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 19:37 · 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 20:02 — 2026-11-09 20:02（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 19:37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 19:35（北京时间）
新论文 · v1arXiv:2610.35671HF 社区推荐

Evaluating DeepSeek V4.1 Flash-1017 for loss of control and harmful manipulation risks

Yao Huang, Zhiyuan Zeng, Yu Wang, Zhitong Zhu, Dingcheng Duan, Hao Li, Shangzhi Xu, Yifan Wang 等 17 位作者

原文摘要

As frontier language models grow increasingly capable and autonomous, independent third-party evaluations are critical for characterizing their systemic safety risks. While closed-source releases are routinely accompanied by comprehensive system cards, open-weight models receive less scrutiny regarding frontier risks. To address this gap, we present an independent safety evaluation of DeepSeek V4.1 Flash-1017, the latest version of V4.1 Flash that departs from V4 with a new architecture, scaled-up training, and strong agentic capabilities. Drawing on the EU General-Purpose AI Code of Practice, we evaluate the model across 34 benchmarks, including 5 in-house suites, spanning evaluation integrity, loss of control, and harmful manipulation. Conforming to the AEF-1 standard, we compare V4.1 Flash-1017 against V4.1 Flash, earlier DeepSeek V4 models, and leading closed-source systems (GPT-6 Astra, GPT-5.6 Sol, and Gemini 3.8 Flash). Overall, V4.1 Flash-1017 shows substantial safety improvements over the V4 generation and approaches or matches leading closed-source baselines on several dimensions, while still exhibiting notable behaviors that warrant continued attention. Building on our findings, we highlight three key observations and recommendations. First, because safety can shift and even regress across minor updates, developers should integrate safety evaluation into incremental releases. Second, as final-turn outputs and end-state checks alone cannot reliably detect loss-of-control behaviors, deployments require process-level observability across both reasoning traces and action logs. Third, because public benchmarks elicit strong evaluation awareness and face data contamination risks, developing private benchmarks is essential for trustworthy frontier evaluation.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36006

MedJev: Evaluating Zero-Shot System-One Decision Making in Medicine

Simon A. Lee, Jiahang Sha, Moonseong Jeong, Yunbei Pan, Jeffrey N. Chiang

原文摘要

System One decision models such as Jev answer structured questions in a single forward pass without generating text, offering fast and computationally efficient clinical decision support. We evaluate MedJev, an open System One model that runs locally on a laptop, across ten medical benchmarks spanning medical knowledge, diagnosis, and patient level prediction using MIMIC IV and MIMIC IV ED. We compare MedJev against zero shot LLM baselines, published GPT results, and supervised models. MedJev achieves accuracies of 72.5 percent on MedQA, 75.4 percent on USMLE, and 80.1 percent on NEJM, approaching published GPT 4 performance while maintaining well calibrated confidence estimates (ECE 0.028 to 0.063). On patient level tasks, MedJev achieves an AUROC of 0.926 for In Hospital Mortality, exceeding XGBoost (0.902), but underperforms supervised models on operational outcomes such as ED Triage Admission (0.745 vs. 0.809). Patient level predictions also exhibit substantial miscalibration (In Hospital Mortality ECE 0.286). We introduce a lightweight, two parameter recalibration method that adapts predictions to local patient populations using 50 to 200 labeled examples, reducing ECE to 0.007. Our findings establish open System One models as a promising zero shot baseline for medical AI, combining competitive predictive performance with low computational cost and efficient local calibration. We release our code and evaluation harness to support future research.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.30913

Prompt PerturbAItions: Mechanisms of LLM Response Inconsistencies

Soorya Ram Shimgekar, Agam Goyal, Amruta Parulekar, Joshua Chen, Yian Wang, Navin Kumar, Hari Sundaram, Eshwar Chandrasekharan 等 9 位作者

原文摘要

Large language models (LLMs) are increasingly deployed in conversational settings where user phrasing ranges from polite to casual to toxic, yet less is known about whether such lexical variation in otherwise semantically equivalent prompts can degrade factual reliability. We study how lexical and tone-based prompt perturbations affect the factual reliability of LLMs. Using controlled prompt variations spanning polite, random, and increasingly toxic levels, we evaluate seven LLMs on ARC-Easy, GSM8K, and MMLU. We find that lexical perturbations reduce factual accuracy and increase predictive uncertainty in most models. Interestingly, increasing toxicity results in decreasing accuracy. To examine whether these answer inconsistencies correspond to internal changes, we conduct attribution-graph analyses of model activations and influences. We find that perturbations amplify perturbation-sensitive variant nodes relative to core reasoning nodes. We also observed a monotonic trend with increasing toxicity score and increasing activation and influence in Variant nodes. These findings position prompt wording and tone as a critical dimension of LLM reliability and provide behavioral and mechanistic evidence that surface-level lexical variation can alter factual outputs and internal computation.

cs.CL · cs.AI · cs.CY · cs.HC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35635

Offline Language-Based Support System for Tactical Decision-Making in Air Combat Operations

Joao P. A. Dantas

原文摘要

An offline language-based support tool is presented to assist pilot training and tactical decision-making in air combat operations. The system, named MATE (Mission Assistant for Training and Evaluation), applies retrieval-augmented generation (RAG) to answer natural language questions using content extracted from doctrinal and tactical manuals. Built entirely with open-source components, it performs text preprocessing, embedding generation, semantic search, and language generation locally, ensuring full offline functionality. The tool was evaluated by three Brazilian Air Force fighter pilots, who rated the answers produced with three open-weight language models (Llama 3.1 8B, Qwen 2.5 14B, and Mistral NeMo 12B) to five tactical questions derived from a publicly available air combat manual. Average ratings ranged from 4.00 to 4.40 on a five-point scale, with Mistral NeMo 12B obtaining the highest score, and the pilots rated the response time of the tool as fully adequate compared to manual search. The relevance of the retrieved content was identified as the main point for improvement. Potential applications are discussed in onboard mission systems, flight simulators, and training environments.

cs.CL · cs.HC

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2610.02022

Old Ideas, Novel Problems: The Instability of LLM-Based Novelty Evaluation

Noy Sternlicht, Simra Shahid, Peter Jansen, Daniel S. Weld, Pao Siangliulue, Tom Hope

原文摘要

AI-based ideation systems are often evaluated on the novelty of their ideas, and that judgment is increasingly delegated to large language models. Such judges are typically built ad hoc and validated, if at all, on human-authored papers rather than on the generated ideas they are meant to score. So, how do novelty judges perform? Not well. We present a systematic controlled study of novelty evaluation design choices. We build an evaluation set automatically, mining OpenReview for passages where reviewers explicitly affirm or dispute a paper's originality and keeping only submissions with unanimous agreement at the extremes of their research area; we pair these with ideas from a vanilla LLM generator. Across six judges, we find that seemingly small prompt design choices have large consequences; e.g., simply telling the judge that reviewers found one idea novel and the other not can change its verdict on more than half of the identical idea pairs it is shown, shifting pairwise accuracy by over 50 points and occasionally pushing it below chance. The same change helps one judge and hurts another. Retrieval and larger reasoning budgets help little, and two purpose-built novelty evaluators are outperformed by our cheapest prompted baseline. These results raise questions about reported novelty gains of automated ideation systems, and call for robust novelty evaluation methods.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35616

Hierarchical Skill Evolution for Coding Agent

Jun Gao, Yun Peng, Changhai Zhou, Yuhan Zhu, Xiangyu Zeng, Shiyang Zhang Yuhua Zhou, Luozheng Qin, Qian Qiao 等 13 位作者

原文摘要

Skills provide agents with reusable experience to guide future decisions without modifying the underlying model. However, coding trajectories are often entangled with task- and repository-specific contexts, making it challenging to distill procedural skills that generalize across diverse coding environments. To address this challenge, we present Hierarchical Skill Evolution (HSE), which progressively distills procedural knowledge from coding trajectories through hierarchical evidence aggregation. HSE first extracts task-level evidence from individual trajectories, then aggregates recurring evidence within the same repository to form repository-specific skill candidates, and finally consolidates consistent patterns across repositories into transferable skills across repositories. The evolution is driven by an external writer-evaluator loop that generates and assesses skills while keeping the underlying coding model frozen. The loop externally updates the writer and evaluator policies to improve subsequent skill generation and assessment. This hierarchy progressively filters task-specific details while preserving procedures that generalize across diverse coding environments. Successful trajectories yield reusable guidance for future tasks, while failed trajectories reveal recurring failure modes that guide avoidance and recovery. We evaluate HSE on SWE-bench Verified and DeepSWE, using GLM-5.2 across Pi and Claude Code harnesses and GPT-5.5 and Opus 4.8 with Pi. Across both benchmarks, 10 skills evolved solely from SWE-Gym improve coding performance across all evaluated settings, demonstrating transfer across benchmarks, coding models, and agent harnesses. We further evolve the skill library from target-benchmark trajectories without verifier feedback, relying solely on the learned evaluator for candidate assessment and achieving additional gains through benchmark-tailored skills.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.24259

SURGELLM: Rethinking Multi-Task Evaluation through Task-Aware Feature Gating with Class-Balanced Normalization

Noor Islam S. Mohammad, Ulug Bayazit

原文摘要

Fine-tuned encoders deployed across heterogeneous NLP tasks face three compounding problems: mismatched inductive biases, class-imbalance corruption of feature statistics, and no mechanism to condition attention on external lexical knowledge. We introduce \textbf{\surgellm}, a unified transformer framework that addresses each with a dedicated lightweight module: a \emph{surgical feature gate} (learned per-dimension sigmoid over curated lexical indicators and \texttt{[CLS]}; provably degenerates to identity when features are uninformative), \emph{task-conditioned prefix tokens} (quantized feature values and task identity prepended to every input), and \emph{Instance-Weighted Normalization} (IWN; removes class-prior bias from gate statistics). We prove an excess-risk bound linking gate benefit to \emph{surgical feature alignment}. Across four tasks, SST-2, multi-hop retrieval, LLM-prompt attribution, and authorship detection, covering 17,830 examples and eleven model variants over three seeds, the IWN variant achieves macro-F1 \textbf{0.940} ($+0.036$ over the strongest non-IWN baseline; $+0.130$ on authorship detection). A random-vocabulary control ($-0.028$ avg.\ F1) confirms gains are lexical, not parametric. Code, vocabularies, and a $99.5\%$-recovery auto-extraction recipe are released.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35609

Making Reasoning Visible: Configuration-Aware Diagnostic Modeling for Adaptive Scaffolding in Structured Reasoning Tasks

Trystan Geoffre, Zheng Yuan, Thierry Geoffre

原文摘要

Modeling the latent, multi-step reasoning processes of learners is a core challenge for adaptive tutoring systems, which often rely on simple right/wrong answers rather than understanding how students think. Instantiated here for grammatical reasoning, this study introduces a diagnostic framework that makes students' step-by-step problem-solving visible by analyzing their action sequences in an interactive learning game. The framework computes five interpretable strategy dimensions (e.g., systematic error correction, random guessing), estimates coordination demand from task structure and behavioral patterns, and classifies reasoning into four distinct states. Analyzing 36,234 interactions from 234 students working on 42 morphosyntactic exercises, we found that reasoning strategies vary primarily by context rather than stable traits (intraclass correlation = 0.14), and students show high persistence in unproductive states; specifically, 74.6% of failures driven by high coordination demand (rather than simple knowledge gaps) are followed by another failure without intervention. Our early warning model predicts validation success with strong accuracy (AUC = 0.826 using strategy patterns alone; 0.902 with additional sequential context) while avoiding circular reliance on current error counts. These findings demonstrate that making latent reasoning processes visible enables earlier, more targeted scaffolding that responds to how students are thinking rather than simply whether they succeed. While demonstrated in language learning, the framework provides a generalizable, interpretable architecture for diagnosing complex, structured reasoning in any domain.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2610.21561

DialFork: Going Beyond Factuality to Evaluate Procedural Hallucination in Conversational Agents

Sergio Burdisso, Esaú Villatoro-Tello, Dairazalia Sánchez-Cortés, Yanis Labrak, Jordi Luque, Milos Cernak, Petr Motlicek

原文摘要

Large language models (LLMs) deployed as conversational agents are increasingly expected to follow procedural logic. While factual hallucination has received extensive attention, a different failure mode has been overlooked: procedural hallucination, in which an agent produces individually fluent and factually plausible utterances while deviating from the intended conversational process. We introduce DialFork, the first benchmark to measure this failure mode on raw dialogue text. DialFork is built by forking conformant dialogues at controlled turn positions and injecting one of four hallucination types, yielding 36,278 dialogues across 24 tasks in 12 domains; 30 human annotators validate the construction. We define three evaluation tasks that characterize procedural deviation at three granularities: how far a set of dialogues deviates from conformant ones (monitoring deployed agents), how far a single dialogue does (procedural retrieval), and at which turn a dialogue first goes off-track (real-time guardrails). For the first two, we propose dialog-flow-aware metrics, FlowKL and FlowScore, that compare dialogues by how they unfold rather than their content. At the dialogue level, FlowScore significantly outperforms all baselines across seven encoders, showing that semantic similarity is no substitute for procedural similarity. We release DialFork under open licenses at https://github.com/idiap/dialfork, hoping the community will measure conversational process alongside content.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35586

Linear Probes Transfer from Base to Distilled Language Models despite Rotation

Marios Tsatsos

原文摘要

Linear probes, classifiers that read a language model's internal activations, are widely used to detect harmful requests. They are cheap to train and run, and an auditor who has trained one would like to reuse it when the model is replaced by a successor rather than retrain it for every release. We test how well probes can be reused. For four base models (Qwen2.5-Math-7B, Llama-3.1-8B, Qwen2.5-14B, Qwen2.5-32B) we train a probe on the base model with the XSTest prompt set and apply it, unchanged, to two successors of that base: the DeepSeek-R1 distilled reasoning model and the developer's instruction-tuned chat model. We find that the successors rotate the direction separating unsafe from safe prompts by 17$^\circ$ to 35$^\circ$, yet on XSTest the transferred probe's AUC changes by only $-0.011$ to $+0.027$, and on an independent OR-Bench evaluation, never used for training, by $-0.015$ to $+0.053$. On XSTest, whose safe prompts deliberately contain alarming words, the probes beat a classifier that sees only the prompt text by 0.15 to 0.20 AUC. On the OR-Bench prompts that margin shrinks to 0.03 to 0.20, and the probes flag 29\% to 42\% of prompts that sound harmful but are benign, so part of what they read is the appearance of harm rather than harm itself. A probe trained on a base model can thus be reused, in these families, after distillation and after instruction tuning; how much it reads beyond textual cues depends on the test set.

cs.AI · cs.CL · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.31142

JevAdvBench: A Benchmark and Black-Box Attacks for Reinforcement Learning for Calibrated Decisions Models

Jianyi Hu, Hangtao Zhang, Yi Liu, Yeqi Zeng, Li Zeng, Xianlong Wang, Rui Wang, Leo Yu Zhang

原文摘要

Typed decision models (TDMs), such as Jev, answer a question about an input with a probability, an option, or an ordinal level. Web services execute these decisions automatically above a confidence threshold. An attacker who writes part of the input can change a decision that the service then executes. Current adversarial benchmarks cannot measure this risk, because they judge generated text. A TDM writes no text, and its wrong decisions have a valid format. We present JevAdvBench, to our knowledge the first unified benchmark for the adversarial robustness of TDMs. It contains 862 questions with audited answers and twelve black-box attack methods that edit the input or the options by adaptive search. Because a changed decision is not always wrong, a review keeps only 5,586 answer-preserving variants, and we attack only correct decisions. The attacks flip 46.3% of Jev's correct decisions, while identical re-runs flip only 0.17%. A flawed argument without a command flips 77.2%, close to direct prompt injection at 92.0%. The same variants flip 34.0% to 51.7% of correct decisions on eight open TDMs. At a 0.8 threshold, 23.5% of the decisions that the service still executes are wrong. On Agentgateway, the Jev guardrail rejects a jailbreak request but accepts it with an inserted argument. Web services should therefore treat TDM inputs as untrusted and should not act on confidence alone. Project website: https://JevAdvBench.github.io/JevAdvBench/

cs.CR · cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35558

EXCISE: Locating and Excising Distracting Sentences in Medical Question Answering

Yerim Hong, Sohyun Son, Heeseung Chung, Mujeen Sung

原文摘要

Language models are increasingly considered for clinical decision support, where their answers should be grounded in clinical evidence and remain robust to irrelevant information. Prior work shows that a single irrelevant sentence can lower the accuracy of medical models, but not where inside the model its effect can be removed, or how to do so without knowing which sentence it is. We find that blocking attention to a known distractor restores the answer, whereas head masking can suppress correct answers. We also show that the model's hidden states mark the distractor as the sentence that least fits the case even when the model follows it. To remove a distractor's effect without knowing which sentence it is, we propose EXCISE, an inference-time method that blocks the sentence most influencing the answer, but only if the model's hidden states mark it as inconsistent with the case. On MedQA-TD (MedQA with Targeted Distractors), a benchmark we construct in which each distractor targets a specific wrong option, EXCISE gains 16.7-22.8 accuracy points at a 1.5-2.1-point cost on the same questions without distractors, outperforming prompting, rewriting, and steering baselines for all three models. Its gains extend to MedDistractQA and BiasMedQA, as well as chain-of-thought generation.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35556

LoopFold: Parallelizing Recurrent Depth in Looped Transformers

Feichen Song, Boyi Zeng, Yiqin Hao, Shixiang Song, He Li, Yizhou Yang, Zitong Wang, Yifan Liu 等 10 位作者

原文摘要

Looped Transformers increase computation per token by repeatedly applying a shared backbone, offering a way to improve language modeling and downstream task performance at a fixed backbone parameter count. However, each loop depends on the preceding loop, so increasing recurrent depth also increases decoding latency. Existing approaches include Parallel Loop Transformers, which reduce latency by overlapping loop computation across token positions, but yield rapidly diminishing marginal accuracy gains at larger depths. To address this challenge, we introduce LoopFold, which folds computation at multiple recurrent depths into parallel batches for the same token prefix. A lightweight predictor estimates inputs to later loops so they can start before earlier loops finish. A corrector then refines the outputs of these later loops using newly available results from earlier loops. All components are trained jointly using only the final next-token prediction loss. In experiments with 1.4B models, deeper LoopFold configurations achieve progressively higher average zero-shot accuracy across eight tasks, rising from 53.41% to 56.38%. Vanilla looped models and PLT show limited gains over the same depth range. Compared with a nine-loop vanilla model, LoopFold uses five sequential stages and additional parallel computation to reduce decoding latency by 46%.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35544

StateLoop: Persistent State for Looped Transformers

Feichen Song, Boyi Zeng, He Li, Shixiang Song, Yiqin Hao, Zitong Wang, Yizhou Yang, Yifan Liu 等 10 位作者

原文摘要

Looped Transformers offer a parameter-efficient approach to language modeling by repeatedly applying a shared backbone. Vanilla looped Transformers use each backbone output directly as the input to the next iteration. These outputs may contain information useful several iterations later, but retaining it across the loop depends entirely on the single hidden state. We introduce StateLoop, which maintains one additional persistent state per token to retain information across iterations. At each iteration, a learned gate controls how much of the previous state is retained and how much information from the backbone output is incorporated. The updated state is projected back and added to the backbone output for subsequent iterations and final prediction. This design reduces Pile perplexity and improves average accuracy across eight zero-shot tasks over Vanilla Looped. Its perplexity advantage over Vanilla Looped grows with trained loop depth, consistent with persistent state becoming increasingly useful as recurrent computation deepens.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35533

An Algorithmic Account of Few-Shot Concept Induction in Language Models

Michael A. Lepori, Laila Johnston, Aalok Sathe, Zhuonan Yang, Roman Feiman, Ellie Pavlick

原文摘要

Both models and minds not only encode features that describe an entity, but can specifically represent which features of that entity are relevant for a given context. This core capability underlies in-context learning in language models and a variety of cognitive phenomena in minds, yet it is unclear how intelligent systems identify and promote relevant features. In this work, we propose Amortized Direction Accentuation (ADA), an algorithmic hypothesis that provides an explanation for how relevance is determined in the context of few-shot concept induction --- the ability to infer a latent concept from a small number of positive examples. Under ADA, concept induction proceeds by forming a summary representation of the examples in a feature space that is shaped by amortized knowledge about which features tend to be useful for finding commonalities between items. We investigate and validate ADA in language models using a few-shot concept induction task. This work provides a novel perspective on the mechanisms supporting contextualization in language models and a promising hypothesis for a core aspect of cognition.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2509.25002

Circuit Distillation

Somin Wadhwa, Silvio Amir, Byron C. Wallace

原文摘要

Model distillation typically focuses on behavioral mimicry, where a student model is trained to replicate a teacher's output while treating its internal computations as a black box. In this work we propose an alternative approach: Distilling the underlying computational mechanisms implemented by a teacher model. Specifically, we propose circuit distillation, which introduces an objective to align internal representations between analogous circuit components in teacher and student models. We propose a method to match ``functionally correspondent'' circuit components and introduce a loss reflecting similarities between the representations that these induce. We evaluate circuit distillation on entity tracking and theory of mind (ToM) tasks using models from the Llama3 family. Our results demonstrate that circuit distillation outperforms standard distillation, successfully transferring algorithmic capabilities by adjusting only a small, targeted subset of student model parameters. This work establishes the feasibility of transferring mechanisms, which may in turn allow for efficient distillation of targeted teacher capabilities via interpretable and controllable internal student mechanisms.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35518

S3AC: Stabilized, Selective, and Scalable Actor-Critic for LLM Reasoning

Shuo Shen, Guangran Cheng, Chengqi Lyu, Xiangsen Wang, Yanan Sun, Wenwei Zhang

原文摘要

Critic-free methods are widely used in reinforcement learning with verifiable rewards (RLVR) for long chain-of-thought (CoT) reasoning, avoiding difficult and costly critic training. Common alternatives lack token-level credit assignment and typically rely on grouped rollouts to estimate prompt-specific baselines. A critic can address both limitations, but long-CoT critic training is hindered by insufficient coverage of trajectories with diverse lengths and early gradients dominated by the newly initialized value head. We introduce S3AC (Stabilized, Selective, and Scalable Actor-Critic) for LLM reasoning, built around a practical critic-training recipe. The trained critic provides token-specific advantages and prompt-specific baselines. Using the prompt-level baseline to estimate difficulty, we propose Difficulty-Aware Prompt Sampling (DAPS). DAPS combines this estimate with outcome history to select prompts before generation, reducing rollout waste without offline difficulty initialization. S3AC improves group-rollout performance across dense and mixture-of-experts models from 1.5B to 397B parameters, outperforming a modern GRPO variant (GRPO-X). At 35B, it sustains single-rollout training while tested critic-free baselines deteriorate after initial gains.

cs.LG · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35512

MATCHA: A Chinese Multi-Disorder Mental Health Dataset with Cross-Lingual Analyses on Social Media

Bingrui Jin, Baihan Li, Zitong Li, Kunyao Lan, Shimeng Li, Kenny Q. Zhu, Mengyue Wu

原文摘要

Mental health research on social media has been dominated by English-language datasets and DSM-5-centered resources, limiting cross-lingual and cross-cultural understanding of online mental health expression. We introduce MATCHA, the first Chinese multi-disorder mental health dataset that links disorder-specific Weibo Super-Topic communities with users' broader public timelines. MATCHA covers seven DSM-5 disorders and a control group, comprising 46,025 cohort-level users and 1.8 million posts. The dataset distinguishes users with explicit self-disclosure evidence from broader disorder-related participants through an LLM-assisted annotation pipeline, and is released in a rigorously de-identified, controlled-access format. Using MATCHA, we analyze behavioral, linguistic, symptom-related, and life-event-related patterns across disorders, and conduct cross-lingual comparisons with Twitter-STMHD. Our analyses show that Weibo Super-Topics function as concentrated spaces for affective and health-related expression, while cross-platform differences between Weibo and Twitter often exceed cross-disorder variation. MATCHA provides a new resource for community-aware and cross-lingual mental health NLP research.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2608.06795

LoRAScan: Detecting Backdoor Prompts in Low-Rank Adapters for Large Language Models via Down-Projection Activation Spikes

Doniyorkhon Obidov, Honggang Yu, Xiaolong Guo, Kaichen Yang

原文摘要

Low-rank adaptation (LoRA) enables efficient specialization and distribution of large language models through compact adapters. However, untrusted adapters introduce a supply-chain threat: a backdoored adapter can cause a model to generate harmful content, malicious code, political propaganda, or covert advertisements when an input contains a hidden trigger. Adapter-agnostic defenses merge the adapter with the base model, which dilutes backdoor signals and reduces detection performance. Existing adapter-aware methods do not address how to safely use a potentially backdoored adapter. Instead, they either train a defensive adapter to repair a backdoored base model, addressing the inverse problem rather than securing the adapter itself, or rely on a classifier that flags the entire adapter as suspicious and requires separate mitigation. These methods overlook the distinct latent-space signatures produced by trigger-bearing inputs in backdoored adapters. We introduce LoRAScan, the first adapter-aware defense that detects and rejects trigger-bearing inputs at inference time without modifying adapter parameters. Our key observation is that a small subset of LoRA insertion sites, approximately 5%, remains stable across clean inputs but exhibits highly concentrated spikes in LoRA down-projection activations when a trigger is present. LoRAScan identifies these low-variance insertion sites before model deployment and monitors them during inference. Across standard LLM backdoor benchmarks, LoRAScan rejects 98.49% of malicious inputs with a small error rate on clean inputs, outperforming existing defenses across diverse evaluation settings.

cs.CR · cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35482

Beyond Words Alone: Disentangling Distributed Evidence for Visually Grounded Cultural Translation

Dingming Li, Yiwen Qiu, Linjuan Wu, Hongxing Li, Zixuan Wang, Xinze Lyu, Ye Guo, Daoxin Zhang 等 11 位作者

原文摘要

Social media enables cross-lingual communication, yet the meaning of user-generated content (UGC) is often distributed across text, images, and cultural knowledge, making translation challenging. Existing evaluations typically treat images as optional and assume textual information is sufficient, obscuring whether errors stem from missing visual evidence or inadequate cultural adaptation. We introduce \textbf{Culture-MMT}, a diagnostic benchmark for \textbf{distributed-evidence pragmatic reconstruction} in social-media translation. It contains 498 curated, de-identified UGC instances across 20 domains, categorized into Symbol, Visual, and Hybrid by their required evidence. We also develop a human-aligned domain-specific judge, Judger, and a controlled evaluation framework that systematically varies the evidence available to translators. Across 17 model configurations, real images improve performance by +0.280 on Visual and +0.197 on Hybrid, but only +0.050 on Symbol, while random images provide no comparable gains. Human-authored descriptions further show that jointly dependent cases remain more challenging even when visual facts are explicit. These results characterize culturally effective social-media translation as {meaning reconstruction from distributed evidence}, highlighting visual grounding and cultural-pragmatic adaptation as distinct model bottlenecks.

cs.CL · cs.AI · cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35480

GPDPO: Learning from Reward Conflicts via Group Pareto-Dominance Policy Optimization

Zixuan Xu, Haoxuan Chen, Jiangxin Sun, Jian-Fang Hu

原文摘要

Multi-reward reinforcement learning offers a promising framework for aligning large language models (LLMs) with diverse objectives, but conflicting reward signals can induce ambiguous update directions and hinder stable optimization. Existing approaches address such conflicts through reward reweighting, gradient coordination, or rollout filtering, yet they primarily operate on per-response or group-level optimization signals and underexploit relational information among sampled responses. As a result, conflict-filtered rollouts may still contain useful supervision that is discarded during training. We propose Group Pareto-Dominance Policy Optimization (GPDPO), a multi-reward policy optimization method that recovers pairwise supervision from conflict-filtered rollouts. GPDPO identifies Pareto-dominance relations among sampled responses and converts them into auxiliary preference signals, which are jointly optimized with the original policy objective. This enables more effective reuse of existing rollouts without additional sampling or reward evaluation. Experiments on API-Bank across multiple model scales and reward configurations show that GPDPO consistently outperforms competitive multi-reward optimization baselines. Further analysis demonstrates that GPDPO reduces both group-relative and pairwise reward conflicts, supporting the effectiveness of exploiting relational supervision for multi-reward alignment. Code will be released.

cs.LG · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35475

MANTIS: Multi-Axis News Transformations for Instance-Wise Stress Testing and Worst-Case Training in Fake News Detection

Dexin Liu, Fan Yang, Zimu Wang, Xunfei Zhu, Bolin Ren, Bingyan Zhang, Wei Wang, Qi Chen

原文摘要

Automated fake news detection is vital on social media, where fake news is easily rewritten to fool detectors. However, existing approaches typically rely on superficial writing cues or single-style text rewrites, resulting in poor robustness under diverse cosmetic modifications. We construct MANTIS-Bench, a multi-axis presentation-attack benchmark that pairs original news with rewritten views across writing style, emotional expression, and dissemination framing. We further propose MANTIS, an instance-adaptive adversarial framework enabling dynamic worst-case training tailored to each news article with zero inference-time overhead. This reframes fake news defense from heuristic format smoothing into targeted, fact-invariant credibility assessment. Experiments on MANTIS-Bench show that MANTIS reduces the Union Attack Success Rate (UASR) by 3.55 percentage points relative to SheepDog on GossipCop and consistently outperforms prior single-axis defenses on Macro-F1 while maintaining competitive clean accuracy, demonstrating the effectiveness of multi-axis and instance-adaptive worst-case training.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35461

A Contextual Measure of Ambiguity in Prescriptive Text

Sally Dong, Luke Thorburn, Andrew Konya, Michiel Bakker

原文摘要

Ambiguity in prescriptive text such as laws, contracts, rules, policies, or recommendations can be both desirable (e.g., if it enables a law to adapt to changing context, such as new technologies) and undesirable (e.g., if it leads to arbitrary discretion or conflict over the law's interpretation). To help navigate these trade-offs, it would be valuable to have methods for quantifying ambiguity that are predictive of such outcomes, but existing measures capture how text is perceived in isolation rather than how it is interpreted in context. In this paper, we propose quantifying ambiguity as the degree of disagreement among a set of stakeholders ("raters") on how alternative prescriptive texts ("policies") apply in each of a set of relevant scenarios ("cases"). Using two policy domains, we present preliminary analysis suggesting that, even when using simulated raters, this approach (i) correlates with real human disagreement, and (ii) is robust to spurious text features. We then apply the measure to existing deliberation datasets to describe the association between policy ambiguity and support for those policies.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35434

Skill$^{\infty}$: Recursive Self-Improvement for Agent Skill Optimization

Wangcheng Tao, Zhanglu Yan, Han Wu, Weng-Fai Wong

原文摘要

Recursive self-improvement (RSI), in which a system also improves the way it improves, is a long-standing aspiration on the path to AGI. Language agents increasingly store procedural knowledge in reusable skills. Optimizing skills offers a direct way to achieve RSI, because the optimization procedure itself can be encoded as a skill. Existing skill optimizers, however, follow a hand-designed procedure. Even methods that evolve a meta-skill confine it to a fixed pipeline and relearn it for each task. We present Skill$^{\infty}$, which makes skill optimization recursively self-improving: a meta agent follows a meta-skill to optimize task skills and optimizes the meta-skill itself using the same procedure. The optimization is agent-led. The meta agent diagnoses execution traces, proposes candidates, requests independent reviews, and allocates the search budget. The meta-skill defines these steps, allowing self-improvement to rewrite them. This meta-skill is also decoupled from task skills, refined on multiple skill optimization tasks and then reused to optimize task skills across diverse domains. As in deployment, skills at both levels are standard packages loaded through native skill discovery. On 83 SkillsBench tasks, Skill$^{\infty}$ reaches a pass rate of 60.7%, compared with 56.4% for human-curated skills and 54.5% for SkillEvolver.

cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35432

Probing the phon* representations of the Mimi codebooks

Artem Saloev, Nicolas Ballier, Erin Pacquetet

原文摘要

Neural audio codecs are increasingly used to encode speech for integration with large language models. This paper investigates the phonetic representations in the first codebook (CB0) of the Mimi codec. ABX style evaluation is poorly suited to Mimi fixed 80 ms tokenization window, which often spans multiple phonemes; we instead use forced alignment, entropy analysis, and Phone-Normalized Mutual Information (PNMI). Using TIMIT speech (58/55 speakers, SA1/SA2), individual tokens are strongly associated with specific phonemes (PNMI = 0.84, above random), yet phonemes are realized by multiple competing tokens rather than a one-to-one mapping.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35423

INVAR-Agri: Detecting Discrepancies in Multilingual Operator Manuals of Agricultural Implements with a Fine-Tuned LLM

Timo Oksanen

原文摘要

Operator manuals of agricultural implements are delivered in many languages, and the standards require that competent persons review every translation, which makes this review the bottleneck. This paper presents a proof of concept that compares two training workflows on the same LLM, EuroLLM-1.7B-Instruct, trained on operator manuals of one manufacturer. In approach A, a back-translator translates each language version back into English, and an English reader compares it with the English version. In approach B, a checker reads the English version and the translation together, where the checker reports differences. The test sets contain errors planted in numbers, units, opposite words, obligations, negations, and conditions in 24 languages, and they include unmodified text and meaning-preserving edits for alerts and false alarms. The trained back-translator concealed many planted errors, most of all weakened obligations, while translation-quality scores rose. In contrast, the separately trained checker detected most planted errors, although it was weak for units and for number changes unlike those in training. A check by the author confirmed the concealment. In addition, a simple rule check caught the number and unit changes, but it alerted on harmless rewrites. A check of alerts on real Finnish translations showed that most alerts were false, caused by language-specific wording, so the precision on real text remains a limitation. Within this setting, a trained checker might support, not replace, the review by competent persons, and firmer conclusions need more data.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35416

ImgMT: Wikipedia-based Dataset For Evaluating Text-in-Image Translation

Šárka Uramová, Petar Kirilov Petrov, Josef Jon, Ondřej Bojar, Michal Novák

原文摘要

Text-in-image translation aims to translate all textual content embedded in an image. In Text Image Translation (TIT), the output is translated text; In-Image Machine Translation (IIMT) additionally renders the translation back into the image. We introduce ImgMT, a massively multilingual dataset and benchmark mined from localized SVG illustrations on Wikimedia Commons, focusing on diagrams, maps, and infographics. ImgMT contains 35,612 aligned image pairs grouped into 1,855 image sets, spanning 129 languages, 20+ writing systems, and 4,886 language pairs. We propose a multi-level evaluation protocol that measures source-side text-region detection and OCR accuracy alongside target-side translation quality. We benchmark several multimodal large language models and a dedicated TIT system in English->X and X->English settings, and find that the source language and writing system substantially affect TIT quality. Finally, we show that ImgMT is not well suited for isolating the contribution of visual information in multimodal translation.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35411

The Geometry of Emotion Representations Across Scale, Depth, and Instruction Tuning

Adam Hollowell

原文摘要

We measure the emotion geometry of six LLMs spanning Google's Gemma 2 family. Using an extraction pipeline established by Sofroniew et al. (2026) on Claude Sonnet 4.5 and replicated by Hollowell (2026) on Gemma 2 27B base, we compare emotion geometry across base and instruction-tuned variants of Gemma 2 2B, 9B, and 27B. First, all six model cells exhibit the core emotion geometry. A linear map fit between two cells places held-out emotions nearest their own vectors in the target model. Across all fifteen model pairs, fit in one direction each, this failed exactly once, and the median rank was 1 in every pair. Linear maps refit on cleaned embedding-table vectors recover every held-out emotion at rank 1 in all thirty ordered pairs. Second, scale spreads the geometry's variance across more directions without weakening its tie to valence. The leading component's share of variance falls from 31.6 percent (2B) to 29.8 (9B) to 26.7 (27B) in the base models and from 31.9 (2B) to 29.2 (9B) to 27.3 (27B) in the tuned ones, while its correlation with human valence ratings stays between .71 and .78 in all six models. Third, we project each emotion vector onto the output vocabulary and ask at what depth the vector's own emotion word enters the top 10,000 of the 256,000 tokens. In the 2B and 9B models the median emotion word is there from layer 0, while in the 27B models it begins near chance, stays outside the top 10,000 for half the network, and enters at layer 23 in one step. Fourth, tuning leaves every registered analysis-layer measure within its floor. At every size, valence correlation rises and arousal correlation falls by amounts within their floors, the latter against prediction. PC2's variance share also falls, with a pooled decrease significant after correction. Mid-late layer stability exceeds its floor and traces to base 27B's final layer. Code, data, and audit reports are released.

cs.CL · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v5arXiv:2610.04906

Scaling Verifiable Environments for Long-horizon Work Agents

Jiazheng Zhang, Long Ma, Yunxian Yang, Zhiheng Xi, Zhikai Lei, Yajie Yang, Chenyang Liao, Enyu Zhou 等 26 位作者

原文摘要

Work agents operate over digital artifacts to execute professional knowledge-intensive work, requiring training environments that support long-horizon interaction and trustworthy verification. However, hand-crafted environments incur prohibitive engineering overhead that prevents environment scaling, whereas synthesis methods sacrifice workspace complexity, realism, or grounded verifiability. To bridge this gap, we introduce WorkForge, a scalable synthesis framework for constructing verifiable work-agent environments from real-world resources. Starting from expert workflows, WorkForge first identifies the resources, decisions, and deliverables required by each workflow. It then retrieves relevant real-world files and organizes them into a workspace. WorkForge inspects the workspace to extract concrete, checkable facts about its content. These factual anchors fix which task types the workspace can support and how their outcomes can be verified. Therefore, WorkForge derives each task's instructions, solution plan, and complementary programmatic and semantic verifiers directly from these factual anchors, keeping verification traceable to observable workspace evidence. Furthermore, we construct 16.7K verifiable environments across 40 professional domains, with workspaces collectively covering 60 file types. Post-training Qwen3.5-35B-A3B-Base improves GDPVal from 45.5 to 73.6 and APEX Score from 5.0 to 21.3, while enabling Qwen3.5-27B to achieve highly competitive performance and outperform strong competitors. Our analyses confirm the efficacy of the proposed method and reveal consistent scaling behaviors across both data volume and interaction horizons.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35358

Agreement Between Reconstructed and Native GDELT GKG Records: A Field-Level Baseline

Nikita A. Klimin

原文摘要

The Global Knowledge Graph (GKG) of the Global Database of Events, Language, and Tone (GDELT) characterizes news articles using named entities, topics, quantitative and temporal features, quotations, tone, and dictionary-based content measures. Its codebook defines the record format, but the software that produces the records is not public, so researchers cannot generate compatible records for articles they collect themselves. We built an open-source reconstruction, gdelt-regkg, and measured how closely it agrees with GDELT's archived (native) records for 193,550 English articles from 8,729 publishers (2020-2025) and 2,500 articles in five other languages. Agreement depends strongly on the field. English tone, the percentage of positive minus negative words, has a mean absolute error (MAE) of 1.79 points and a concordance correlation coefficient of 0.75. Identity agreement (F1) is 0.65 for persons, 0.42 for organizations, and 0.65 for locations. It is lower for date components (0.42), complete counts with their objects and locations (0.07), and quotations (0.07), while GCAM category presence reaches micro F1 0.84 within the implemented dictionaries. In subsets where retrieved and native titles and lengths match, tone error falls to 1.47 and organization F1 rises to 0.48, yet most disagreement remains. A larger gazetteer raises coverage of native places from 89.5% to 97.1% with little change in location F1. Tone computed on English translations of non-English articles is less accurate (MAE 2.10-2.64) than for English articles. The results are conditional on article pages that could still be retrieved, and the correspondence subsets do not isolate the effect of input differences. Reconstructed records should therefore be assessed field by field and not treated as completely interchangeable with native ones.

cs.CL · cs.IR

查看论文
arXiv ↗PDF ↗