大模型推理能力

PyBangla at BLP-2025 Task 2: Enhancing Bangla-to-Python Code Generation with Iterative Self-Correction and Multilingual Agents

在BLP-2025任务2中提升孟加拉语到Python代码生成：通过迭代自我纠正和多语言代理

Jahidul Islam, Md Ataullha, Saiful Azad

机构 * Department of Computer Science and Engineering（计算机科学与工程系） 纠错

专题命中 复杂问题求解 ：self-correction(title,abstract)；reasoning(abstract)；分类 cs.CL、cs.AI

AI总结 本文提出BanglaCodeAct框架，通过多代理提示和迭代自我纠正，提升孟加拉语到Python代码生成的性能，实验显示Qwen3-8B结合该框架在开发集和盲测集上分别达到94.0%和71.6%的准确率。

Comments 6 Pages

Route-and-Reason: Scaling Large Language Model Reasoning with Reinforced Model Router

路由与推理：通过强化模型路由扩展大语言模型推理

Chenyang Shao, Xinyang Liu, Yutang Lin, Fengli Xu, Yong Li

机构 * Department of Electronic Engineering, Tsinghua University（清华大学电子工程系） ； BNRist, Tsinghua University（清华大学BNRist） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

AI总结 R2-Reasoner通过强化模型路由器实现高效的大语言模型推理扩展，减少API成本并保持推理准确性。

Advancing Multi-Step Mathematical Reasoning in Large Language Models through Multi-Layered Self-Reflection with Auto-Prompting

通过多层自反思与自动提示推进大语言模型的多步数学推理

André de Souza Loureiro, Jorge Valverde-Rebaza, Julieta Noguez, David Escarcega, Ricardo Marcacini

机构 * Luiz de Queiroz College of Agriculture University of São Paulo（圣保罗大学农业学院） ； School of Engineering and Sciences Tecnologico de Monterrey（蒙特雷技术学院工程与科学学院） ； Institute of Mathematics and Computer Sciences University of São Paulo（圣保罗大学数学与计算机科学研究所） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；CoT(abstract)；分类 cs.CL、cs.LG

AI总结 本文提出MAPS框架，通过多层自反思与自动提示提升大语言模型的多步数学推理能力，实验证明其在多个基准测试中优于标准CoT并接近优化模型。

Comments Accepted for publication in: European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD 2025). Research Track

Synthetic Error Injection Fails to Elicit Self-Correction In Language Models

合成错误注入无法在语言模型中引发自我修正

David X. Wu, Shreyas Kapur, Anant Sahai, Stuart Russell

机构 * ucb（伯克利大学） 纠错

专题命中 复杂问题求解 ：self-correction(title,abstract)；reasoning(abstract)；分类 cs.AI、cs.LG

AI总结 本研究发现合成错误注入无法有效提升语言模型的自我修正能力，揭示了策略性强化学习在激发自我修正方面的独特优势。

Comments 13 pages, 12 figures

You Need Reasoning to Learn Reasoning: The Limitations of Label-Free RL in Weak Base Models

Shuvendu Roy, Hossein Hajimirsadeghi, Mengyao Zhai, Golnoosh Samei

机构 * Sea AI Lab（海思人工智能实验室） ； National University of Singapore（新加坡国立大学） ； Singapore Management University（新加坡管理学院） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.AI、cs.LG

Comments 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop: MATH-AI

Pentest-R1: Towards Autonomous Penetration Testing Reasoning Optimized via Two-Stage Reinforcement Learning

He Kong, Die Hu, Jingguo Ge, Liangxiong Li, Hui Li, Tong Li

机构 * State Key Laboratory of Cyberspace Security Defense, Institute of Information Engineering, Chinese Academy of Sciences（中国科学院网络空间安全防御重点实验室，信息工程研究所） ； School of Cyber Security, University of Chinese Academy of Sciences（中国科学院大学网络与安全学院） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；self-correction(abstract)；分类 cs.AI、cs.LG

Illusions of reflection: open-ended task reveals systematic failures in Large Language Models' reflective reasoning

Sion Weatherhead, Flora Salim, Aaron Belbasis

机构 * University of New South Wales Sydney（新南威尔士大学悉尼分校） ； Aurecon Group（Aurecon集团） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；self-correction(abstract)；分类 cs.AI、cs.LG

Comments Currently under review

ReasoningShield: Safety Detection over Reasoning Traces of Large Reasoning Models

Changyi Li, Jiayi Wang, Xudong Pan, Geng Hong, Min Yang

机构 * Fudan University（复旦大学） ； Shanghai Innovation Institute（上海创新研究院） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；CoT(abstract)；分类 cs.CL、cs.AI

Hound: Relation-First Knowledge Graphs for Complex-System Reasoning in Security Audits

Bernhard Mueller

机构 * Bernhard Mueller（伯纳德·穆勒） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；planning(abstract)；分类 cs.AI、cs.LG

Chain or tree? Re-evaluating complex reasoning from the perspective of a matrix of thought

Fengxiao Tang, Yufeng Li, Zongzong Wu, Ming Zhao

机构 * School of Computer,Science and Engineering（计算机科学与工程学院） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；CoT(abstract)；分类 cs.CL、cs.AI

A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models

Yanbo Wang, Yongcan Yu, Jian Liang, Ran He

机构 * School of Artificial Intelligence, University of Chinese Academy of Sciences（人工智能学院，中国科学院大学） ； NLPR & MAIS, Institute of Automation, Chinese Academy of Sciences（神经语言处理与机器智能中心，中国科学院自动化研究所） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；CoT(abstract)；分类 cs.CL、cs.AI

Comments 38 pages. This survey considers papers published up to June 30, 2025. Work in progress

Implicit Reasoning in Large Language Models: A Comprehensive Survey

Jindong Li, Yali Fu, Li Fan, Jiahong Liu, Yao Shu, Chengwei Qin, Menglin Yang, Irwin King, Rex Ying

机构 * Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州）） ； Jilin University（吉林大学） ； The Chinese University of Hong Kong（香港中文大学） ； Yale University（耶鲁大学） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

XFinBench: Benchmarking LLMs in Complex Financial Problem Solving and Reasoning

Zhihan Zhang, Yixin Cao, Lizi Liao

机构 * School of Computing and Information Systems, Singapore Management University（计算与信息系统学院，新加坡管理大学） ； Institute of Trustworthy Embodied AI, Fudan University（可信具身人工智能研究院，复旦大学） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；planning(abstract)；分类 cs.CL、cs.LG

Pragmatic Inference Chain (PIC) Improving LLMs' Reasoning of Authentic Implicit Toxic Language

Xi Chen, Shuo Wang

机构 * Nanyang Technological University（南洋理工大学） ； University of Macau（澳门大学） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；CoT(abstract)；分类 cs.CL、cs.AI

Comments 14 pages, 4 figures, 2 tables

Accelerating LLM Reasoning via Early Rejection with Partial Reward Modeling

Seyyed Saeid Cheshmi, Azal Ahmad Khan, Xinran Wang, Zirui Liu, Ali Anwar

机构 * University of Minnesota（明尼苏达大学） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；math reasoning(abstract)；分类 cs.AI、cs.LG

RLVMR: Reinforcement Learning with Verifiable Meta-Reasoning Rewards for Robust Long-Horizon Agents

Zijing Zhang, Ziyang Chen, Mingxiao Li, Zhaopeng Tu, Xiaolong Li

专题命中 复杂问题求解 ：reasoning(title,abstract)；planning(abstract)；分类 cs.AI、cs.LG

Reasoning about Uncertainty: Do Reasoning Models Know When They Don't Know?

Zhiting Mei, Christina Zhang, Tenny Yin, Justin Lidard, Ola Shorinwa, Anirudha Majumdar

机构 * Princeton University（普林斯顿大学） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

EmbRACE-3K: Embodied Reasoning and Action in Complex Environments

Mingxian Lin, Wei Huang, Yitang Li, Chengjie Jiang, Kui Wu, Fangwei Zhong, Shengju Qian, Xin Wang, Xiaojuan Qi

专题命中 复杂问题求解 ：reasoning(title,abstract)；planning(abstract)；分类 cs.CL、cs.AI

Comments Project page: https://mxllc.github.io/EmbRACE-3K/

PLAN-TUNING: Post-Training Language Models to Learn Step-by-Step Planning for Complex Problem Solving

Mihir Parmar, Palash Goyal, Xin Liu, Yiwen Song, Mingyang Ling, Chitta Baral, Hamid Palangi, Tomas Pfister

专题命中 复杂问题求解 ：planning(title,abstract)；reasoning(abstract)；分类 cs.CL、cs.AI

Comments 15 Pages

Do Thinking Tokens Help or Trap? Towards More Efficient Large Reasoning Model

Bowen Ding, Yuhan Chen, Futing Wang, Lingfeng Ming, Tao Lin

机构 * Zhejiang University（浙江大学） ； Boston University（波士顿大学） ； ByteDance（字节跳动） ； School of Engineering, Westlake University（西溪大学工程学院） ； Research Center for Industries of the Future, Westlake University（西溪大学未来产业研究中心） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；math reasoning(abstract)；分类 cs.CL、cs.AI

Comments 13 pages, 5 figures

Data Shifts Hurt CoT: A Theoretical Study

Lang Yin, Debangshu Banerjee, Gagandeep Singh

机构 * Department of Computer Science University of Illinois Urbana-Champaign（计算机科学系伊利诺伊大学厄巴纳-香槟分校） 纠错

专题命中 复杂问题求解 ：CoT(title,abstract)；reasoning(abstract)；分类 cs.AI、cs.LG

Comments Comparison to v1: upgraded the quality of a figure

SafeCoT: Improving VLM Safety with Minimal Reasoning

Jiachen Ma, Zhanhui Zhou, Chao Yang, Chaochao Lu

机构 * Shanghai AI Laboratory（上海人工智能实验室） ； Fudan University（复旦大学） 纠错

专题命中 复杂问题求解 ：reasoning(title)；chain-of-thought(abstract)；CoT(abstract)；分类 cs.AI、cs.LG

Graph Counselor: Adaptive Graph Exploration via Multi-Agent Synergy to Enhance LLM Reasoning

Junqi Gao, Xiang Zou, YIng Ai, Dong Li, Yichen Niu, Biqing Qi, Jianxing Liu

机构 * Shanghai Artificial Intelligence Laboratory（上海人工智能实验室） ； School of Mathematics, Harbin Institute of Technology（哈尔滨工业大学数学学院） ； Department of Control Science and Engineering, Harbin Institute of Technology（哈尔滨工业大学控制科学与工程学院） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；planning(abstract)；分类 cs.CL、cs.AI

Comments Accepted by ACL 2025

Auto-Patching: Enhancing Multi-Hop Reasoning in Language Models

Aviv Jan, Dean Tahory, Omer Talmi, Omar Abo Mokh

机构 * Tel Aviv University（特拉维夫大学） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.LG

Comments 8 pages, 5 figures

Cosmos-Reason1: From Physical Common Sense To Embodied Reasoning

NVIDIA, :, Alisson Azzolini, Junjie Bai, Hannah Brandon, Jiaxin Cao, Prithvijit Chattopadhyay, Huayu Chen, Jinju Chu, Yin Cui, Jenna Diamond, Yifan Ding, Liang Feng, Francesco Ferroni, Rama Govindaraju, Jinwei Gu, Siddharth Gururani, Imad El Hanafi, Zekun Hao, Jacob Huffman, Jingyi Jin, Brendan Johnson, Rizwan Khan, George Kurian, Elena Lantz, Nayeon Lee, Zhaoshuo Li, Xuan Li, Maosheng Liao, Tsung-Yi Lin, Yen-Chen Lin, Ming-Yu Liu, Xiangyu Lu, Alice Luo, Andrew Mathau, Yun Ni, Lindsey Pavao, Wei Ping, David W. Romero, Misha Smelyanskiy, Shuran Song, Lyne Tchapmi, Andrew Z. Wang, Boxin Wang, Haoxiang Wang, Fangyin Wei, Jiashu Xu, Yao Xu, Dinghao Yang, Xiaodong Yang, Zhuolin Yang, Jingxu Zhang, Xiaohui Zeng, Zhe Zhang

机构 * NVIDIA 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.AI、cs.LG

Chain-of-Thought Tokens are Computer Program Variables

Fangwei Zhu, Peiyi Wang, Zhifang Sui

机构 * School of Computer Science, State Key Laboratory of Multimedia Information Processing, Peking University（计算机科学学院、多媒体信息处理国家重点实验室、北京大学） 纠错

专题命中 复杂问题求解 ：chain-of-thought(title)；reasoning(abstract)；CoT(abstract)；分类 cs.CL、cs.AI

Recursive Decomposition with Dependencies for Generic Divide-and-Conquer Reasoning

Sergio Hernández-Gutiérrez, Minttu Alakuijala, Alexander V. Nikitin, Pekka Marttinen

机构 * Department of Computer Science Aalto University（计算机科学系 阿尔托大学） 纠错

专题命中 复杂问题求解 ：reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.AI、cs.LG

Phi-4-reasoning Technical Report

Marah Abdin, Sahaj Agarwal, Ahmed Awadallah, Vidhisha Balachandran, Harkirat Behl, Lingjiao Chen, Gustavo de Rosa, Suriya Gunasekar, Mojan Javaheripi, Neel Joshi, Piero Kauffmann, Yash Lara, Caio César Teodoro Mendes, Arindam Mitra, Besmira Nushi, Dimitris Papailiopoulos, Olli Saarikivi, Shital Shah, Vaishnavi Shrivastava, Vibhav Vineet, Yue Wu, Safoora Yousefi, Guoqing Zheng

专题命中 复杂问题求解 ：reasoning(title,abstract)；planning(abstract)；分类 cs.CL、cs.AI

Reasoning without Regret

Tarun Chitra

专题命中 复杂问题求解 ：reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.AI、cs.LG

SafeMLRM: Demystifying Safety in Multi-modal Large Reasoning Models

Junfeng Fang, Yukai Wang, Ruipeng Wang, Zijun Yao, Kun Wang, An Zhang, Xiang Wang, Tat-Seng Chua

专题命中 复杂问题求解 ：reasoning(title,abstract)；self-correction(abstract)；分类 cs.AI、cs.LG