大模型推理能力 - arXivDaily 专题

2510.01116 2026-03-17 cs.LG 89%

Eliciting Chain-of-Thought Reasoning for Time Series Analysis using Reinforcement Learning

通过强化学习激发时间序列分析的链式推理

Felix Parker, Nimeesha Chan, Chi Zhang, Kimia Ghobadi

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

AI总结本文提出COUNTS框架，通过强化学习训练LLM进行时间序列任务的链式推理，提升复杂时间数据处理能力。

2602.13979 2026-02-17 cs.CL 89%

Chain-of-Thought Reasoning with Large Language Models for Clinical Alzheimer's Disease Assessment and Diagnosis

基于大语言模型的链式推理在临床阿尔茨海默病评估与诊断中的应用

Tongze Zhang, Jun-En Ding, Melik Ozolcer, Fang-Ming Hung, Albert Chih-Chieh Yang, Feng Liu, Yi-Rou Ji, Sang Won Bae

机构 * Stevens Institute of Technology（斯蒂文斯理工学院）； Surgical Trauma Intensive Care Unit（外科创伤重症护理科）； Institute of Brain Science（脑科学研究所）； National Yang Ming Chiao Tung University（国立阳明交通大学）

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

AI总结本文提出利用大语言模型的链式推理进行阿尔茨海默病的评估与诊断，通过生成推理路径提升诊断稳定性和性能，F1分数提升达15%。

2601.19061 2026-01-29 cs.CR cs.LG 89%

Thought-Transfer: Indirect Targeted Poisoning Attacks on Chain-of-Thought Reasoning Models

思维转移：链式思维推理模型的间接针对性污染攻击

Harsh Chaudhari, Ethan Rathbun, Hanna Foerster, Jamie Hayes, Matthew Jagielski, Milad Nasr, Ilia Shumailov, Alina Oprea

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

AI总结本文提出了一种名为思维转移的新型攻击方法，通过操纵不同任务学习的推理痕迹，对链式思维推理模型进行间接针对性污染攻击，成功影响目标任务的输出并提升模型性能。

2512.22631 2025-12-30 cs.CL 89%

Evaluating GRPO and DPO for Faithful Chain-of-Thought Reasoning in LLMs

评估GRPO和DPO在LLM中的忠实链式推理能力

Hadi Mohammadi, Tamas Kozak, Anastasia Giachanou

机构 * Utrecht University, Department of Information and Computing Sciences（乌特勒支大学信息与计算科学系）

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

AI总结本文评估GRPO和DPO在提升LLM链式推理忠实性方面的性能，发现GRPO在大模型中表现更优，有助于开发更透明可信的推理方法。

2512.18605 2025-12-23 cs.AI 89%

Reflective Confidence: Correcting Reasoning Flaws via Online Self-Correction

反思性信心：通过在线自我纠正修正推理缺陷

Qinglin Zeng, Jing Yang, Keze Wang

机构 * Sun Yat-sen University（中山大学）

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

AI总结本文提出反思性信心框架，通过在线自我纠正修正推理缺陷，提升数学推理任务的准确性。

Comments Under submission

2505.16782 2025-11-04 cs.CL 89%

Reasoning Beyond Language: A Comprehensive Survey on Latent Chain-of-Thought Reasoning

Xinghao Chen, Anhao Zhao, Heming Xia, Xuan Lu, Hanlin Wang, Yanjun Chen, Wei Zhang, Jian Wang, Wenjie Li, Xiaoyu Shen

机构 * Department of Computing, The Hong Kong Polytechnic University（计算机系，香港理工大学）； Ningbo Digital Twin Institute, Eastern Institute of Technology（宁波数字孪生研究院，东部技术研究所）

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

2510.00579 2025-10-02 cs.CL 89%

CoT Vectors: Transferring and Probing the Reasoning Mechanisms of LLMs

Li Li, Ziyi Wang, Yongliang Wu, Jianfei Cai, Xu Yang

机构 * School of Computer Science & Engineering, Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications (Southeast University), Ministry of Education, China（计算机科学与工程学院，新一代人工智能技术及其交叉应用关键实验室（东南大学），教育部，中国）； Data Science & AI Department at Faculty of IT, Monash University, Australia（信息学院数据科学与人工智能部门，墨尔本大学，澳大利亚）

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

Comments 22 pages, 7 figures

2501.04341 2025-09-25 cs.CL 89%

Understanding Before Reasoning: Enhancing Chain-of-Thought with Iterative Summarization Pre-Prompting

Dong-Hai Zhu, Yu-Jie Xiong, Jia-Chen Zhang, Xi-Jiong Xie, Chun-Ming Xia

机构 * School of Electronic and Electrical Engineering, Shanghai University of Engineering Science（电子工程学院，上海工程技术大学）； School of Information Science and Engineering, Ningbo University（信息科学与工程学院，宁波大学）

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

2506.04182 2025-06-05 cs.CL 89%

Long or short CoT? Investigating Instance-level Switch of Large Reasoning Models

Ruiqi Zhang, Changyi Xiao, Yixin Cao

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

2502.12134 2025-05-28 cs.CL 89%

SoftCoT: Soft Chain-of-Thought for Efficient Reasoning with LLMs

Yige Xu, Xu Guo, Zhiwei Zeng, Chunyan Miao

机构 * Joint NTU-UBC Research Centre of Excellence in Active Living for the Elderly（联合NTU-UBC老年人积极生活卓越研究中心）； College of Computing and Data Science（计算与数据科学学院）

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

Comments Camera-ready for ACL 2025 (main conference)

2505.15927 2025-05-23 stat.ML cs.LG 89%

CoT Information: Improved Sample Complexity under Chain-of-Thought Supervision

Awni Altabaa, Omar Montasser, John Lafferty

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

2410.05695 2024-10-29 cs.CL 89%

Unlocking the Capabilities of Thought: A Reasoning Boundary Framework to Quantify and Optimize Chain-of-Thought

Qiguang Chen, Libo Qin, Jiaqi Wang, Jinxuan Zhou, Wanxiang Che

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

Comments Accepted at NeurIPS 2024 (Oral)

2305.16582 2024-03-26 cs.CL 89%

Beyond Chain-of-Thought, Effective Graph-of-Thought Reasoning in Language Models

Yao Yao, Zuchao Li, Hai Zhao

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

2305.03453 2023-12-19 cs.CL 89%

T-SciQ: Teaching Multimodal Chain-of-Thought Reasoning via Mixed Large Language Model Signals for Science Question Answering

Lei Wang, Yi Hu, Jiabang He, Xing Xu, Ning Liu, Hui Liu, Heng Tao Shen

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

Comments AAAI 2024

2312.08762 2023-12-15 cs.AI 89%

Multi-modal Latent Space Learning for Chain-of-Thought Reasoning in Language Models

Liqi He, Zuchao Li, Xiantao Cai, Ping Wang

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

2305.17306 2023-05-30 cs.CL cs.AI cs.LG 89%

Chain-of-Thought Hub: A Continuous Effort to Measure Large Language Models' Reasoning Performance

Yao Fu, Litu Ou, Mingyu Chen, Yuhao Wan, Hao Peng, Tushar Khot

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

Comments Preprint. Code at https://github.com/FranxYao/chain-of-thought-hub

2606.17687 2026-06-17 cs.CL cs.AI 新提交 89%

SuCo: Sufficiency-guided Continuous Adaptive Reasoning

SuCo: 充分性引导的连续自适应推理

Jiahao Wang, Bingyu Liang, Chenhao Hu, Longhui Zhang, Xuebo Liu, Min zhang, Jing Li, Xuelong Li

机构 * Harbin Institute of Technology (Shenzhen)（哈尔滨工业大学（深圳））

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

AI总结针对大型推理模型生成过长思维链导致计算浪费的问题，提出最小充分CoT概念，并构建两阶段训练框架SuCo，通过自适应充分性阈值和强化学习优化推理长度，在数学、代码和科学基准上同时提升准确率和效率。

Comments Accepted to ICML 2026. 18 pages

2605.09346 2026-05-12 cs.CL cs.AI 89%

RuPLaR : Efficient Latent Compression of LLM Reasoning Chains with Rule-Based Priors From Multi-Step to One-Step

RuPLaR : 通过基于规则的先验概率实现LLM推理链的高效潜在压缩

Xiaocheng Luo, Kang Wang, Zaifu Zhan, Yuechi Zhou, Xiangyu Duan

机构 * School of Computer Science and Technology（计算机科学与技术学院）； Department of Electrical and Computer Engineering（电气与计算机工程系）

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

AI总结本文提出RuPLaR框架，通过基于规则的先验概率指导LLM生成单阶段潜在推理令牌，提升推理效率与准确性，实验表明其比现有方法提升11.1%的准确率且消耗更少token。

Comments 15 pages, 15 figures

2605.20201 2026-05-25 cs.CL cs.AI cs.LG 89%

Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning

基于代理思维链调优的长上下文推理

Miao Li, Irina Saparina, Alexander Gurung, Mirella Lapata

机构 * School of Informatics, University of Edinburgh（爱丁堡大学信息学院）

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

AI总结提出ProxyCoT训练框架，通过将短代理上下文中的推理能力迁移到完整长上下文中，以提升大语言模型在长上下文复杂推理任务上的表现。

Comments Long paper, ACL 2026 (Main conference)

2509.22221 2026-02-03 cs.CV 89%

Towards Faithful Reasoning in Remote Sensing: A Perceptually-Grounded GeoSpatial Chain-of-Thought for Vision-Language Models

迈向遥感中的可信推理：一种基于感知的地理空间思维链用于视觉-语言模型

Jiaqi Liu, Lang Sun, Ronghao Fu, Bo Yang

机构 * Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education（教育部符号计算与知识工程重点实验室）

专题命中复杂问题求解：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract)

AI总结本研究提出Geo-CoT框架，通过两阶段策略提升遥感VLMs的推理能力，实现可验证的多步骤分析，显著优于现有模型。

2511.17027 2025-11-24 cs.SE 89%

ReVul-CoT: Towards Effective Software Vulnerability Assessment with Retrieval-Augmented Generation and Chain-of-Thought Prompting

ReVul-CoT：通过检索增强生成与链式推理提示实现有效的软件漏洞评估

Zhijie Chen, Xiang Chen, Ziming Li, Jiacheng Xue, Chaoyang Gao

专题命中复杂问题求解：chain-of-thought(title,abstract)；CoT(title,abstract)；reasoning(abstract)

AI总结 ReVul-CoT通过整合检索增强生成与链式推理提示，提升软件漏洞评估的准确性和效率。

2507.02199 2025-09-30 cs.CL cs.AI cs.LG 89%

Latent Chain-of-Thought? Decoding the Depth-Recurrent Transformer

Wenquan Lu, Yuechuan Yang, Kyle Lee, Yanshu Li, Enqi Liu

机构 * Brown University（布朗大学）； Harvard University（哈佛大学）

专题命中复杂问题求解：chain-of-thought(title,abstract)；reasoning(abstract,comments)；planning(abstract,comments)；CoT(abstract)

Comments First Workshop on the Application of LLM Explainability to Reasoning and Planning at COLM 2025

2505.19877 2025-05-27 cs.CV 89%

Vad-R1: Towards Video Anomaly Reasoning via Perception-to-Cognition Chain-of-Thought

Chao Huang, Benfeng Wang, Jie Wen, Chengliang Liu, Wei Wang, Li Shen, Xiaochun Cao

机构 * Shenzhen Campus of Sun Yat-sen University（中山大学深圳校区）； Harbin Institute of Technology, Shenzhen（哈尔滨工业大学深圳研究院）； Hong Kong Polytechnic University（香港理工大学）

专题命中复杂问题求解：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract)

Comments 9 pages, 4 figures

2404.10513 2024-11-28 cs.CL cs.AI cs.LG 89%

CoTAR: Chain-of-Thought Attribution Reasoning with Multi-level Granularity

Moshe Berchansky, Daniel Fleischer, Moshe Wasserblat, Peter Izsak

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

Comments Findings of the Association for Computational Linguistics: EMNLP 2024

2407.04078 2024-07-18 cs.CL cs.AI cs.LG 89%

DotaMath: Decomposition of Thought with Code Assistance and Self-correction for Mathematical Reasoning

Chengpeng Li, Guanting Dong, Mingfeng Xue, Ru Peng, Xiang Wang, Dayiheng Liu

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

Comments Work in progress

2603.16331 2026-05-25 cs.LG 89%

Decoding the Critique Mechanism in Large Reasoning Models

解码大型推理模型中的批判机制

Hoang Phan, Quang H. Nguyen, Hung T. Q. Le, Xiusi Chen, Heng Ji, Khoa D. Doan

机构 * VinUni-Illinois Smart Health Center（VinUniversity-伊利诺伊州智能健康中心）； University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）

专题命中复杂问题求解：reasoning(title,abstract)；CoT(abstract,abstract_cn)；chain-of-thought(abstract)；self-correction(abstract)

AI总结本文通过插入算术错误研究大型推理模型如何从错误中恢复，发现存在隐藏的批判向量，通过引导潜在表示可提升错误检测和测试时扩展性能。

2505.13975 2026-04-28 cs.CL 89%

DRP: Distilled Reasoning Pruning with Skill-aware Step Decomposition for Efficient Large Reasoning Models

DRP：基于技能感知步骤分解的 distilled 推理剪枝用于高效大推理模型

Yuxuan Jiang, Dawei Li, Francis Ferraro

机构 * University of Maryland, Baltimore County（马里兰大学巴尔的摩分校）； Arizona State University（亚利桑那州立大学）

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

AI总结本文提出DRP框架，结合推理时剪枝与调优-based知识蒸馏，提升大推理模型的效率与准确性，在数学推理任务中实现显著的token效率提升。

Comments Published on ACL 2026

2506.18957 2025-06-25 cs.AI cs.CL cs.LG 89%

A Comment On "The Illusion of Thinking": Reframing the Reasoning Cliff as an Agentic Gap

Sheraz Khan, Subha Madhavan, Kannan Natarajan

专题命中复杂问题求解：reasoning(title,abstract)；chain-of-thought(abstract)；CoT(abstract)；self-correction(abstract)

Comments 10 pages, 2 figures, Comment on "The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity" (arXiv:2506.06941v1)

2608.11573 2026-08-13 cs.CL cs.AI 新提交 88%

Reinforcing Step-level Reasoning for Effective Self-Correction in LLMs

强化步骤级推理以实现大语言模型的有效自校正

Vu Duc Anh, Nhat M. Hoang, Do Xuan Long, Cong-Duy Nguyen, Ponhvoan Srey, Luu Anh Tuan

机构 * Nanyang Technological University（南洋理工大学）； National University of Singapore（新加坡国立大学）； VinUniversity ； Institute for Infocomm Research (IR), A*STAR（新加坡科技研究局信息通信研究院）

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

AI总结针对大语言模型自校正的核心挑战，提出SFS-DPO及教师辅助变体SFS-DPO-R框架，经多模型多域评估，其性能优于现有步骤级训练基线，可提升自校正频率与有效性。

2601.05459 2026-01-12 cs.CL cs.AI 88%

Do LLMs Need Inherent Reasoning Before Reinforcement Learning? A Study in Korean Self-Correction

LLMs是否需要在强化学习前具备内在推理能力？韩国自我修正研究

Hongjin Kim, Jaewook Lee, Kiyoung Lee, Jong-hun Shin, Soojong Lim, Oh-Woog Kwon

机构 * ETRI

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

AI总结本研究探讨LLMs在强化学习前是否需要具备内在推理能力，通过韩国自我修正研究发现，对齐模型内部推理与韩语输入关键，提升多语言推理效果。

Comments IJCNLP-AACL 2025 (Main), Outstanding Paper Award

1. 复杂问题求解 5001 篇

Eliciting Chain-of-Thought Reasoning for Time Series Analysis using Reinforcement Learning

Chain-of-Thought Reasoning with Large Language Models for Clinical Alzheimer's Disease Assessment and Diagnosis

Thought-Transfer: Indirect Targeted Poisoning Attacks on Chain-of-Thought Reasoning Models

Evaluating GRPO and DPO for Faithful Chain-of-Thought Reasoning in LLMs

Reflective Confidence: Correcting Reasoning Flaws via Online Self-Correction

Reasoning Beyond Language: A Comprehensive Survey on Latent Chain-of-Thought Reasoning

CoT Vectors: Transferring and Probing the Reasoning Mechanisms of LLMs

Understanding Before Reasoning: Enhancing Chain-of-Thought with Iterative Summarization Pre-Prompting

Long or short CoT? Investigating Instance-level Switch of Large Reasoning Models

SoftCoT: Soft Chain-of-Thought for Efficient Reasoning with LLMs

CoT Information: Improved Sample Complexity under Chain-of-Thought Supervision

Unlocking the Capabilities of Thought: A Reasoning Boundary Framework to Quantify and Optimize Chain-of-Thought

Beyond Chain-of-Thought, Effective Graph-of-Thought Reasoning in Language Models

T-SciQ: Teaching Multimodal Chain-of-Thought Reasoning via Mixed Large Language Model Signals for Science Question Answering

Multi-modal Latent Space Learning for Chain-of-Thought Reasoning in Language Models

Chain-of-Thought Hub: A Continuous Effort to Measure Large Language Models' Reasoning Performance

SuCo: Sufficiency-guided Continuous Adaptive Reasoning

RuPLaR : Efficient Latent Compression of LLM Reasoning Chains with Rule-Based Priors From Multi-Step to One-Step

Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning

Towards Faithful Reasoning in Remote Sensing: A Perceptually-Grounded GeoSpatial Chain-of-Thought for Vision-Language Models

ReVul-CoT: Towards Effective Software Vulnerability Assessment with Retrieval-Augmented Generation and Chain-of-Thought Prompting

Latent Chain-of-Thought? Decoding the Depth-Recurrent Transformer

Vad-R1: Towards Video Anomaly Reasoning via Perception-to-Cognition Chain-of-Thought

CoTAR: Chain-of-Thought Attribution Reasoning with Multi-level Granularity

DotaMath: Decomposition of Thought with Code Assistance and Self-correction for Mathematical Reasoning

Decoding the Critique Mechanism in Large Reasoning Models

DRP: Distilled Reasoning Pruning with Skill-aware Step Decomposition for Efficient Large Reasoning Models

A Comment On "The Illusion of Thinking": Reframing the Reasoning Cliff as an Agentic Gap

Reinforcing Step-level Reasoning for Effective Self-Correction in LLMs

Do LLMs Need Inherent Reasoning Before Reinforcement Learning? A Study in Korean Self-Correction