跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+979新增论文
+781其中 AI 相关
+119已有论文修订
+95新增 AI 导读

arXiv 最近完成扫描:2026-11-20 08:48 · 全站最近入库:2026-11-20 19:58各信源进度 →

统计口径与数据新鲜度

统计 2026-11-20 02:48 — 2026-11-21 02:48(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-20 08:48最近发现新增版本:2026-11-20 19:58库内论文最新更新时间:2026-11-20 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。

28,961 已收录论文24,030 AI 相关8,912 有修订112 HF 社区发现

浏览论文库

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 79 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-20 08:46(北京时间)
版本更新 · v3arXiv:2605.18565

MINTEval:评估长时程智能体系统在多目标干扰下的记忆能力

MINTEval: Evaluating Memory under Multi-Target Interference in Long-Horizon Agent Systems

Hyunji Lee, Justin Chih-Yao Chen, Joykirat Singh, Zaid Khan, Elias Stengel-Eskin, Mohit Bansal

AI 速读 基于摘要

作者提出 MINTEval 基准,用于评测长时程、信息频繁更新且相互干扰的场景下,记忆增强智能体的召回与多证据聚合推理能力。

cs.CL · cs.AI

查看原文摘要

Real-world agents operate over long and evolving horizons, where information is repeatedly updated and may interfere across memories, requiring accurate recall and aggregated reasoning over multiple pieces of information. However, existing benchmarks focus on static, independent recall and fail to capture these dynamic interactions between evolving memories. In this paper, we study how current memory-augmented agents perform in realistic, interference-heavy, long-horizon settings across diverse domains and question types. We introduce MINTEval (Long-Horizon Memory under INTerference Evaluation), a benchmark featuring (1) long, highly interconnected contexts with frequently updated information that induces substantial interference, (2) diverse domains (state tracking, multi-turn dialogue, Wikipedia revisions, and GitHub commits), enabling evaluation of domain generalization, and (3) diverse question types that assess robustness to interference, including (i) single-target recall tasks requiring retrieval of a specific target from long contexts, and (ii) multi-target aggregation tasks requiring reasoning over multiple relevant pieces of information. Overall, MINTEval has 15.6k question-answering pairs over long-horizon contexts averaging 138.8k tokens and extending up to 1.8M tokens per instance. We evaluate 7 representative systems, including vanilla long-context LLMs, RAG, and memory-augmented agent frameworks. Across all systems, we observe consistently low performance (avg. 27.9% accuracy), especially on questions requiring aggregated reasoning over multiple pieces of evidence. Our analysis shows that performance is primarily limited by retrieval and memory construction. Furthermore, current memory systems struggle to recall and reason over earlier facts that are revised or interfered with by subsequent context, with accuracy degrading as the number of intervening updates increases.

新论文 · v1arXiv:2611.26358

Beam Model Technical Report

Reflection Team, Adam Abdulhamid, Osama Abuelsorour, Yasmin Alameddine, Carlo Alfano, Duarte Alves, Brandon Amos, Selvaraj Anandaraj 等 230 位作者

原文摘要

Beam is a sparse Mixture-of-Experts (MoE) model with 501.3B total parameters and 23.4B active parameters, specifically built and optimized for coding, reasoning, and agentic workloads. Beam delivers frontier level reasoning, while requiring 3-4x less inference compute than many models of its class. Pretrained on 23.8T diverse, curated, high-quality tokens, Beam features a 1M-token context window achieved through a novel two-stage midtraining phase, built on verifiable, correctness-anchored data. Post-training leveraged a large-scale, fully asynchronous reinforcement learning (RL) campaign across 10.5K GB300 GPUs, generating over 100M rollouts over 4 weeks of training. We believe this is one of the largest scale RL runs conducted by any open lab to date. Across our evaluation suite, capabilities continued to improve as we increased RL compute, with no sign of a plateau. Supported by deep hardware-model co-design, FP8/NVFP4 quantization, and a 1.93B-parameter speculative decoding drafter (DFlash 2), Beam maximizes serving throughput with efficient, flexible deployment. Averaged across conversation turns and harm categories, Beam refuses or pivots on 88.9% of harmful requests while complying with 81.5% of harmless requests, achieving the highest refusal/pivot rate on harmful requests among open-weight models on the Pareto frontier.

cs.CL · cs.AI

版本更新 · v4arXiv:2505.03563

Say It Another Way: Auditing LLMs with a User-Grounded Automated Paraphrasing Framework

Cléa Chataigner, Rebecca Ma, Prakhar Ganesh, Yuhao Chen, Afaf Taïk, Elliot Creager, Golnoosh Farnadi

原文摘要

Large language models (LLMs) are highly sensitive to subtle changes in prompt phrasing, posing challenges for reliable auditing. Prior methods often apply unconstrained prompt paraphrasing, which risk missing linguistic and demographic factors that shape authentic user interactions. We introduce AUGMENT (Automated User-Grounded Modeling and Evaluation of Natural Language Transformations), a framework for generating controlled paraphrases, grounded in user behaviors. AUGMENT leverages linguistically informed rules and enforces quality through checks on instruction adherence, semantic similarity, and realism, ensuring paraphrases are both reliable and meaningful for auditing. Through case studies on the BBQ and MMLU datasets, we show that controlled paraphrases uncover systematic weaknesses that remain obscured under unconstrained variation. These results highlight the value of the AUGMENT framework for reliable auditing.

cs.CL

版本更新 · v4arXiv:2506.16552

Revela: Dense Retriever Learning via Language Modeling

Fengyu Cai, Tong Chen, Xinran Zhao, Sihao Chen, Hongming Zhang, Sherry Tongshuang Wu, Iryna Gurevych, Heinz Koeppl

原文摘要

Dense retrievers play a vital role in accessing external and specialized knowledge to augment language models (LMs). Training dense retrievers typically requires annotated query-document pairs, which are costly to create and scarce in specialized domains (e.g., code) or in complex settings (e.g., requiring reasoning). These practical challenges have sparked growing interest in self-supervised retriever learning. Since LMs are trained to capture token-level dependencies through a self-supervised learning objective (i.e., next token prediction), we can analogously cast retrieval as learning dependencies among chunks of tokens. This analogy naturally leads to the question: How can we adapt self-supervised learning objectives in the spirit of language modeling to train retrievers? To answer this question, we introduce Revela, a unified and scalable training framework for self-supervised retriever learning via language modeling. Revela models semantic dependencies among documents by conditioning next token prediction on local and cross-document context through an in-batch attention mechanism. This attention is weighted by retriever-computed similarity scores, enabling the retriever to be optimized as part of language modeling. We evaluate Revela on domain-specific (CoIR), reasoning-intensive (BRIGHT), and general-domain (BEIR) benchmarks across various retriever backbones. Without annotated or synthetic query-document pairs, Revela surpasses larger supervised models and proprietary APIs on both CoIR and BRIGHT. It achieves BEIR's unsupervised SoTA with ~1000x less training data and 10x less compute. Performance increases with batch size and model size, highlighting Revela's scalability and its promise for self-supervised retriever learning.

cs.IR · cs.CL

新论文 · v1arXiv:2611.20335

When Should Decision Models Compare Candidates?

Fangzhi Li, Quan Shi, Wenting Yang, Yueqiao Chen, Jiayu Chang, Shuting Zhao, Zhaowei Xu, Yizhao Chen 等 10 位作者

原文摘要

Language-model agents frequently select among candidate actions, such as tools and web elements. Should decision models score candidates independently or allow their scores to depend on other candidates? We investigate this question by comparing three approaches: Pointwise scoring, Setwise scoring, and Candidate Attention. Across five agent-decision benchmarks, Pointwise scoring remains competitive, while Setwise scoring performs better on some tool-selection and binary-decision tasks. Candidate Attention does not consistently improve upon Pointwise scoring, despite allowing cross-candidate interaction. Using two controlled synthetic tasks, we further examine whether models learn to use candidate-set information. With targeted relational training, Setwise models learn candidate dependencies and generalize to shifted inputs, while Candidate Attention shows weaker transfer. Our results show that access to other candidates does not guarantee effective comparison, and that its benefits depend on both task requirements and training.

cs.CL · cs.LG

版本更新 · v2arXiv:2601.02236

CD4LM: Consistency Distillation and aDaptive Decoding for Diffusion Language Models

Yihao Liang, Ze Wang, Ximeng Sun, Jiang Liu, Jialian Wu, Hao Chen, Xiaodong Yu, Emad Barsoum 等 10 位作者

原文摘要

Diffusion language models (DLMs) accelerate generation by decoding multiple tokens in parallel, but pushing them to fast decoding often sacrifices accuracy. We attribute this to a static-to-dynamic mismatch: training optimizes local denoising under fixed schedules, while fast inference skips steps and encounters irregular mask patterns. Existing accelerators largely address only one side of this problem. Training-free inference methods speed up decoding without changing the model, leaving the mismatch unresolved; training-side distillation methods adapt the model, but often tie the student to a specific decoding trajectory, limiting composability and still suffering accuracy loss. We propose CD4LM, a composable framework for accurate DLM acceleration. Its core training component, Discrete-Space Consistency Distillation (DSCD) trains a trajectory-invariant student on diverse masked states, reducing the mismatch without fixing a decoding policy. We pair DSCD with a lightweight ConfidenceAdaptive Decoder, while keeping the trained student compatible with external inference-side accelerators. On GSM8K, CD4LM preserves accuracy while achieving a 5.18x speedup; stacking CD4LM with per-step cache accelerators further pushes speedup to 9.73x at teacher-level accuracy, suggesting a scalable path for combining training-side robustness with future decoding advances. Code is available at https://github.com/yihao-liang/CDLM.

cs.CL

版本更新 · v2arXiv:2605.22355

TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation

Hanyu Guo, Jiedong Yang, Chao Chen, Longfei Xu, Kaikui Liu, Xiangxiang Chu

原文摘要

Public transit route planning traditionally depends on structured map infrastructure and complex routing engines, and no existing dataset supports training models to bypass this dependency. We present TransitLM, a large-scale dataset of nearly 13 million transit route planning records from four Chinese cities covering 120,845 stations and 13,666 lines, released as a continual pre-training corpus and benchmark data for three evaluation tasks with complementary metrics. Experiments show that an LLM trained on TransitLM produces structurally valid routes at high accuracy and implicitly grounds arbitrary GPS coordinates to appropriate stations without any explicit mapping. These results demonstrate that transit route planning can be learned entirely from data, enabling end-to-end, map-free route generation directly from origin-destination information. The dataset and benchmark are available at https://huggingface.co/datasets/GD-ML/TransitLM, with evaluation code at https://github.com/HotTricker/TransitLM.

cs.CL · cs.AI · cs.LG

新论文 · v1arXiv:2611.16264

Label-Efficient Document-Grounded Factuality Detection via Cross-Paraphrastic Invariance Learning

Chao Chen, Shanshan Lin, Dongsheng Hong, Sibo Ju, Wenjie Yang, Sihong Xie, Xiangwen Liao

原文摘要

Large language models (LLMs) can generate fluent claims that are unsupported by a given source document, motivating accurate and efficient document-grounded factuality detection. Existing LLM-based evaluators often require repeated inference and incur high computational cost, whereas compact factuality detectors are efficient at inference time but commonly rely on large labeled or synthetically generated corpora. We propose CPIL (Cross-Paraphrastic Invariance Learning), a two-stage Siamese framework that converts a small labeled set into relational supervision without acquiring additional factuality labels. CPIL constructs positive pairs from distinct label-preserving paraphrastic views of the same document-claim instance and negative pairs from opposite-label examples associated with the same source document. It further introduces quantile-aware semi-hard negative mining, which derives a boundary from the positive-pair similarity distribution and balances informative negatives with broader coverage. In the two-stage training procedure, Stage 1 learns an encoder from these task-aware pairs, and Stage 2 jointly fine-tunes the encoder and a linear classification head for binary factuality prediction. At inference time, CPIL requires no paraphrase generation, pair construction, or Siamese comparison and uses only one encoder-classifier forward pass. On the 11-task LLM-AggreFact benchmark, CPIL-2% and CPIL-3% achieve macro-F1 scores of 81.04 and 81.60 while using only 2-3% of the FactCG-released labeled pool. Both settings outperform all compared compact detectors and two vanilla LLM references. Ablations and sensitivity analyses support the contributions of task-aware pair construction, dynamic negative mining, and multi-pivot paraphrasing.

cs.CL · cs.AI

新论文 · v1arXiv:2611.15049

Science Prediction Universe

Zengji Tu, Xiang Liu, Jiarui Sun, Ruoyu Wu, Zhihui Qi, Chengda Wu, Yutao Zhou, Bo Yang 等 36 位作者

原文摘要

Humanity has long sought to explain natural phenomena. Reliable prediction of experimental outcomes is a central test of our understanding of the processes that produce them. If artificial general intelligence (AGI) is to surpass human scientific capabilities, it must also predict experimental outcomes more reliably than humans. Such capabilities could accelerate scientific discovery across disciplines, help address humanity's most pressing challenges, and improve human well-being. We introduce Science Prediction Universe (SPU), a benchmark of hidden experimental outcomes with 100 public verified tasks from chemistry, materials science, biology and medicine, and 16 private tasks from unpublished wet-lab experiments. Given an experimental setup and candidate conditions, an agent works with any resources it needs to rank the candidates by a measured outcome. We evaluate thirteen agent configurations spanning twelve frontier models across two harnesses (Codex and Claude Code). Current agents remain far from saturating wet-lab outcome prediction, with substantial performance disparities across models: on the public benchmark, configurations separate into two distinct tiers, with task-adjusted Spearman correlations ranging from 0.227 for the lowest-performing configuration to 0.478 for the best. Analysis of 1,155 scored public trajectories (227,103 tool calls) suggests that the bottleneck lies in scientific judgment rather than tool execution: agents rely on shared monotone priors, agreeing far more with one another than with experimental reality (0.52 vs. 0.33), and frequently compute trivial physical or chemical proxies that do not influence the final outcome.

cs.AI · cs.CL

新论文 · v1arXiv:2611.14364

ARE-bench:在真实研究者会话与工作区中评测智能体的研究能力

ARE-bench: Benchmarking Agents' Research Ability in Real-World Researcher Sessions and Workspaces

Jiahang Lin, Wanying Ren, Wanghan Xu, Huayu Sha, Yang Shu, Zhonghang Lu, Chengjun Pan, Yi Dong 等 27 位作者

AI 速读 基于摘要

ARE-bench 用真实研究者的会话与未清理工作区构建 75 个任务,检验研究智能体能否在延续项目时遵守先前决策与工作区状态。

cs.CL

查看原文摘要

Research agents achieve increasingly strong benchmark scores, yet these scores do not necessarily translate into reliable delivery in ongoing research projects. Many research-agent benchmarks emphasize standalone tasks and relatively self-contained workflows in clean or curated workspaces. However, researchers ask agents to complete one step within an ongoing project, where requests depend on earlier session decisions and workspaces contain unfinished files, outdated versions, and failed attempts. We introduce ARE-bench to evaluate whether research agents advance ongoing projects while respecting prior decisions and existing workspace state, grounded in real session context and in-progress workspaces. ARE-bench includes 75 tasks built on agent sessions and full workspaces from recruited researchers' own projects, spanning seven research stages, 33 requirements, and 150 sub-requirements, with 1,173 binary checklist items in total. Each task is anchored at a real user request, preserves its requirements, and summarizes the earlier requests still in effect as background. The complete workspace is replayed to that boundary without cleanup, preserving artifacts from earlier turns. Each checklist is compiled from an intent-aligned rubric and credits any valid solution. Across seven frontier models, the strongest passes every checklist item on only 66.2% of tasks, and Design and Execution rank among the hardest stages for every model. Three models pass about 69% of tasks when only delivery contracts and general norms are checked, but just 28.4% to 40.0% once historical and material or workspace requirements are also checked. Evaluating delivery without these contextual requirements can therefore overestimate research agents and mask differences between them.

新论文 · v1arXiv:2611.12392

More Than Just Style: Benchmarking Personality Influence in Large Language Models

Yuhao Liu, Mengyu Wang, Xiao Chen, Lu Liu

原文摘要

In addition to the effects of explicit input prompts and instructions, large language model outputs are influenced by underlying behavioral preferences similar to human personality tendencies. These tendencies can affect how models interpret situations and make decisions. However, their influence is difficult to disentangle from the effects of model capabilities and instructions. Therefore, we introduce SPACE, a benchmark framework combining a controlled evaluation protocol with complementary datasets for personality verification, capability assessment, and hierarchical situational tasks. Controlled comparisons across guidance conditions help distinguish personality influence from measured capability changes and immediate instruction effects. Experimental results demonstrate that although prompting can shift generation style, models' internal personality tendencies are persistent and not readily overridden by opposing instructions. Furthermore, these underlying tendencies have an observable influence on LLM task performance, highlighting the need for greater attention to these characteristics and potential applications. Training code and SPACE benchmark specifications are available at https://github.com/Niklauseik/SPACE-training.

cs.CL

新论文 · v1arXiv:2611.12041

SMO: Compiling Natural Language Skills into Graphs of Reusable State-Move-Outcome Bricks

Chao Chen, Jiashuo Zhang, Yunhang Shen, Liangyu Lian, Rongrong Ji

原文摘要

Large language model agents increasingly rely on reusable natural-language skills, yet growing skill collections remain accumulations of isolated text artifacts that are difficult to compare, deduplicate, and compose. We argue that the reusable content of a skill lies not in its surface wording but in its procedural transitions: which operation to execute under a given state and which observable outcome to produce. Based on this observation, we propose State-Move-Outcome (SMO), a compilation-like framework that decomposes raw skills into reusable procedural bricks and organizes admitted bricks as an externalized symbolic structure. When compiling a skill, SMO first retrieves and composes existing bricks, dynamically connects them through Outcome-to-State conditions, and creates a candidate brick only for an uncovered procedure; the candidate becomes reusable only after validation and admission. Experiments on SkillsBench v1.1 demonstrate that SMO represents 232 rawSkills from 87 underlying tasks with 95 shared bricks, reducing the corresponding skill representations from 366,472 to 149,154 tokens--a $\mathbf{59.3}\times$ reduction and $\mathbf{2.46}\times$ compression. Across 232 skill-conditioned evaluation instances, SMO raises the average task score from $\mathbf{0.6379}$ to $\mathbf{0.7478}$ and increases 5/5 Pass instances from $\mathbf{138}$ to $\mathbf{164}$, while matching or outperforming rawSkill on $\mathbf{208}$ of $\mathbf{232}$ instances. It also reduces average runtime token consumption by $\mathbf{38.38\%}$. These results show that compiling natural-language skills into shared procedural units can jointly improve skill reuse, representation efficiency, and downstream execution.

cs.CL

新论文 · v1arXiv:2611.09770

ROSE:面向多模态大语言模型终身知识编辑的风险感知正交子空间导航

ROSE: Risk-Aware Orthogonal Subspace Navigation for Lifelong Knowledge Editing in Multimodal Large Language Models

Lingyun Song, Ziyao Chen, Kang Pan, Xiaolin Han, Yifei Zhang, Xiaoqi Wang, Yudai Pan, Xuequn Shang

AI 速读 基于摘要

提出ROSE框架,通过将参数更新限制在已编辑子空间的正交补中,并引入风险掩码与多粒度知识融合,以缓解多模态大模型终身知识编辑中的局部性侵蚀与灾难性遗忘。

cs.CL · cs.AI · cs.CV

查看原文摘要

Multimodal Large Language Models (MLLMs) require continuous factual updates to stay current, yet lifelong knowledge editing remains a daunting challenge. The primary obstacle is the severe Locality Erosion and Catastrophic Forgetting, where sequential edits inevitably interfere with pre-trained capabilities and prior updates. In this paper, we propose \textbf{ROSE}, a \textbf{R}isk-aware \textbf{O}rthogonal \textbf{S}ubspace navigation framework for lifelong knowledge \textbf{E}diting. ROSE addresses forgetting by aiming to restrict parameter updates to the orthogonal complement of previously edited subspaces. Crucially, it introduces a risk-guided mechanism that utilizes a dynamically computed risk mask to safeguard specific factual parameter subspaces essential for model stability. During inference, ROSE employs a multi-granularity knowledge integration strategy, featuring prototype-anchored semantic gating and perturbation-driven subspace fusion to precisely activate edited knowledge for relevant queries while limiting interference on unrelated inputs. Comprehensive experiments on two prominent multimodal benchmarks show that ROSE is competitive with state-of-the-art approaches across five key metrics, while maintaining outstanding stability over $1,000$ sequential updates. Our code is available at https://github.com/sunlitsong/ROSE.

版本更新 · v2arXiv:2604.17139

The Consensus Trap: Collaborative Interleaved Generation to Defend against a Corrupted Majority

Jiayuan Liu, Shiyi Du, Weihua Du, Xinhao Cheng, Mingyu Guo, Vincent Conitzer

原文摘要

Majority voting offers a simple way to combine answers from multiple large language model (LLM) agents. When a majority follows instructions favoring the same incorrect answer, however, agreement reinforces the error. We call this failure the "consensus trap": a minority may identify the error, yet its reasoning cannot change a vote over completed answers. We introduce chunk-interleaved Round-Robin (RR) collaboration, in which agents take turns extending one unfinished reasoning trajectory. A minority of agents with correct reasoning can insert corrections that subsequent speakers read before the answer is complete. RRMaj combines several such trajectories by a final vote. We characterize a majority barrier for outcome-level voting over final answer labels and derive a local sufficient condition for expected progress when the minority's cumulative corrections outweigh opposing corrupted updates. Across model families and verifiable reasoning benchmarks, RRMaj substantially improves accuracy under majority corruption while retaining comparable average performance under minority corruption in the main advisory-instruction setting. RR also strengthens full-response debate. Scoring the shared reasoning between turns shows that, on average, continuations from uncorrupted agents increase support for the correct answer over the incorrect target. Parallel local serving adds modest latency in the measured deployment. These findings show how language shared during reasoning, and the timing of that exchange, can help a minority correct a group's answer.

cs.CL · cs.AI · cs.MA

版本更新 · v2arXiv:2610.38076

Prompt Lottery in LLM-as-a-Judge: Measuring and Mitigating Sensitivity to Syntactic Rephrasing

Ilija Subasic, Brett Levert, Andrew Rabinovich, Sajjad Pakdamansavoji, Zhao Chen

原文摘要

LLM-as-a-Judge has become a standard for evaluating large language models, yet most approaches implicitly assume that a single judge prompt functions as a stable evaluation instrument. Prior work has primarily explored deliberate prompt modifications such as scoring rubrics, role descriptions, or chain-of-thought reasoning in order to optimize accuracy. In contrast, we investigate whether validated prompt variants intended to preserve evaluation semantics while differing in surface realization (\textit{paraphrastic rephrasings}) yield consistent judgments under a fixed evaluation task. We introduce \textit{Prompt Lottery}, a phenomenon in which paraphrastic rephrasings produce different LLMaJ judgments. Across a large-scale study spanning diverse evaluation tasks, judge families, and model backbones, we find that these variations can alter evaluation outcomes and, in some settings, reverse relative model rankings. A user study with 10 human annotators shows that LLMaJ judges exhibit 6-9$\times$ larger agreement degradation under prompt rephrasings than humans, falling below the generally accepted threshold in usability studies. This indicates that single-prompt evaluation is insufficiently robust for model ranking. Finally, we show that multi-prompt aggregation substantially improves robustness by emphasizing judgments that remain consistent across prompt variants.

cs.CL · cs.AI

新论文 · v1arXiv:2611.08219

Agent-R²:通过真实来源数据合成扩展智能体以完成现实生产力工作

Agent-R$^2$: Scaling Agents for Realistic Productivity Work via Real-Source Data Synthesis

Teng Pan, Zixuan Wang, Yuchen Yan, Qifei Wu, Yao Chen, Yingqi Qu, Ruiqing Zhang, Yongliang Shen

AI 速读 基于摘要

提出基于真实职业种子、工作流证据与真实网页文件的数据合成框架 Agent-R²,并训练 4B 智能体以提升现实生产力任务表现。

cs.CL

查看原文摘要

Large language models have made rapid progress in agentic reasoning and tool use, raising expectations for agents that can take over realistic productivity work from humans. However, current agents still struggle to finish such tasks, where success requires following long-horizon professional workflows, understanding heterogeneous file contexts, and producing usable deliverables under fine-grained requirements. A key bottleneck is the lack of scalable post-training data that reflects real work intent, authentic file environments, and deliverable-level evaluation. To close this gap, we introduce Agent-R$^2$, a $\textbf{r}$eal-source data synthesis framework that scales training data for $\textbf{r}$ealistic productivity work. Instead of relying on manually specified workflows or LLM-generated files, Agent-R$^2$ grounds synthesis throughout the pipeline in real occupational seeds, corpus-derived workflow evidence, and authentic web-sourced files. From 10k web-corpus candidates, Agent-R$^2$ distills 4k task instances that pair real multi-format attachments with fine-grained, deliverable-level rubrics. On this data, we train Agent-R$^2$-4B via supervised fine-tuning and rubric-based reinforcement learning, improving both end-to-end productivity performance and file-grounded reasoning. Trained on only 4k instances, Agent-R$^2$-4B surpasses Qwen3-235B-A22B-Thinking-2507 on both GDPval (51.0$\%$ vs. 48.0$\%$) and OfficeQA (14.8$\%$ vs. 13.8$\%$). Further analysis shows that Agent-R$^2$ mitigates shortcut behavior, where agents produce superficially plausible deliverables before sufficiently grounding, constructing, or verifying them, while promoting sustained exploration.

新论文 · v1arXiv:2611.06043

Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription

Daniel Huang, Qi He, Shaoxiong Duan, Yingjie Li, Qiao Cheng, Yitao Duan

原文摘要

Streaming speech recognition is increasingly consumed by machines, such as translators and voice agents, that act on each token as it appears. A token that is later revised costs wasted work or a wrong action. In this work, we present Confucius4-R2T2 (short for Real Real-Time Transcription), a stable-prefix streaming ASR model that never modifies its earlier output while maintaining high accuracy and low latency. R2T2 is obtained by post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio while freezing its pre-trained audio encoder and projector. The adaptation is based on the concept of Longest Stable Prefix (LSP), a new notion that we introduce in this work. We prove that LSP is monotone under the prefix order, consistent with the full-audio output, and maximal among labels that equal a whole earlier output. Learning from LSP labels converts a full-context model into a stable-prefix streaming one without any architectural change. Extensive evaluations show that R2T2 brings streaming recognition close to the base model's full-context quality while delivering stable text with low emission lag. Its gains extend beyond the post-training languages, demonstrating a practical path from full-context ASR to stable streaming. We open-source the model weights and code at https://huggingface.co/netease-youdao/Confucius4-R2T2 and https://github.com/netease-youdao/Confucius4-R2T2.

cs.CL · cs.AI · cs.SD

新论文 · v1arXiv:2611.05804

Matching the Training-Time Weight Rule in Inference-Time Expert Deactivation

Hao Cheng, Zhiyu Yang, Bo Ma, Hanzhang Jia

原文摘要

Expert deactivation at inference time changes both token routing and expert weights, and its effect on performance depends on how these changes are implemented. We compare removal, freezing, remaining-expert renormalization, and unnormalized re-routing on identical disabled sets across five instruction-tuned checkpoints from four model families, with a base OLMoE checkpoint as an additional comparison. We develop a weight-sum model that relates the loss of expert contributions and changes in branch scale to the weight rule used during training. With half of the experts deactivated according to activation-based rankings, remaining-expert renormalization yields the highest perplexity in models trained with raw router probabilities, whereas unnormalized re-routing yields the highest perplexity in models trained with normalized weights. Perplexity ranges from 1.4 to 73 times the intact baseline when the training rule is retained, and from 14 to 28,740 times the baseline under a rule mismatch. Task outcomes vary across evaluation types: on OLMoE, GSM8K strict-match accuracy falls to zero under all four implementations. Extending the model to changes in the number of active experts yields a further prediction, supported by the OLMoE experiments: the perplexity penalty from additional renormalization increases as fewer experts are activated.

cs.CL · cs.AI

新论文 · v1arXiv:2611.05759

PAYWORTHY: Can LLMs Predict Which Freelance Jobs an AI Agent Will Complete?

Mohammad Reza Rezaei, Dingxian Wang, Brett Levert, Celeste Fejzo, Andrew Rabinovich, Zhao Chen

原文摘要

Before an AI agent takes on a paid job, someone has to judge whether it is likely to succeed. We test whether frontier language models can make this call before any attempt. We propose PAYWORTHY which contains 1{,}494 real freelance job postings, sampled from 18.8M postings made between 2020 and 2026 on a freelancing platform. An agent attempts each job from the posting alone in a single run, and an LLM judge acting as the client decides whether to accept the work. We ask 18 models to forecast how likely the agent is to succeed and measure how well each model's forecasts rank the accepted jobs above the rejected ones (AUC, where 0.5 is a random ranking). The answer depends on how the agent is described. Told what the agent cannot do, every model ranks the jobs well (AUC 0.66--0.86). Told instead that it is a generic agent with a browser and code execution, the same models rank the same jobs about as well as a random ordering (0.45--0.56). These results hold under judges from four other model families (Claude, GLM, Kimi and Qwen), and the informed forecasts also rank the verdicts of two human raters. Code is available at https://github.com/upwork/Payworthy

cs.AI · cs.CL

新论文 · v1arXiv:2611.04687

CardioVascBench:Prioritized Diagnosis Generation from Real-World Chinese Cardiovascular Records

Minghao Kong, Muyao Chen, Jianhao Wu, Yuqiang Zhang, Xiangbin Meng, Zirong Liu, Ting Cao, Kaida Ning 等 10 位作者

原文摘要

Identifying the primary diagnosis alone does not fully characterize a model's diagnosis-generation performance. We introduce CardioVascBench, a retrospective benchmark for prioritized diagnosis generation from real-world Chinese cardiovascular inpatient records. It contains 3,262 de-identified cases, with 2,174 for training and 1,088 for evaluation, linking clinical narratives and textual ECG reports to expert-verified, ordered discharge diagnoses standardized to ICD-10. An English translation of the evaluation set supports aligned cross-lingual comparison. Our evaluation protocol combines terminology normalization and rank-aware scoring to separately assess primary-diagnosis identification, reference-diagnosis coverage, and list precision. Evaluation across 36 systems shows that higher primary-diagnosis accuracy does not necessarily imply higher list precision against discharge references. We also provide CardioKG-Agent, a knowledge-enhanced baseline combining entity normalization, constrained graph retrieval, and guideline evidence. In a staged comparison of three baseline models, supervised fine-tuning plus the agent achieves a 73.5% higher mean composite score than fine-tuning alone, representing a system-level gain. These results measure retrospective agreement with recorded diagnoses; agreement between metric rankings and physician judgments remains to be validated. CardioVascBench supports research on diagnosis prioritization and knowledge-enhanced generation.

cs.CL

新论文 · v1arXiv:2611.04079

AlgoPilot: Cross-Paradigm Reasoning in Language Models via Strategy Selection and Guidance

Yilun Hao, Ruixiao Yang, Hudson Hilal, Yongchao Chen, Kaizhi Qian, Yada Zhu, Yang Zhang, Chuchu Fan

原文摘要

Real-world problems span diverse domains, including program synthesis, symbolic reasoning, planning, and optimization, and often require fundamentally different solution paradigms. A central challenge for LLM-based reasoning is twofold: identifying an appropriate problem-solving approach and formulating it correctly. In this paper, we propose AlgoPilot, a cross-paradigm reasoning framework that enables adaptive selection of the solving approaches and structured problem formulation prior to execution. The AlgoPilot framework contains three components: 1) a SteerLM, trained via supervised fine-tuning (SFT) and reinforcement learning (RL), that adaptively selects the solving strategy (paradigm and modes) based on input problems, 2) the corresponding expert GuideLM, trained via SFT, that generates structured formulation guidance tailored to the selected strategy, and 3) an ExecutionLM that generates solutions conditioned on both the selected strategy and its formulation. Unlike prompting-based methods that rely on fixed reasoning templates, our approach learns to adaptively choose and structure problem-solving strategies based on the input. We build AlgoPilot on Qwen3-8B and evaluate across 5 problem categories and 13 benchmarks with 66 unique domains, comparing against both the latest similarly scaled small models and large LLMs. AlgoPilot significantly improves average accuracy for the same model from 38.4% to 72.9%, outperforming the best large LLM prompting baseline GLM-5 by 11.7%. Ablation studies confirm the effectiveness of each component. Notably, we show that the learned steering and formulation modules can transfer across execution models, suggesting their potential to improve other LLMs. Project page: https://sites.google.com/view/algopilot.

cs.AI · cs.CL · cs.LG

新论文 · v1arXiv:2611.03634

WebHarbor: Docking Real Websites for Evolving GUI Agent Environments

Zhaoyang Wang, Qianhui Wu, Boyu Gou, Demi Wang, Jiacheng Zhang, Peichun Hua, Qize Li, Shi Qiu 等 57 位作者

原文摘要

GUI web agents require realistic and interactive environments that remain stable during both evaluation and training. Live website evaluation suffers from changing content, access restrictions, and limited coverage of authenticated workflows, while existing synthetic environments for training often lack functional depth and visual fidelity. We propose WebHarbor, a suite of 153 executable and dockerized website environments that faithfully reconstruct the visual interfaces, content, and functional workflows of real-world web services under deterministic control. We build WebHarbor through task-driven human-agent collaboration, where environments evolve as new tasks expand their functional scope, coding agents implement the required changes, and human contributors audit visual fidelity, functional correctness, and task validity. The resulting suite comprises 3,518 tasks spanning multi-step and post-authentication workflows, where existing works often fail to cover. Beyond reliable evaluation, WebHarbor provides high-fidelity synthetic web environments for cost-efficient reinforcement learning (RL). We demonstrate that a Qwen 4B-level GUI agent trained with RL in WebHarbor generalizes to unseen tasks and performs competitively with leading open-source GUI web agents across four benchmarks. The code and data are publicly available at https://github.com/aiming-lab/WebHarbor.

cs.AI · cs.CL · cs.LG

新论文 · v1arXiv:2611.03496

Online Label-Free Gradient Matching for Continual Language Model Training

Sajjad Pakdamansavoji, Mohammad Reza Rezaei, Ilija Subasic, Pedro Garzon, Radu Paradovschi, Zhao Chen

原文摘要

Language models in customer-facing products must be continually updated, since the data they serve shifts over time. As retraining on the full accumulated data is impractical, each update should instead run on a small set of examples carrying the training signal of all data seen so far. The strongest current methods build this set by gradient matching, which synthesizes examples whose training gradients match those of the full data. However, they require ground-truth labels, run once on a fixed snapshot rather than on a growing stream, and are slow on large data. We introduce \textbf{OLGA}, which lifts these limitations. It distills unlabeled data by drawing the gradient matching targets from the model's own next-token predictions. Across five text classification benchmarks with phi-1.5, it improves accuracy by 10 percentage points over the strongest selection baseline using as few as five examples. OLGA also distills online data streams by keeping a running rehearsal memory of synthetic examples, which is re-distilled together with each new batch as it arrives. In the online setting, it outperforms the strongest selection-based continual baseline by 6.6 stream-AUC percentage points. Finally, our method scales to large data by pairing distillation with selection in a hybrid pipeline, letting selection handle the mild stage of compression and distillation the extreme one. It cuts runtime by 30\% versus distillation alone at similar accuracy, and beats the best selection-only pipeline by up to 12.6 percentage points.

cs.AI · cs.CL

版本更新 · v2arXiv:2610.34103

T2D: Target-to-Draft Model Pruning & Adaptation for Efficient Speculative Decoding

Shahriar Kabir Nahin, Xiao Ye, Muhao Chen, Ben Zhou, Anshuman Chhabra

原文摘要

Speculative decoding significantly accelerates autoregressive text generation in Large Language Models (LLMs), but its efficacy depends largely on two key factors: the distributional alignment between the draft and the target model, and the draft model's latency during inference. However, existing approaches typically utilize independently pretrained small draft models or seek to prune the target model to a manually chosen architecture, thereby failing to explicitly balance the trade-off between draft model alignment and token generation efficiency. To address this gap, we formulate draft model design as an optimization problem and propose T2D, a framework that constructs draft models directly from the target model using latency-aware structured pruning and knowledge distillation. In this manner, T2D obtains draft models that are jointly optimized for both measured inference latency and cross-entropy loss on the target model's responses, resulting in superlative speculative decoding performance. Through extensive experiments on multi-domain speculative decoding benchmarks and across several target model families of differing scale, we demonstrate how T2D prunes up to 92\% of the target model to obtain highly performant draft models compared to competitive baselines. Moreover, draft models obtained via T2D can also be utilized by other speculative decoding token-alignment and knowledge distillation methods in a plug-and-play fashion, leading to further gains. Our findings underscore the importance of jointly navigating the draft model architecture-alignment frontier to unlock the full potential of speculative decoding for LLMs.

cs.CL · cs.AI · cs.LG

新论文 · v1arXiv:2610.40541

论基于预填充的解码长度预测的根本极限

On the Fundamental Limits of Prefill-Based Decoding Length Prediction

Haowen Shi, Xuzhe Yan, Fahao Chen, Peng Li

AI 速读 基于摘要

作者指出基于预填充阶段隐状态预测LLM解码长度的精度受自回归解码固有随机性限制,并给出理论界与基准评测建议。

cs.AI · cs.CL · cs.IT

查看原文摘要

Predicting how many tokens a large language model (LLM) will decode, using only prefill-phase hidden states, is a key primitive for inference scheduling. We show that current benchmarks conflate predictor quality with task hardness: when the target model, dataset, or sampling configuration changes, reported accuracy collapses even though the predictor is unchanged. We trace this instability to an overlooked error source---the inherent stochasticity of autoregressive decoding---and derive bounds that every prefill-side method must obey. Our analysis treats decoding as a stopped stochastic process. The per-step stopping probability of a response is not constant but ramps up over hundreds of steps, so the decode length inherits a central limit theorem from the many local decisions that determine where a response ends; this yields per-prompt normality with an explicit error term, together with an affine variance law $σ=αμ+b$ whose slope is the coefficient of variation of the number of response segments. Empirically, the normal law reproduces the per-prompt MAE and the central mass of the decode-length distribution to within a few percent, and the residual departures from normality (mild right skew, slight excess kurtosis) are the ones predicted by the sampling-noise correction. Under fixed model, data, and sampling conditions, a broad design space of predictors (MLP, EGTP, FiLM, soft-label, and Mixture-of-Experts variants) converges to similar performance close to the bound, confirming that \emph{system error dominates model error}. We argue for a benchmarking protocol that fixes the model, data, and sampling parameters and that reports relative-error metrics alongside absolute ones. Code and analysis tools are available at https://github.com/ShwStone/DecodingPredict.

版本更新 · v3arXiv:2605.24279

ContextEcho:长程智能体编程会话中人格漂移的基准

ContextEcho: A Benchmark for Persona Drift in Long Agentic-Coding Sessions

Xianzhong Ding, Yangyang Yu, Changwei Liu, Bill Zhao, Le Chen, Tao Chen

AI 速读 基于摘要

作者提出 ContextEcho 基准与可复用测试框架,用于刻画长程智能体编程会话中编程助手人格的漂移现象。

cs.CL · cs.SE

查看原文摘要

Frontier language models are trained to exhibit stable programming assistant personas, yet those personas may drift over long agentic coding sessions in production. After a long real-world coding session, a model that hedges preferences in a fresh chat ("I don't have preferences") may instead express them directly ("Python -- the feedback loop is instant..."). Existing persona-stability studies focus on scripted chat of tens of turns, leaving long production coding sessions involving thousands of tool-using agent steps, repeated compaction, and hours of interaction largely uncharacterized. We introduce ContextEcho, a benchmark and reusable harness with a 25-probe identity suite, a snapshot-then-probe protocol that evaluates conversation state without perturbing the main session, judge-based and judge-free metrics, and a corpus of 57 long agentic-coding sessions collected with consent and redacted for release across Claude Code and Codex CLI. We use one representative session for a 23-model cross-provider study and all 57 sessions for corpus-scale trajectory analysis across frontier and open-weight models. Results show that persona drift occurs across providers and emerges early in real agentic interaction, while in-session compaction neither reliably resets nor amplifies it. A single-shot anchor restores the original assistant persona, though placebo controls suggest a generic register-reset effect. On same-task continuations, drifted prefixes improve tool-call fidelity -- which we attribute to retained task context rather than drift itself. In tool-free chat, however, the same prefixes break output-format contracts and inflate output length. These findings motivate ContextEcho as an open-source framework for auditing persona drift in long coding sessions without retraining.

新论文 · v1arXiv:2610.39086

HALO: Human-Grounded Granular Evaluation of Proactive Agents

Wenqi Pei, Xinyu Wang, Jinrui Liu, Cynthia Zhuge, Jie Sheng, Zihan Wang, Tenghao Huang, Muhao Chen

原文摘要

Proactive agents must decide whether, when, and how far to assist without an explicit request. Existing evaluations often rely on automated scores or binary verdicts, offering limited human diagnosis of evidence use and intervention quality. To address these gaps, we introduce HALO, Human-Grounded GrAnular EvaLuation of PrOactive Agents. HALO-Arena covers 4 proactive need types, Dependency, Blockage, Risk, and Opportunity, across 10 everyday domains and 50 personal environments. Scenarios require agents to retrieve evidence from structured databases or unstructured documents before judging whether assistance is warranted and choosing its scope. On HALO-Arena, we construct HALO-8K, a dataset of 8,000 responses from 10 models across 400 scenarios, annotated by 20 experts for action level, acceptability verdicts, and human diagnosis. For rejected responses, human diagnosis covers timing, intensity, and other issues. Recorded reasoning tokens and retrieval rounds measure the effort behind these responses. The best proprietary and open-source models achieve 72.88% and 56.63% acceptability, respectively. Our central finding is that over-proactivity is the dominant observed failure mode, with Premature timing or Excessive intensity accounting for 61.20% of rejected responses. Further analyses yield complementary findings on the balance between intervention and restraint, evidence acquisition, and the calibration of timing and intensity.

cs.CL

新论文 · v1arXiv:2610.38204

Two Birds with One Stone: Skill-Augmented Generation via Unified Retrieval and Utilization

Zihan Luo, Jiahui Kang, Wei Zhou, Qilin Fan, Mingyan Li, Chao Chen

原文摘要

Skills, as reusable external knowledge, have demonstrated substantial potential to enhance the ability of Large Language Models (LLMs) to solve complex tasks. However, existing skill-augmented generation methods typically rely on separate models for skill retrieval and skill utilization, resulting in disconnected optimization of these two processes. To address this gap, we propose Skill-Uni, a unified reinforcement learning framework that jointly optimizes skill retrieval and utilization. Specifically, we adopt a generative skill retrieval paradigm that enables a single model to both learn how to retrieve relevant skills and leverage them for task solving. To effectively train this unified framework, we introduce a two-stage training strategy consisting of warm-up instruction tuning and GRPO, allowing the model to progressively improve its skill retrieval and utilization capabilities. Extensive experiments on five skill-related benchmarks demonstrate the effectiveness of Skill-Uni, yielding average performance gains of 66.59% and 34.82% over the base Qwen3-4B and LLaMA3.1-8B models, respectively. Besides, Skill-Uni is compatible with various skill retrieval engines, demonstrating its flexibility and generalizability. Additional in-depth analyses also provide insights into the effectiveness of Skill-Uni.

cs.CL

版本更新 · v2arXiv:2608.17843

Decodable but Not Actionable: Auditing the Representation-Behavior Gap in Frozen LLMs for Geometric Constraints

Man Liang, Xinzhao Cheng, Faizan Wajid

原文摘要

Large language models (LLMs) perform well on structured reasoning tasks, but what they encode and whether encoded information guides their behavior remain unclear. We introduce a controlled auditing framework that uses parametric CAD to distinguish failures of encoding, expression, and activation-level control. Applied to eight pretrained models and matched randomly initialized controls, our framework reveals that pretraining contributes substantially more to linear decoding of local relations than to global constraint status, which is already highly decodable from random representations. Yet strong local decodability does not translate into the models' zero-shot predictions, which remain far below supervised probe performance. Restoring clean internal activations can repair probe predictions, whereas simple activation steering does not provide reliable targeted control. With only 400 labeled relation pairs, low-rank fine-tuning substantially improves generation on three representative models, showing that the representation--behavior gap can be reduced through adaptation. Together, these findings show why decodability alone is insufficient evidence that encoded information will be expressed or controllable in model behavior.

cs.CL · cs.AI

版本更新 · v4arXiv:2512.13399

Dual-Evolving Reinforcement Learning for Automated Reward Design

Sitao Cheng, Tianle Li, Xuhan Huang, Xunjian Yin, Difan Zou

原文摘要

Crafting effective reward signals remains a central challenge in reinforcement learning (RL) for complex reasoning. Existing methods for automated reward optimization search with a fixed proposer which never learns from its costly trials. We present Dual-Evolving Reinforcement Learning (DERL), a dual-training framework for autonomous discovery of reward structures. An outer-loop proposer composes a reward from task primitives, and an inner policy trains with this reward. The proposer itself learns by policy gradient from the validation performance of the inner policy. Thus, DERL utilizes past-trial feedback to shape its reward-proposal distribution, an empirical meta-gradient of task success. We validate DERL on embodied (ALFWorld), scientific (ScienceWorld), and mathematical (GSM8K, MATH) reasoning. On multi-turn agent benchmarks, DERL outperforms strong baselines, significantly on out-of-distribution tasks. In mathematics, DERL improves on the outcome reward where directly added primitives hurt. Further analyses under matched compute show that our proposer effectively captures the meta-gradient to discover useful reward signals.

cs.AI · cs.CL