近发现新增版本：2026-11-09 13:22库内论文最新更新时间：2026-11-07 02:59

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。

18,172 已收录论文15,506 AI 相关4,527 有修订112 HF 社区发现

浏览论文库

研究日报 / 周报 →按信源浏览 →

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 5,189 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-09 18:33（北京时间）
新论文 · v1arXiv:2610.36759

DocSkill: Skill-Integrated Agentic Reasoning for Multilingual Multi-Hop Document Question Answering

Jing Huang, Zhiya Tan, Changtao Miao, Shutao Gong, Yang Wang, Joey Tianyi Zhou, Jianshu Li, Ying Yan

原文摘要

Multilingual multi-hop document question answering (MMHop-DocQA) requires locating and integrating evidence across complex layouts, document pages, and diverse languages. Existing approaches include single-pass processing and adaptive tool use, but learning to coordinate evidence acquisition, cross-page integration, and answer verification remains challenging. To address this challenge, we present DocSkill, an agentic framework with a learned skill-orchestration policy for adaptive evidence acquisition and reasoning. The policy selects, revisits, or skips skills according to the evolving evidence state. To construct supervision for skill orchestration, bounded Monte Carlo Tree Search (MCTS) explores alternative skill trajectories offline. These trajectories provide demonstrations for supervised fine-tuning (SFT) and preference pairs for Direct Preference Optimization (DPO). We further introduce MMHop-DocBench, a multilingual multi-hop document QA dataset with training and test splits, containing 60K QA pairs across 10 languages and eight document domains. Its HopGen pipeline builds hierarchical page-element graphs to derive traceable cross-page evidence paths and generate grounded QA pairs with page- and element-level evidence annotations. Experiments show that DocSkill with Qwen3-VL-8B achieves 41.0 accuracy and 49.2 F1 on MMHop-DocBench, exceeding the strongest evaluated baseline by 5.5 and 6.3 points, respectively, and achieving competitive performance on MMLongBench-Doc and DocBench.

cs.CL · cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2606.04874

Agent Planning Benchmark: A Diagnostic Framework for Planning Capabilities in LLM Agents

Haoyu Sun, Wenxuan Wang, Mingyang Song, Jujie He, Weinan Zhang, Yang Liu, Yang Yang, Yu Cheng

原文摘要

Planning is central to LLM agents: before acting, an agent must decompose goals, select tools, reason over constraints, and decide when a task is infeasible. Yet existing agent evaluations often report only end-to-end success, making it difficult to determine whether failures stem from planning or execution. We introduce Agent Planning Benchmark (APB), a planning-specific diagnostic benchmark with 4,209 multimodal cases across 22 domains and five settings, covering holistic planning, feedback-conditioned step-wise planning, and robustness under extraneous tools, broken tools, and unsolvable tasks. Across 12 MLLMs, APB reveals systematic weaknesses in long-horizon planning, tool-noise robustness, calibrated refusal, and inference-time refinement. We further validate APB on 200 ToolSandbox tasks and 200 $τ^2$-bench tasks, where APB-guided refinement consistently improves plan correctness, plan grade, and downstream execution metrics across three representative models. APB thus serves as an upstream diagnostic complement to execution benchmarks. The APB benchmark and code are available in \href{https://github.com/Mikivishy/AgentPlanningBenchmark}{this URL}.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36734

CAMR: Compatibility-Aware Memory Replay for Continual Learning in Large Language Models

Jingpu Yang, Tianhao Li, Fengxian Ji, Yufeng Wang, Shuxin Zheng, Xiaoqing Zhang

原文摘要

Continual learning requires large language models to acquire new knowledge and capabilities while retaining what they have learned. Fine-tuning and replay support acquisition and retention, but deciding whether incoming examples need supervision and which historical examples to revisit remains a challenge. We propose Compatibility-Aware Memory Replay (CAMR), an adaptive framework that jointly controls current-example supervision and historical replay. CAMR combines compatibility checks with a learned memory gate to guide an online policy over three actions: learning incoming examples with general replay, jointly reinforcing compatible related examples, and skipping direct supervision of verified mastered duplicates. The policy learns from acquisition, retention, and training-cost feedback under a bounded per-arrival replay budget. We evaluate CAMR with Qwen3-8B-Base on continual factual updating and cross-task and cross-domain learning. On WikiBigEdit, CAMR consistently leads evaluated methods in Original Match from 3K to 100K facts, reaching 93.66% after just 3K facts and surpassing Full-history Replay (LoRA) after 100K facts. It also achieves the highest final scores on all four selected TRACE tasks and maintains or improves earlier-task performance relative to post-acquisition levels in a sequence spanning financial question answering, mathematical reasoning, code generation, and summarization.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2602.15896

Every Little Helps: Building Knowledge Graph Foundation Model with Fine-grained Transferable Multi-modal Tokens

Yichi Zhang, Zhuo Chen, Lingbing Guo, Wen Zhang, Huajun Chen

原文摘要

Multi-modal knowledge graph reasoning (MMKGR) aims to predict the missing links by exploiting both graph structure information and multi-modal entity contents. Most existing works are designed for a transductive setting, which learns dataset-specific embeddings and struggles to generalize to new KGs. Recent knowledge graph foundation models (KGFMs) improve cross-KG transfer, but they mainly exploit structural patterns and ignore rich multi-modal signals. We address these gaps by proposing a token-based foundation model (TOFU) for MMKGR, which exhibits strong generalization across different MMKGs. TOFU discretizes structural, visual, and textual information into modality-specific tokens. TOFU then employs a hierarchical fusion architecture with mixture-of-message mechanisms, aiming to process these tokens and obtain transferable features for MMKGR. Experimental results on 17 transductive, inductive, and fully-inductive MMKGs show that TOFU generally outperforms strong KGFM and MMKGR baselines, delivering strong performance on unseen MMKGs.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.37719

What a Weight Update Buys: Leverage Across Training Stages

Shikhar Shiromani

原文摘要

Post-training moves a language model's weights by about a thousandth of their norm and changes its behavior sharply; pretraining moves them by a third and buys a slow drift. We introduce leverage, an exact measure of what a unit of weight change buys: the Fisher quadratic form of an update, computed with one forward-mode pass per direction, divided by the same form for random directions of the same size at the same checkpoint. Leverage orders released Olmo-3, Tülu-3, Zephyr, and OLMo-2 checkpoints over four orders of magnitude: pretraining windows sit at the random level on prose, SFT five times above it, RL-Zero and RLVR at 400-1,500, and DPO at 400-23,000. With this instrument we find that pretraining accumulates into noise: windows of 30k to 1M steps are no more consequential on prose than random directions of their size, as a noisy quadratic model of SGD predicts. We find that training precision changes what an update looks like: training bf16 parameters without an fp32 master copy freezes 36-99% of coordinates, lowers leverage up to sevenfold, and creates an alignment between update magnitudes and the diagonal Fisher that fp32 training does not show. We derive and verify a law for reuse: the sign-flip null predicts the relative damage of random drop-and-rescale (DARE) at keep rate $p$ as $(1-p)/(p\,ρ_{\mathrm{sign}})$, within a factor of two of measurement wherever DARE succeeds. Two controls make these readings reliable: every controlled SFT and DPO trajectory with fp32 master weights loses leverage as it moves, and two re-runs of a released DPO stage from its anchor with its published recipe land within 1% of the released displacement and 4% of its leverage, where a plain recipe reads five to six times lower; leaving one component out traces the level to the length-normalized loss and its batch size of 128 together.

cs.LG · cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2505.17399

FullFront: Benchmarking MLLMs Across the Full Front-End Engineering Workflow

Haoyu Sun, Huichen Will Wang, Jiawei Gu, Linjie Li, Yu Cheng

原文摘要

Front-end engineering involves a complex workflow where engineers conceptualize designs, translate them into code, and iteratively refine the implementation. While recent benchmarks primarily focus on converting visual designs to code, we present FullFront, a benchmark designed to evaluate Multimodal Large Language Models (MLLMs) \textbf{across the full front-end development pipeline}. FullFront assesses three fundamental tasks that map directly to the front-end engineering pipeline: Webpage Design (conceptualization phase), Webpage Perception QA (comprehension of visual organization and elements), and Webpage Code Generation (implementation phase). Unlike existing benchmarks that use either scraped websites with bloated code or oversimplified LLM-generated HTML, FullFront employs a novel, two-stage process to transform real-world webpages into clean, standardized HTML while maintaining diverse visual designs and avoiding copyright issues. Extensive testing of state-of-the-art MLLMs reveals significant limitations in page perception, code generation (particularly for image handling and layout), and interaction implementation. Our results quantitatively demonstrate performance disparities across models and tasks, and highlight a substantial gap between current MLLM capabilities and human expert performance in front-end engineering. The FullFront benchmark and code are available in https://github.com/Mikivishy/FullFront.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36717

A Harness Reducing over-Defensive Writing in Agents by Isolating Objections

Xiaochen Chang

原文摘要

When expanding a report, writing with agents may introduce a stronger claim than the article makes and then disclaim it. This is called over-defensive writing. We assume that this is because the agent is writing for responding to an adversarial AI reviewer, instead of a human reader. The defensive context may also influnce the consequent when we simply ban defensive writing in the output through prompts. In order to reduce over-defensive writing, we produce a harness that separates the reviewer, filter, and writer. The writer agent receives only the problems choosen by the filter, which is an independent context. We evaluate this harness on 25 business-project drafts expanded into formal reports, the result shows that the defensive propositions decreased from 59 to 2, which is a 96.61\% reduction, no unsupported factual additions were identified while the source-claim retention rises from 80.35% to 93.12%.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2606.05158

Streaming Communication in Multi-Agent Reasoning

Zhen Yang, Xiaogang Xu, Wen Wang, Cong Chen, Xander Xu, Ying-Cong Chen

原文摘要

Multi-agent reasoning systems adopt a generate-then-transfer paradigm that forces end-to-end latency to scale linearly with pipeline depth. We introduce StreamMA, a multi-agent reasoning system that streams each reasoning step to downstream agents as soon as it is generated, pipelining adjacent agents and thus reducing latency. Surprisingly, this pipelining also improves effectiveness: because multi-step reasoning quality is non-uniform and early steps are more reliable than later ones, working with these reliable early steps instead of the full chain prevents error-prone late steps from misleading downstream agents. We formalize both advantages with the first closed-form joint analysis of stream, serial, and single protocols, deriving the effectiveness ordering, speedup upper bound, and cost ratio. Across eight reasoning benchmarks spanning mathematics, science, and code, three frontier LLMs (Claude Opus 4.6, GPT-5.4, and GLM-5.2), and three topologies (Chain, Tree, Graph), StreamMA outperforms both baselines (avg.+7.3pp, max +22.4pp on HMMT2026; Claude Opus 4.6-High). Beyond these contributions, we discover a step-level scaling law: increasing per-agent steps consistently improves both effectiveness and efficiency, a new scaling dimension orthogonal to and composable with agent-count scaling.

cs.CL · cs.AI · cs.MA

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36709

Is Token Confidence Just Chain Length? A Signal-Processing Audit of Reasoning Traces

Pragna Mallappa

原文摘要

Reasoning language models answer math problems by writing long chains of thought, and sampling many chains per problem and taking a majority vote raises accuracy at a large cost in generated text. A popular way to cut that cost scores each chain by the model's confidence in each token it writes, where a token is a word or part of a word. The method averages this confidence over a sliding window of 2048 tokens and keeps or stops chains based on the lowest average. We treat the confidence, recorded token by token, as a signal in the sense of audio or radio, and examine the method with standard signal-processing tools such as spectral analysis and statistical tests for detecting a sudden change. We sampled 16 chains for each of 500 problems from two math benchmarks with a small open reasoning model. The windowed-minimum score turns out to measure mostly how long a chain is. Ranking chains by length alone does as well as or better than the confidence score, and once length is accounted for the score carries no information about correctness. Signal processing explains why: the minimum of a longer stretch of a slowly drifting signal is lower by chance alone, and the confidence signal drifts slowly enough that this effect dominates. The spectrum does hold a small signal that length does not explain, since incorrect chains carry more rapid token-to-token fluctuation, but it is too weak to improve the vote. For stopping chains early, neither the sliding average nor a classical change detector beats the simpler choice of sampling fewer chains.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2510.20256

Uncertainty-Guided Multimodal Consensus for Multimodal Sentiment Analysis

Guowei Zhong, Junjie Li, Huaiyu Zhu, Ruohong Huan, Yun Pan

原文摘要

Multimodal sentiment analysis (MSA) has made considerable progress in recent years. Existing MSA methods mainly learn shared multimodal representations or train unimodal predictors with a common multimodal label. However, assigning the shared label to every modality can obscure modality-specific sentiment, while direct fusion may overemphasize unreliable predictions when modalities disagree. To address these issues, we propose Uncertainty-Guided Multimodal Consensus (UMC), which first recovers continuous unimodal pseudo-labels and then forms a reliability-aware consensus across modalities. Our method consists of two key components: i) Continuous Prototype-Guided Label Recovery (CPLR), which combines heteroscedastic predictions with ordered sentiment prototypes and uses confidence and label conflict to refine continuous unimodal pseudo-labels, and ii) Reliability and Precision Consensus (RPC), which estimates sample-dependent modality reliability from predictive statistics and cross-modal disagreement and combines it with predictive precision through a closed-form consensus solver. Extensive experiments conducted on four publicly available MSA datasets demonstrate the effectiveness and superiority of our proposed model.

cs.CV · cs.CL · cs.LG · cs.MM

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.19350

Pruning via Causal Attribution Preserves Reasoning Performance in Large Language Models

Amogh Sheth, Andrew Lin, Biruk Assefa, Yi Wen Huang, Yuhao Ge

原文摘要

Large language models (LLMs) excel at multi-step reasoning but incur substantial inference cost. We introduce Causal Attribution Pruning (CAP), a training-free method that identifies critical attention heads by measuring their causal impact on reasoning tasks and uses these head-level scores to guide fine-grained weight pruning. For each attention head, CAP estimates the expected performance degradation when the head is masked during forward passes on a small calibration set of reasoning problems. These causal scores are then converted into weight-level importance values for the corresponding projection matrices. Unlike magnitude-only or activation-based criteria, CAP's interventional measurement directly captures each head's functional contribution, yielding relative accuracy gains of up to 61% over Wanda on ARC-Challenge at 20% sparsity. We evaluate CAP on GSM8K, StrategyQA, and ARC-Challenge using Llama-3-8B-Instruct and Mistral-7B-Instruct at 10%, 20%, and 50% sparsity. At moderate sparsity (10-20%), CAP improves over Wanda in most model-benchmark configurations. with especially large gains on ARC-Challenge for Llama-3. Our results suggest that attention-head-level causal attribution can better preserve reasoning performance on downstream benchmarks than correlational pruning criteria at equivalent sparsity, while remaining limited by coarse MLP attribution at 50% sparsity.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2610.07853

Lost in the bf16 Cast: Exporting Ternary Language Models Can Revert Most Low-Learning-Rate Code Changes

Avichal Sahai, Nishant Raj, Animesh Srivastava

原文摘要

Ternary language models such as BitNet b1.58, Falcon-E and BitCPM are fine-tuned through higher-precision latents and deployed as ternary codes by an export that, in the labs' documented pipelines, first casts latents to bf16. We audit releases from three labs. In released checkpoints, fp32 quantization of the shipped latents disagrees with deployed codes on 0.83-1.77% of codes (Falcon-E, BitCPM) and 1.530% (2B-4T); most Falcon-E and BitCPM disagreements are bf16 products exactly on the threshold, which ties-to-even maps to zero, and the unmodified onebitllms exporter reproduces all four Falcon-E releases byte for byte. At fine-tuned endpoints with ULP-matched rates, the documented export lowers greedy GSM8K strict accuracy from 58.79% to 0.78% (Falcon-E-1B-Base) and 36.13% to 0.39% (BitCPM-CANN-0.5B), and a bf16 save and reload lowers 2B-4T's by 27.54 points while its last-number accuracy rises at that endpoint. The strict decrease recurs at each of three new seeds in all three models and has the same direction on 607 reserve test rows, but it is recipe-dependent: at Falcon-E's online-loss-selected rate the export was non-inferior, and after a longer recipe no difference was detected. Two remedies, writing the trained codes directly or adjusting bf16 inputs so unchanged tools emit them, met an inherited 4-point strict-accuracy non-inferiority criterion in all three models, also for the pooled mean over three fixed new trajectories per model and, separately, on the reserve rows. In a frozen study of bf16 latents in fine-tuning (one task, one trajectory per arm), native bf16 writes left 2B-4T's endpoint codes unchanged and changed one Falcon-E code; emulated stochastic-rounding and compensated writers changed codes and were non-inferior to fp32 training at that margin. In two model families, randomized threshold-distance interventions support distance-dependent selection of changed codes.

cs.LG · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36679

Specialization, Diversity, and Scaling in Child-Scale Language Models

Steven Y. Feng, Michael C. Frank

原文摘要

Modern language models are trained on far more language than children experience, but we know much less about learning from small, naturalistic datasets. We train small GPT-2 models on transcripts of speech with children from dense home recordings (BabyView) and on synthetic child-directed dialogue (TinyDialogues), varying the amount, source, and diversity of training data. Models trained on one child's language environment predict that environment substantially better than other children's environments. Much of this specialization is lexical and distributional: a simple 5-gram model reproduces the cross-child pattern (r=.91). Training on data from multiple children clearly improves prediction of pooled held-out language relative to training on a single child's input without a similarly clear improvement on grammatical benchmarks. Across data sizes, naturalistic and synthetic training inputs also produce different grammatical and lexical-semantic patterns, with important differences between grammatical phenomena. Together, these results show that small-data language learning depends not only on how much language a model sees, but on which environments it comes from and how well that input matches what the model is asked to predict.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36668

A Quantitative Approach to Asch Conformity in Frontier Large Language Models

Senthil Rajasekaran, Stylianos Loukas Vasileiou

原文摘要

The Asch conformity experiment has recently gained considerable attention as a way of measuring conformity in large language models (LLMs). While previous studies have focused on demonstrating conformity in LLMs, our work adds a further quantitative layer to this approach. Instead of sampling a frontier LLM on a suite of questions, our approach samples a single question repeatedly, leveraging the probabilistic nature of LLMs to produce a distribution of answers. In doing so, we can compare the distributions of answers before and after the model is presented with the manufactured consensus. Our experiments confirm that frontier LLMs do conform, reinforcing prior literature, and, through our distribution-to-distribution comparisons, provide additional quantitative data on the conditions that lead to conformity and the degree to which it occurs.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36667

Japanese Discourse Structure Analysis Based on Segmented Discourse Representation Theory Using LLMs

Satomu Tomisaki, Yosuke Mikami, Amane Watahiki, Hitomi Yanaka

原文摘要

Discourse parsing based on Segmented Discourse Representation Theory (SDRT) is crucial for capturing the structural dynamics of dialogues. However, most prior research focuses on English, and its effectiveness for Japanese remains unclear owing to the lack of Japanese SDRT corpora. To address this issue, we construct Japanese SDRT corpora by translating existing English SDRT corpora using an LLM. Through a comparative evaluation of prompt-based and fine-tuned LLMs, we show that the models achieve comparable performance between English and Japanese, suggesting the cross-lingual applicability of SDRT-based discourse parsing. In order to investigate why the English and Japanese parsers achieve similar performance, we conduct ablation analyses on discourse markers and contextual inputs. Our results reveal that most of the Japanese discourse markers we examined have little impact, and the models' predictions depend heavily on contextual discourse relations rather than on the content of the context clauses. This structural reliance might explain the comparable performance despite the substantial typological differences between English and Japanese.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.18420

Agaram 2.5B: The First Tamil-First Language Model Trained from Scratch with Tool Calling. A Playbook for Under-Represented Languages

Ramchand Kumaresan, Madhusudhanan G, Kaushik Kezendirane

原文摘要

We present Agaram 2.5B, a Tamil-first language model trained from scratch that converses in Tamil script, romanized Tanglish and English, and calls tools. Our first model, Agaram 1.5B, guided the new base, tokenizer and fine-tuning recipe: we pretrained a new 2.53B-parameter base on 50 billion tokens, 58% of them Tamil, with a tokenizer that needs 1.55 tokens per Tamil word where Qwen2.5's needs 9.44. A four-day public preview of our first 2.5B chat model showed where it fell short with real users, and we rebuilt supervised fine-tuning around what we saw, on 222,662 records. The released model scores above Sarvam-1 in all six translation directions we measured (FLORES chrF 56.9 into Tamil and 55.4 into English), chooses the expected tool on all 25 prompts of a tool set that was used during checkpoint selection and whose schemas appear in training, with exact arguments on 16, and was preferred over the preview model by an LLM judge on 100 sealed human-written prompts (net preference +30 points, 95% interval +14.5 to +45.0). On FLORES it also scores above Google's Gemma 4 E4B-it and our fine-tuned variant of it in both directions, with 2.53B parameters against E4B's 4.5B effective parameters. Factual knowledge and reliable refusal of harmful requests remain open problems, which we set out as future work. The base and chat weights and GGUF files are released under Apache-2.0, and this report gives a full account of the data, training and evaluation as a playbook for building language models for under-represented languages.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2608.29249

Validating FKG.in: Soundness Assessment in LLM-Augmented Indian Food Knowledge

Saransh Kumar Gupta, Armaan Shah, Lipika Dey, Partha Pratim Das, Ramesh Jain

原文摘要

The online culinary ecosystem is increasingly populated by recipe content generated, modified, or summarized by LLMs. Although often convincing, such outputs may contain hallucinated ingredients, misrepresented quantities, or culturally implausible combinations, limiting their suitability for downstream applications and knowledge graph construction. In this paper, we present a semi-automated soundness assessment workflow for validating structured recipe data extracted from informal culinary sources and augmented by LLMs. Developed as part of FKG(.in), a knowledge graph of Indian food, the pipeline identifies and addresses common failure modes, including structural inconsistencies, semantic and logical incoherence, and deviations from the source text, through a multi-stage process combining formal grammars, vocabulary-based checks, statistical heuristics, Set Transformer-based coherence modeling, and retrieval-based verification. Although evaluated on Indian recipes, the proposed methods are applicable to broader multilingual and multicultural culinary domains. We provide a practical, auditable, and application-agnostic approach to improving the reliability of machine-readable food knowledge infrastructures in the era of LLM-generated content.

cs.AI · cs.CL · cs.IR · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36617

Balanced State Subspaces for Mamba-2 State Compression

Shang Tingyu, Cao Wenfei

原文摘要

Mamba-2 compresses all context into a fixed-size recurrent state that dominates decoding memory and bandwidth. Existing compressors quantize state values or prune state coordinates, scoring each channel independently in the native basis. We examine trained Mamba-2 states and find that energy is almost rank-one, but the output does not read the energetic directions: energy concentrates in a bias-like direction accumulated by the recurrence, while the content the output reads lives in a low-dimensional subspace that is not aligned with the state's coordinates. Even output-aware scores are computed coordinate by coordinate, so they miss the directions that determine the input--output map. Compression should therefore target the write--read subspace rather than individual coordinates. We propose \method{}, a post-training transform that instantiates balanced truncation with full controllability and observability Gramians and applies the rotation exactly via Mamba-2's scalar-identity transition. Because the useful rank varies widely across layers, we further introduce HARP, which allocates ranks per layer from the Hankel spectrum and protects retrieval-sensitive layers, improving both perplexity and retrieval over uniform budgets. Calibration takes a single forward pass, an order of magnitude faster than GHOST's released protocol, with lower peak memory. With HARP, \method{} halves the state dimension at an average WikiText-2 perplexity increase of $0.3$ points ($1.8\%$) across 130M--2.7B, versus $1.6$ ($9.3\%$) for GHOST, and the gap widens at 25\% retention.

cs.LG · cs.AI · cs.CL · stat.AP · stat.ME

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.07755

Rethinking State Tracking in Recurrent Models Through Error Control Dynamics

Jiwan Chung, Heechan Choi, Seon Joo Kim

原文摘要

The theory of state tracking in recurrent architectures has predominantly focused on expressive capacity: whether a fixed architecture can theoretically realize a set of symbolic transition rules. We argue that equally important is error control, the dynamics governing hidden-state drift along the directions that distinguish symbolic states. We prove that affine recurrent networks, a class of models encompassing State-Space Models and Linear Attention, cannot correct errors along state-separating subspaces once they preserve state representations. Consequently, practical affine trackers do not learn robust state tracking; rather, they learn finite-horizon solutions governed by accumulated state-relevant error. We characterize the mechanics of this failure, showing that tracking remains readable only while the accumulating within-class spread remains small relative to the between-class separation. We demonstrate empirically on group state-tracking tasks that this breakdown is predictable: tracking collapses when the distinguishability ratio crosses the readability threshold. Across trained models, the point of this crossing predicts the horizon at which downstream accuracy fails. These results establish that robust state tracking is determined not only by an architecture's theoretical expressivity but crucially by its error control.

cs.LG · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36582

Structured Maps for Landmark-Based Route Instructions with LLMs

Mifu Matsumoto, Yuki Nakanishi, Tomoya Hashiguchi, Hiroaki Ohshima

原文摘要

To generate landmark-based route instructions, models must understand both map geometry and human landmark preferences. Although large language models (LLMs) show great abilities in language tasks, the best way to structure map data for these models is not yet established. This paper proposes a structured text representation for landmark selection tasks in route navigation, utilizing SVG format to encode spatial features. The structured text maintains a geometric connection with rendered map images, allowing language models to directly access spatial and semantic data. Using a new dataset created from human-written route descriptions and OpenStreetMap data, we evaluate models on two tasks: landmark selection at turns and around destinations. Experiments with five LLMs show that the proposed structured text input performs as well as or better than image-based PNG inputs. In addition, task-specific fine-tuning enables smaller local LLMs to effectively leverage map layouts to achieve high selection accuracy. These results demonstrate that structured text is an efficient and lightweight interface for spatial reasoning in human-oriented navigation.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2512.03620

SELF: A Robust Singular Value and Eigenvalue Approach for LLM Fingerprinting

Hanxiu Zhang, Qiang Xu, Yue Zheng

原文摘要

Safeguarding the Intellectual Property (IP) of Large Language Models (LLMs) against model theft represents a critical challenge in contemporary AI research. Although fingerprinting is essential for establishing forensic evidence for model derivation, existing methods are susceptible to false claim attacks, weight manipulations, or high false positive rates, undermining their utility. To overcome these limitations, we propose SELF, \rev{a novel intrinsic weight-based fingerprinting scheme tailored for robust IP forensics via adversarial model lineage verification}. SELF eliminates dependency on input and inherently resists false claims, achieving robust \rev{provenance detection} through unique, scalable and transformation-invariant fingerprint extraction via singular value and eigenvalue decomposition of LLM attention weights. Experimental results demonstrate that SELF maintains high IP infringement detection accuracy while significantly outperforming state-of-the-art baselines in reducing false positives and showing strong robustness against various downstream modifications, including quantization, pruning, and fine-tuning attacks. Our code is available at https://github.com/HanxiuZhang/SELF_v4.

cs.CR · cs.AI · cs.CL · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36569

Decoupling Consistency and Conflict across MLLM Layers for Fake News Detection

Kai Qian, Dongjie Xu, Weijie Shi, Jiaqi Wang, Haolun Ma, Jia Zhu, Jiajie Xu

原文摘要

The trustworthiness of Web content faces a growing threat from multimodal fake news. Real news usually shows aligned image-text semantics, while fake news often contains a few conflicts at varying levels, such as mismatches in objects, scenes, events, and claims. However, existing methods usually treat consistency and conflict in a coarse and static way, limiting their ability to capture sparse conflicts and adapt to varying conflict levels. Inspired by the distinct knowledge encoded across layers in large language models, we propose Multi-Level Consistency and Conflict Decoupling (MLCCD), which captures conflicts at different levels from different layers of multimodal large language models (MLLMs). Specifically, we first use the MLLM's final decision representation to select visual and textual tokens relevant to the task, preventing irrelevant content from diluting sparse conflicts. Then, we preserve tokens whose representations change sharply between adjacent layers, capturing the knowledge each layer adds. Finally, we decouple this knowledge into consistency and conflict evidence and aggregate each across layers with its own router, adapting to varying conflict levels. Experiments on three benchmarks show that MLCCD outperforms existing methods.

cs.CV · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36557

QESChunker: A Single Objective Unifies Overlapping and Non-Overlapping Chunking for RAG

Yifan Zhao, Qianyang Zhao, Xinyu Guan, kai wei, Yuming Deng

原文摘要

Chunking is a fundamental preprocessing step in retrieval-augmented generation (RAG), yet most existing chunkers rely primarily on heuristics. We introduce the Question--Evidence-Supervised Chunker (QESChunker), to our knowledge the first method for constructing static chunk indices from question--evidence supervision. QESChunker constructs document-specific chunk indices by optimizing over human-annotated or LLM-generated question--evidence pairs, without access to held-out test queries. We formulate non-overlapping and overlapping index construction through a unified question--evidence utility-maximization objective, turning chunking into a supervised combinatorial optimization problem. Under the non-overlap constraint, the objective decomposes over chunks and can be optimized exactly in polynomial time by dynamic programming. When overlap is allowed, the objective becomes monotone submodular; exact optimization is NP-hard, but our greedy achieves a $(1-1/\sqrt{e})$-approximation. Across two self-constructed and three public datasets, both variants outperform existing chunking methods in complete-evidence recall in nearly all evaluated settings, under the same overlap regime and a fixed hybrid retriever. These results recast chunking as supervised, budget-aware index construction for evidence retrieval.

cs.IR · cs.CL · math.OC

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2607.28862

TextCloak: Safeguarding Data from Unauthorized LLM Fine-Tuning via RL-Driven Unlearnable Examples

Chengshuai Zhao, Pingchuan Ma, Dawei Li, Bohan Jiang, Zhiyuan Yu, Zhen Tan, Huan Liu

原文摘要

The development of Large Language Models (LLMs) has led to significant advances across a wide range of language tasks, while simultaneously raising growing concerns about unauthorized online data exploitation and privacy leakage. Unlearnable examples (UEs) emerge as a promising defense by injecting crafted perturbations into data such that models trained on them exhibit degraded utility. However, existing methods for text protection are primarily designed for classification tasks and discriminative language models, which limits their broad applicability (e.g., instruction following or reasoning) in LLMs. In this work, we propose **TextCloak**, an RL-driven data synthesis framework that safeguards textual data against unauthorized LLM fine-tuning. TextCloak employs a *UE generative policy* that transforms clean text into unlearnable examples while preserving semantic fidelity and linguistic naturalness. To optimize the policy, we introduce **GRPO-UE**, which simulates the attacker's LLM fine-tuning and rewards downstream degradation via group-relative policy optimization. This overall formulates a bi-level optimization loop, allowing the generator to discover effective protective patterns. Comprehensive experiments on six publicly available datasets and nine state-of-the-art LLMs demonstrate that TextCloak consistently impairs unauthorized LLM fine-tuning while maintaining text utility for legitimate use. Further analyses confirm its transferability and robustness across model architectures, training configurations, and adaptive attacks, highlighting its broad applicability as a practical defense for online content.

cs.CL · cs.AI · cs.CR · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.34717

ReMCTS: Reflection-Enhanced Monte Carlo Tree Search for Code Generation

Huifei Wang, Xinying Huang, Yiheng Sun, Yifan Yuan

原文摘要

Open-weight large language models (LLMs) can generate function-level programs from natural-language prompts, but plausible candidates still fail on hidden semantics and repeat mistakes across repair attempts. We present ReMCTS, an execution-grounded, memory-augmented, LLM-guided MCTS-style search framework. It organizes program candidates as tree states, retains branch-local debugging context, retrieves failure experience across branches, and distinguishes failed checks from unavailable evidence. On HumanEval and MBPP-Sanitized, visible-test ReMCTS improves over direct generation in 8 of 10 model-dataset pairs under held-out evaluation, whereas proxy-only search is less stable. Controlled tree-search, sampling, repair, and memory ablations characterize the source and limits of these gains. A 30-task HumanEval-X C++ pilot further demonstrates compatibility with compiler-backed execution, but does not constitute a broad multilingual evaluation.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2604.09711

Decoupling and Regulating Modality Heads for Fake News Detection under Missing Modality

Kai Qian, Dongjie Xu, Weijie Shi, Yezhao Xu, Jiaqi Wang, Hanghui Guo, Mengze Li, Hao Chen 等 10 位作者

原文摘要

Multimodal news often loses its image or text as it spreads across Web platforms. Existing methods treat this as information loss, assuming that the original veracity label remains valid. This assumption fails for fake news reusing authentic material, whose remainder may be genuine. Detectors should therefore decouple modality cues and adapt their contributions to modality availability. However, coarse news labels provide little guidance for modality decoupling, while training on complete inputs can induce modality bias. To address these challenges, we propose Head-wise Modality Decoupling and Regulation (HMDR) for multimodal large language models. First, Modality Head Decoupling decouples modality representations through soft head grouping by attention preference. Second, Modality Head Regulation regulates head contributions through compensation for imbalanced modality attention. Finally, Modality Invariance Learning promotes invariance by aligning representations of each modality under replacement of the other. Experiments on two benchmarks show that HMDR improves detection under missing modalities while maintaining performance on complete inputs.

cs.CV · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36538

Agentic Data Sampling Framework for Multilingual ASR via Linguistic-Aware Error-Driven Optimization

Zhiheng Qian, Junhao Du, Jing Peng, Hanqi Li, Kunyang Peng, Feng Xue, Lu Chen, Shuai Wang 等 9 位作者

原文摘要

Multilingual automatic speech recognition (ASR) has achieved rapid progress with large-scale pretrained models and multilingual speech corpora. Previous studies have proved the pivotal role of data sampling in multilingual ASR performance. However, existing sampling strategies largely rely on time-consuming manual, heuristic-based designs. To address this, we propose a linguistic-aware agentic data sampling framework that introduces a data sampling LLM agent into the multilingual ASR training loop. The agent is driven by a linguistically grounded rubric that analyzes ASR errors from phonological, morphological, and syntactic perspectives, and uses the resulting feedback together with training statistics, budget constraints, and algorithm profiles to select domain-level, sample-level, and token-level sampling strategies. This design turns evaluation results into adaptive data-selection decisions rather than treating WER/CER as final endpoints only. We conduct experiments on twelve Common Voice languages using Whisper-medium as the ASR backbone. The results show that our method achieves performance better than full-data training while using less training data and lower computational resources, and substantially accelerates training convergence compared to conventional sampling strategies.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36533

RoboAuto: Autonomous Robot Program Search through Multi-Level Physical Understanding

Buzhao Liu, Jianing Ye, Xiu Yuan, Yangtong Feng, Chengsong Huang, Tyler Dinh, Yevgeniy Vorobeychik, Ning Zhang 等 9 位作者

原文摘要

Recently, LLM-based autonomous agents have shown strong capabilities in iteratively generating, evaluating, and refining executable solutions. In robotic manipulation, Code-as-Policy lets these agents search over modular and editable robot programs. However, effective program search requires physical understanding: agents must understand how objects can be manipulated, how scene geometry constrains actions, and why interactions succeed or fail. Existing methods rely mainly on execution feedback for program refinement, often leaving these physical questions unresolved. We introduce RoboAuto, a framework for robot program search that explicitly organizes physical understanding at the object, scene, and interaction levels through three components. First, a hierarchical knowledge base combines executable skills with object-centric knowledge of affordances, spatial relations, and action effects, learned from embodied data and online task solving. Second, scene grounding connects this knowledge to the current environment by identifying relevant objects and resolving action-critical quantities. Third, multi-stage planning with self-designed experiments checks whether each stage produces its expected physical effects and actively tests competing hypotheses to resolve physical uncertainties. RoboAuto achieves superior performance across five benchmark families, attaining an 89.1% success rate on short-horizon LIBERO-PRO tasks and more than twice the mean success rate of ASPIRE on LIBERO-PRO Long and LIBERO-Plus. It also outperforms ASPIRE and $π_{0.5}$ on three challenging long-horizon real-robot tasks requiring precise manipulation. We further show that the knowledge base transfers across simulators and embodiments, and that program search performance scales with its size. Project page: https://roboauto-ai.github.io/

cs.RO · cs.AI · cs.CL · cs.CV · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2610.03827

The Score Is Not the Structure: Brain Alignment and Cross-Lingual Transfer

Saman Rahbar

原文摘要

Similarity scores are often offered as evidence that a model shares structure with the brain or across languages. We ask what such a score reads when that structure is removed, or when the instrument measuring it does not work, in two settings. Across seventeen languages, a grammaticality probe transfers worse between more distant languages (r = -0.66). But the probe's own accuracy falls along the same axis and is at chance in four languages, four of the five most distant. Dropping them halves the explained variance, and because it also narrows the range of distances, the design cannot say how much of the gradient is the instrument. Counting the 272 language pairs as independent gives p = 0.0006 for a steering effect that is null when the seventeen languages are the unit (p = 0.155). In brain alignment, a training objective raises a language model's similarity to fMRI responses from 0.10 to 0.34 (reliability ceiling 0.54), yet targets with the brain correspondence destroyed still reach 0.31. A shuffled target scores about k/n against the real one, for rank k and n sentences, a baseline that can be computed before any model is trained. Both interventions work, yet we detect no distance-graded steering effect and no syntactic benefit from alignment. Before reading a correspondence score, measure how much of it survives without the correspondence.

cs.CL · cs.AI · cs.LG · q-bio.NC

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.01271

Choosing Coding-Agent Benchmarks for Practice: Lessons from Task Demands and Agent Behaviour

Radin Shayanfar, Keheliya Gallaba, Ahmed E. Hassan

原文摘要

Software teams evaluating coding agents need evidence that reflects the changes they expect those agents to make. Benchmark scores summarize how often an agent resolves tasks under broad labels such as bug fixing and feature implementation, but provide little guidance on whether the evaluated work resembles a team's workload. To support this decision, we characterize 2,487 tasks from five public benchmarks and analyze 14,922 runs of six agent configurations. We introduce the Spread-Novelty-Centrality (SNC) profile to describe how the reference solution is distributed across files and directories, how much it adds relative to what it removes, and the dependencies, history, and complexity of the code it touches. We also compare the scope of agent patches and their file exploration. Every pair of benchmarks differs on at least two SNC axes, including benchmarks with the same task label. Among tasks resolved by Claude Opus 4.6, patches are larger than the reference on FeatBench and smaller on FeatureBench. Across the six configurations, resolved runs tend to involve less distributed changes and code with lower values on the Centrality indicators, while the patch sizes associated with resolution differ between the Claude and Qwen configurations. These observations suggest that practitioners should select evaluation tasks based on their change characteristics, inspect the information provided in task descriptions, and assess patch scope alongside test outcomes. We explain how teams can apply this analysis to completed changes and identify the additional evidence needed before using it for task assignment or review automation.

cs.SE · cs.CL

查看论文
arXiv ↗PDF ↗