大模型推理能力 - arXivDaily 专题

2502.01694 2025-03-04 cs.AI cs.LG stat.ML 91%

Metastable Dynamics of Chain-of-Thought Reasoning: Provable Benefits of Search, RL and Distillation

Juno Kim, Denny Wu, Jason Lee, Taiji Suzuki

专题命中规划推理：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract)；verifier(abstract)

Comments 55 pages, 3 figures

2404.11041 2024-06-19 cs.AI cs.LG 91%

On the Empirical Complexity of Reasoning and Planning in LLMs

Liwei Kang, Zirui Zhao, David Hsu, Wee Sun Lee

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；CoT(abstract)

2404.03891 2024-04-08 cs.RO cs.AI cs.LG 91%

Can only LLMs do Reasoning?: Potential of Small Language Models in Task Planning

Gawon Choi, Hyemin Ahn

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；CoT(abstract)

Comments 8 pages, 11 figures

2310.12342 2023-11-16 cs.CL cs.AI 91%

Eliminating Reasoning via Inferring with Planning: A New Framework to Guide LLMs' Non-linear Thinking

Yongqi Tong, Yifan Wang, Dawei Li, Sizhe Wang, Zi Lin, Simeng Han, Jingbo Shang

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；CoT(abstract)

2605.27773 2026-05-28 cs.CL cs.AI cs.LG 90%

Do Models Know Why They Changed Their Mind? Interpretability and Faithfulness of Chain-of-Thought Under Knowledge Conflict

模型是否知道它们为何改变主意？知识冲突下思维链的可解释性与忠实性

Pruthvinath Jeripity Venkata

机构 * Independent Researcher（独立研究员）

专题命中规划推理：CoT(summary_cn,abstract)；chain-of-thought(title,abstract)；reasoning(abstract)；分类 cs.CL、cs.AI、cs.LG

AI总结通过引入内省忠实性，研究在知识冲突下语言模型的思维链推理是否忠实反映其决策机制，发现CoT高度稳定但置信度携带微弱真实信号。

Comments 12 pages, 8 tables, 3 appendices

2503.06892 2025-03-11 cs.RO 90%

SafePlan: Leveraging Formal Logic and Chain-of-Thought Reasoning for Enhanced Safety in LLM-based Robotic Task Planning

Ike Obi, Vishnunandan L. N. Venkatesh, Weizheng Wang, Ruiqi Wang, Dayoon Suh, Temitope I. Amosa, Wonse Jo, Byung-Cheol Min

专题命中规划推理：chain-of-thought(title,abstract)；reasoning(title)；planning(title)；CoT(abstract)

2601.03154 2026-04-21 cs.CL 90%

Decoupling the Effect of Chain-of-Thought Reasoning: A Human Label Variation Perspective

解耦链式推理的影响：从人类标签变异视角出发

Beiduo Chen, Tiancheng Hu, Caiqi Zhang, Robert Litschko, Anna Korhonen, Barbara Plank

机构 * MaiNLP ； Center for Information and Language Processing, LMU Munich（信息与语言处理中心，慕尼黑大学）； LMU Munich, Germany（慕尼黑大学，德国）； Munich Center for Machine Learning (MCML), Munich, Germany（慕尼黑机器学习中心（MCML），慕尼黑，德国）； Language Technology Lab, University of Cambridge, United Kingdom（语言技术实验室，剑桥大学，英国）

专题命中规划推理：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract,abstract_cn)；分类 cs.CL

AI总结本文通过系统解耦实验探讨了链式推理对分布任务的影响，发现推理内容对最终准确性贡献大，而模型先验对分布排名起主导作用。

Comments Accepted by ACL 2026 Findings, 21 pages, 10 figures

2601.11340 2026-04-16 cs.CL 90%

Neural Chain-of-Thought Search: Searching the Optimal Reasoning Path to Enhance Large Language Models

神经链式推理搜索：寻找最优推理路径以增强大语言模型

Guoming Ling, Zhongzhan Huang, Yupei Lin, Junxin Li, Shanshan Zhong, Hefeng Wu, Liang Lin

机构 * Sun Yat-sen University（中山大学）

专题命中规划推理：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract,abstract_cn)；分类 cs.CL

AI总结本文提出神经链式推理搜索框架，通过动态搜索最优推理策略提升大语言模型性能，实验表明在多个基准测试中准确率提升超3.5%，生成长度减少超22%。

2602.22624 2026-02-27 cs.CV cs.AI 90%

Instruction-based Image Editing with Planning, Reasoning, and Generation

基于指令的图像编辑与规划、推理和生成

Liya Ji, Chenyang Qi, Qifeng Chen

机构 * HKUST（香港科技大学）

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；CoT(abstract)

AI总结本文提出一种多模态模型，通过链式思考规划、编辑区域推理和编辑，提升基于指令的图像编辑能力，以应对更复杂的真实场景。

Comments 10 pages, 7 figures

Journal ref Proceedings of the IEEE/CVF International Conference on Computer Vision, 2025, Page 17506--17515

2602.15460 2026-02-18 cs.LG cs.CV 90%

On the Out-of-Distribution Generalization of Reasoning in Multimodal LLMs for Simple Visual Planning Tasks

在简单视觉规划任务中多模态大语言模型推理的分布外泛化

Yannic Neuhaus, Nicolas Flammarion, Matthias Hein, Francesco Croce

机构 * Tübingen AI Center -- University of Tübingen（图宾根人工智能中心 -- 图宾根大学）； EPFL（瑞士联邦理工学院）； ELLIS Institute Finland -- Aalto University（芬兰ELLIS研究所 -- 阿尔托大学）

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；CoT(abstract)

AI总结研究多模态大语言模型在简单视觉规划任务中推理能力的分布外泛化，发现结合多种文本格式的推理方法效果最佳，纯文本模型表现优于图像输入模型。

2511.22532 2025-12-01 cs.CV cs.AI 90%

CoT4AD: A Vision-Language-Action Model with Explicit Chain-of-Thought Reasoning for Autonomous Driving

CoT4AD: 一种具有显式推理链的视觉-语言-动作模型用于自动驾驶

Zhaohui Wang, Tengbo Yu, Hao Tang

机构 * Peking University（北京大学）

专题命中规划推理：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract)；planning(abstract)

AI总结 CoT4AD通过引入显式推理链提升自动驾驶中的视觉-语言-动作模型的推理能力，实现更精确的因果推理和决策。

Comments 10 pages, 3 figures

2504.00010 2025-10-20 cs.LG cs.GR cs.MA 90%

LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration

Yuyao Zhang, Jinghao Li, Yu-Wing Tai

机构 * Dartmouth College（达特茅斯学院）； CUHK（香港大学）

专题命中规划推理：reasoning(title,abstract)；CoT(title,abstract)；chain-of-thought(abstract)；planning(abstract)

Comments 26 pages

2401.17686 2024-10-22 cs.CL 90%

Deductive Beam Search: Decoding Deducible Rationale for Chain-of-Thought Reasoning

Tinghui Zhu, Kai Zhang, Jian Xie, Yu Su

专题命中规划推理：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract)；verifier(abstract)

Comments COLM 2024

2410.01428 2024-10-03 cs.CL 90%

Can We Further Elicit Reasoning in LLMs? Critic-Guided Planning with Retrieval-Augmentation for Solving Challenging Tasks

Xingxuan Li, Weiwen Xu, Ruochen Zhao, Fangkai Jiao, Shafiq Joty, Lidong Bing

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；math reasoning(abstract)

Comments Work in progress

2603.26807 2026-07-30 cs.IR cs.AI cs.CL 版本更新 90%

GroupRAG: Cognitively Inspired Group-Aware Retrieval and Reasoning via Knowledge-Driven Problem Structuring

GroupRAG: 基于知识驱动的问题结构化进行的认知启发式组感知检索与推理

Xinyi Duan, Yuanrong Tang, Jiangtao Gong

机构 * Tsinghua University（清华大学）

专题命中规划推理：CoT(summary_cn,abstract)；reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

AI总结 GroupRAG通过知识驱动的关键点分组，实现组感知的检索与推理，提升语言模型在现实场景中的表现，实验表明其优于RAG和CoT基线方法。

Comments 9 pages, 3 figures

2604.12076 2026-04-15 cs.CL cs.AI cs.CY 90%

Narrative over Numbers: The Identifiable Victim Effect and its Amplification Under Alignment and Reasoning in Large Language Models

数字之外的叙事：可识别受害者效应及其在对齐和推理中的放大

Syed Rifat Raiyan

机构 * Systems and Software Lab (SSL), Department of Computer Science and Engineering（系统与软件实验室（SSL），计算机科学与工程系）

专题命中规划推理：CoT(summary_cn,abstract)；reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

AI总结研究探讨了大语言模型中可识别受害者效应的普遍存在及其受对齐训练的影响，发现指令调优模型表现出极端效应，而推理专精模型则逆转了该效应，且标准CoT提示放大了该效应。

Comments Under review, 49 pages, 20 figures, 11 tables

2507.23773 2026-05-22 cs.AI cs.CL cs.LG cs.RO 90%

General Agentic Planning Through Simulative Reasoning with World Models

通过世界模型的模拟推理实现通用代理规划

Mingkai Deng, Jinyu Hou, Zhiting Hu, Eric Xing

机构 * Institute of Foundation Models (IFM)（基础模型研究所）； Carnegie Mellon University（卡内基梅隆大学）； UC San Diego（南加州大学）

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI、cs.LG

AI总结本文提出通过模拟推理实现通用代理规划，利用世界模型进行未来状态预测，提升决策能力，通过SiRA架构在不同任务中取得更高任务完成率。

Comments Winner of Berkeley LLM Agents Hackathon (Fundamentals Track); code available at https://github.com/sailing-lab/sira

2407.18553 2024-08-01 cs.IR 90%

REAPER: Reasoning based Retrieval Planning for Complex RAG Systems

Ashutosh Joshi, Sheikh Muhammad Sarwar, Samarth Varshney, Sreyashi Nag, Shrivats Agrawal, Juhi Naik

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；CoT(abstract)

2601.10775 2026-04-14 cs.CL cs.GT cs.LG 90%

LLMs for Game Theory: Entropy-Guided In-Context Learning and Adaptive CoT Reasoning

大型语言模型用于博弈论：基于熵的上下文学习与自适应推理

Tommaso Felice Banfi, Sashenka Gamage

专题命中规划推理：reasoning(title,abstract)；CoT(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.LG

AI总结本文提出一种基于LLM的框架，通过熵引导的链式推理和自适应上下文检索，提升离散博弈任务的推理能力，实验显示其显著提高决策质量。

Comments Published at the AAAI 2026 Bridge: Logical and Symbolic Reasoning in Language Models (OpenReview)

2512.24014 2026-01-01 cs.CL cs.AI 90%

iCLP: Large Language Model Reasoning with Implicit Cognition Latent Planning

iCLP: 基于隐式认知潜在规划的大语言模型推理

Sijia Chen, Di Niu

机构 * Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））； University of Alberta（阿尔伯塔大学）

专题命中规划推理：planning(title,abstract)；reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

AI总结 iCLP通过隐式认知潜在规划提升大语言模型的推理准确性和效率，实现跨领域泛化与可解释性。

Comments 9 pages, 6 figures. The source code is publicly available at https://github.com/AgenticFinLab/latent-planning

2608.01319 2026-08-04 cs.AI 新提交 90%

Cognitive Demand Steering for Adaptive Meta-Reasoning in Large Language Models

大语言模型自适应元推理的认知需求引导

John Scoville, Shengzhuang Chen, Yejin Bang, Stefan Winzeck, Jonathan Richard Schwarz

机构 * Thomson Reuters Foundational Research（汤森路透基础研究部）； Imperial College London（帝国理工学院）

专题命中规划推理：CoT(summary_cn,abstract)；reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.AI

AI总结该研究提出无需训练的元推理框架CDS，通过认知量表刻画需求，在三类大模型和六个基准上较直接调用、标准CoT分别提升21.9%、9%准确率，难数学编码任务增益显著。

Comments 21 pages, 3 figures

2510.08992 2026-05-14 cs.LG 90%

Constraints-of-Thought: A Framework for Constrained Reasoning in Language-Model-Guided Search

思维约束：一种在语言模型引导搜索中进行约束推理的框架

Kamel Alrashedy, Vriksha Srihari, Zulfiqar Zaidi, Ridam Srivastava, Pradyumna Tambwekar, Matthew Gombolay

机构 * Georgia Institute of Technology（佐治亚理工学院）

专题命中规划推理：reasoning(title,abstract)；CoT(abstract,abstract_cn)；chain-of-thought(abstract)；planning(abstract)

AI总结本文提出Const-o-T框架，通过约束推理提升语言模型在多步骤任务中的规划效率和决策可靠性，适用于风险游戏、CAD代码生成和算术推理等复杂领域。

2605.06040 2026-05-08 cs.AI cs.CL 90%

Novelty-based Tree-of-Thought Search for LLM Reasoning and Planning

基于新颖性的树状思维搜索用于大语言模型推理与规划

Leon Hamm, Zlatan Ajanovic

机构 * RWTH Aachen（亚琛工业大学）

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

AI总结本文提出基于新颖性的树状思维搜索方法，通过引入新颖性概念优化语言领域推理与规划任务，减少搜索范围和token成本。

2601.21611 2026-01-30 cs.IR cs.AI cs.CL 90%

Thinking Broad, Acting Fast: Latent Reasoning Distillation from Multi-Perspective Chain-of-Thought for E-Commerce Relevance

深度思考，快速行动：多视角链式推理的潜在推理蒸馏用于电子商务相关性

Baopu Qiu, Hao Chen, Yuanrong Wu, Changtong Zan, Chao Wei, Weiru Zhang, Xiaoyi Zeng

机构 * Alibaba International Digital Commerce Group（阿里巴巴国际数字商业集团）； Zhejiang University（浙江大学）

专题命中规划推理：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract)；分类 cs.CL、cs.AI

AI总结本文提出多视角链式推理蒸馏方法，通过改进的教师模型和轻量级学生模型提升电子商务搜索相关性建模的准确性和效率。

Comments 12 pages, 6 figures, Accepted by WWW2026 industry track

2512.20586 2025-12-24 cs.AI cs.CL cs.HC 90%

Automated stereotactic radiosurgery planning using a human-in-the-loop reasoning large language model agent

利用人类在回路的推理大型语言模型代理进行自动化立体定向放射手术计划

Humza Nusrat, Luke Francisco, Bing Luo, Hassan Bagher-Ebadian, Joshua Kim, Karen Chin-Snyder, Salim Siddiqui, Mira Shah, Eric Mellon, Mohammad Ghassemi, Anthony Doemer, Benjamin Movsas, Kundan Thind

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

AI总结本文提出利用人类在回路的推理大型语言模型代理进行自动化立体定向放射手术计划，通过对比推理模型与非推理模型，展示了推理模型在剂量优化和计划行为上的优势。

2510.18817 2025-10-22 cs.CL cs.AI 90%

Fine-Tuned Thoughts: Leveraging Chain-of-Thought Reasoning for Industrial Asset Health Monitoring

Shuxin Lin, Dhaval Patel, Christodoulos Constantinides

机构 * IBM Research（IBM研究院）

专题命中规划推理：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract)；分类 cs.CL、cs.AI

Comments Accepted at EMNLP 2025

2507.16815 2025-09-19 cs.CV cs.AI cs.LG cs.RO 90%

ThinkAct: Vision-Language-Action Reasoning via Reinforced Visual Latent Planning

Chi-Pin Huang, Yueh-Hua Wu, Min-Hung Chen, Yu-Chiang Frank Wang, Fu-En Yang

机构 * NVIDIA ； National Taiwan University（国立台湾大学）

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；self-correction(abstract)；分类 cs.AI、cs.LG

Comments NeurIPS 2025. Project page: https://jasper0314-huang.github.io/thinkact-vla/

2502.12521 2025-02-19 cs.AI cs.LG 90%

Inference-Time Computations for LLM Reasoning and Planning: A Benchmark and Insights

Shubham Parashar, Blake Olson, Sambhav Khurana, Eric Li, Hongyi Ling, James Caverlee, Shuiwang Ji

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；logical reasoning(abstract)；分类 cs.AI、cs.LG

2412.09666 2024-12-16 cs.AI cs.CL 90%

Systematic Analysis of LLM Contributions to Planning: Solver, Verifier, Heuristic

Haoming Li, Zhaoliang Chen, Songyuan Liu, Yiming Lu, Fei Liu

专题命中规划推理：planning(title,abstract)；verifier(title,abstract)；reasoning(abstract)；分类 cs.CL、cs.AI

2402.00658 2024-10-16 cs.AI cs.CL 90%

Learning Planning-based Reasoning by Trajectories Collection and Process Reward Synthesizing

Fangkai Jiao, Chengwei Qin, Zhengyuan Liu, Nancy F. Chen, Shafiq Joty

专题命中规划推理：reasoning(title,abstract)；planning(title,abstract)；logical reasoning(abstract)；分类 cs.CL、cs.AI

Comments 17 pages, 9 figures. EMNLP 2024

1. 规划推理 10953 篇

Metastable Dynamics of Chain-of-Thought Reasoning: Provable Benefits of Search, RL and Distillation

On the Empirical Complexity of Reasoning and Planning in LLMs

Can only LLMs do Reasoning?: Potential of Small Language Models in Task Planning

Eliminating Reasoning via Inferring with Planning: A New Framework to Guide LLMs' Non-linear Thinking

Do Models Know Why They Changed Their Mind? Interpretability and Faithfulness of Chain-of-Thought Under Knowledge Conflict

SafePlan: Leveraging Formal Logic and Chain-of-Thought Reasoning for Enhanced Safety in LLM-based Robotic Task Planning

Decoupling the Effect of Chain-of-Thought Reasoning: A Human Label Variation Perspective

Neural Chain-of-Thought Search: Searching the Optimal Reasoning Path to Enhance Large Language Models

Instruction-based Image Editing with Planning, Reasoning, and Generation

On the Out-of-Distribution Generalization of Reasoning in Multimodal LLMs for Simple Visual Planning Tasks

CoT4AD: A Vision-Language-Action Model with Explicit Chain-of-Thought Reasoning for Autonomous Driving

LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration

Deductive Beam Search: Decoding Deducible Rationale for Chain-of-Thought Reasoning

Can We Further Elicit Reasoning in LLMs? Critic-Guided Planning with Retrieval-Augmentation for Solving Challenging Tasks

GroupRAG: Cognitively Inspired Group-Aware Retrieval and Reasoning via Knowledge-Driven Problem Structuring

Narrative over Numbers: The Identifiable Victim Effect and its Amplification Under Alignment and Reasoning in Large Language Models

General Agentic Planning Through Simulative Reasoning with World Models

REAPER: Reasoning based Retrieval Planning for Complex RAG Systems

LLMs for Game Theory: Entropy-Guided In-Context Learning and Adaptive CoT Reasoning

iCLP: Large Language Model Reasoning with Implicit Cognition Latent Planning

Cognitive Demand Steering for Adaptive Meta-Reasoning in Large Language Models

Constraints-of-Thought: A Framework for Constrained Reasoning in Language-Model-Guided Search

Novelty-based Tree-of-Thought Search for LLM Reasoning and Planning

Thinking Broad, Acting Fast: Latent Reasoning Distillation from Multi-Perspective Chain-of-Thought for E-Commerce Relevance

Automated stereotactic radiosurgery planning using a human-in-the-loop reasoning large language model agent

Fine-Tuned Thoughts: Leveraging Chain-of-Thought Reasoning for Industrial Asset Health Monitoring

ThinkAct: Vision-Language-Action Reasoning via Reinforced Visual Latent Planning

Inference-Time Computations for LLM Reasoning and Planning: A Benchmark and Insights

Systematic Analysis of LLM Contributions to Planning: Solver, Verifier, Heuristic

Learning Planning-based Reasoning by Trajectories Collection and Process Reward Synthesizing