大模型推理能力

Reasoning Runtime Behavior of a Program with LLM: How Far Are We?

Junkai Chen, Zhiyuan Pan, Xing Hu, Zhenhao Li, Ge Li, Xin Xia

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.CL

Comments Accepted by 47th IEEE/ACM International Conference on Software Engineering (ICSE 2025). Our REval leaderboard is available at https://r-eval.github.io

A Neurodiversity-Inspired Solver for the Abstraction \& Reasoning Corpus (ARC) Using Visual Imagery and Program Synthesis

James Ainooson, Deepayan Sanyal, Joel P. Michelson, Yuan Yang, Maithilee Kunda

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.AI

Trusta: Reasoning about Assurance Cases with Formal Methods and Large Language Models

Zezhong Chen, Yuxin Deng, Wenjie Du

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.AI

Comments 38 pages

Large Language Models on the Chessboard: A Study on ChatGPT's Formal Language Comprehension and Complex Reasoning Skills

Mu-Tien Kuo, Chih-Chung Hsueh, Richard Tzong-Han Tsai

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.CL

Reasoning about Causality in Games

Lewis Hammond, James Fox, Tom Everitt, Ryan Carey, Alessandro Abate, Michael Wooldridge

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.AI

Comments Published in Artificial Intelligence (2023)

An Argumentation-Based Legal Reasoning Approach for DL-Ontology

Zhe Yu, Yiwei Lu

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.AI

Comments 16 pages, 1 figure

Pushing the Limits of Rule Reasoning in Transformers through Natural Language Satisfiability

Kyle Richardson, Ashish Sabharwal

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.CL

Comments Accepted to AAAI-2022, AAAI preprint

Negation in Cognitive Reasoning

Claudia Schon, Sophie Siebert, Frieder Stolzenburg

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.CL

Comments 19 pages, 5 figures, 4 tables, extended version

Journal ref In Stefan Edelkamp, Ralf Moeller, and Elmar Rueckert, editors, KI 2021: Advances in Artificial Intelligence - 44th German Conference on AI, LNAI 12873, pages 217-232. Springer, 2021

Proof-Carrying Plans: a Resource Logic for AI Planning

Alasdair Hill, Ekaterina Komendantskaya, Ronald P. A. Petrick

专题命中 代码与定理证明 ：planning(title,abstract)；分类 cs.AI

Comments PPDP 2020, 13 pages, 9 figures

Towards United Reasoning for Automatic Induction in Isabelle/HOL

Yutaka Nagashima

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.AI

Comments This is the pre-print of our short-paper accepted at the 34th Annual Conference of the Japanese Society for Artificial Intelligence, 2020 (https://www.ai-gakkai.or.jp/jsai2020/en)

Learning Dynamics of Attention: Human Prior for Interpretable Machine Reasoning

Wonjae Kim, Yoonho Lee

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.LG

Comments 20 pages, 18 figures, 2 tables

Stream Reasoning-Based Control of Caching Strategies in CCN Routers

Harald Beck, Bruno Bierbaumer, Minh Dao-Tran, Thomas Eiter, Hermann Hellwagner, Konstantin Schekotihin

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.AI

Comments 21 pages, 8 figures

Proceedings First Workshop on Causal Reasoning for Embedded and safety-critical Systems Technologies

Gregor Gössler, Oleg Sokolsky

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.AI

Journal ref EPTCS 224, 2016

Representing and Reasoning about Game Strategies

Dongmo Zhang, Michael Thielsher

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.AI

A Formal Analysis of Required Cooperation in Multi-agent Planning

Yu Zhang, Subbarao Kambhampati

专题命中 代码与定理证明 ：planning(title,abstract)；分类 cs.AI

Rational Nonmonotonic Reasoning

Carl Kadie

专题命中 代码与定理证明 ：reasoning(title,abstract)；分类 cs.AI

Comments Appears in Proceedings of the Fourth Conference on Uncertainty in Artificial Intelligence (UAI1988)

Compile to Compress: Boosting Formal Theorem Provers by Compiler Outputs

编译以压缩：通过编译器输出提升形式定理证明器

Guchan Li, Rui Tian, Hongning Wang

机构 * Department of Computer Science and Technology, Tsinghua University, Beijing, China（清华大学计算机科学与技术系） 纠错

专题命中 代码与定理证明 ：reasoning(abstract)；test-time compute(abstract)；verifier(abstract)；分类 cs.AI、cs.LG

AI总结 利用编译器将大量证明尝试压缩为结构化失败模式，提出一种学习-精炼框架，通过树搜索基于验证器反馈局部修正错误，在可比测试时预算下在PutnamBench上达到最先进性能。

EconProver: Towards More Economical Test-Time Scaling for Automated Theorem Proving

Mukai Li, Linfeng Song, Zhenwen Liang, Jiahao Xu, Shansan Gong, Qi Liu, Haitao Mi, Dong Yu

机构 * Tencent（腾讯） ； The University of Hong Kong（香港大学） 纠错

专题命中 代码与定理证明 ：reasoning(abstract)；chain-of-thought(abstract)；CoT(abstract)；分类 cs.CL、cs.AI

Uncovering Latent Chain of Thought Vectors in Language Models

Jason Zhang, Scott Viteri

专题命中 代码与定理证明 ：reasoning(abstract)；chain-of-thought(abstract)；CoT(abstract)；分类 cs.CL、cs.AI

Comments This work was presented at the Workshop on Neural Network Weights as a New Data Modality at ICLR 2025

Proof-of-Execution Memory: Defending LLM Agents Against Forged-Reasoning Attacks by Verifying What Actually Happened

执行证明内存：通过验证实际发生的内容防御大语言模型智能体的伪造推理攻击

Md Habibur Rahman, Jaeho Kim

专题命中 代码与定理证明 ：reasoning(title,abstract)

AI总结 该研究针对大语言模型智能体的伪造推理攻击，提出执行证明内存（PoEM）防御方案，通过维护防篡改的HMAC链式分类账验证实际执行步骤，使攻击成功率降至0%，且误报率极低、开销微小。

Comments 8 pages, 6 figures, 5 tables. Code: https://github.com/bithabib/ai_security

RepoReasoner: Evaluating Repository-Level Code Reasoning Ability of Long-Context Language Models

RepoReasoner：评估长上下文语言模型的仓库级代码推理能力

Yanlin Wang, Suiquan Wang, Yanli Wang, Bowen Zhang, Daya Guo, Jiachi Chen, Zibin Zheng

专题命中 代码与定理证明 ：reasoning(title,abstract)

AI总结 研究针对现有代码推理基准测试局限，引入RepoReasoner评估仓库级代码推理，通过多阶段管道构建基准，评估输出预测和调用链预测能力，发现当前LLMs在仓库级推理存在局限，为未来相关研究提供方向。

Comments Published in FSE'2026

Planning to Hammer: Difficulty-Aware Decomposition for Automating Rocq Proofs

规划锤击：面向自动化Rocq证明的难度感知分解

Ning Zhang, Nongyu Di, Zenan Li, Yuan Yao, Xiaoxing Ma

专题命中 代码与定理证明 ：planning(title,abstract)

AI总结 提出Quarry框架，通过LLM规划证明分解并利用难度模型排序子目标，结合CoqHammer自动证明，在Rocq基准测试中成功率提升7%-13%。

Comments 26 pages, 8 figures; submitted to OOPSLA 2026

Theoria: Rewrite-Acceptability Verification over Informal Reasoning States

Theoria: 非正式推理状态上的重写-可接受性验证

Michael Saldivar, Ben Slivinski

机构 * Independent Researchers（独立研究者） 纠错

专题命中 代码与定理证明 ：reasoning(title)；分类 cs.CL、cs.AI、cs.LG

AI总结 提出Theoria验证架构，通过将候选解重写为带显式理由的序列化状态转换并验证变更完整性，在HLE-Verified Gold上以91.4%精确率认证105个问题，优于整体式LLM评判。

Personalized and Robust Proactive Robot Assistance with Uncertainty-Guided LLM Reasoning

个性化且鲁棒的主动机器人辅助：基于不确定性引导的大语言模型推理

Alvaro Gonzalez, M. H. Hasan Shovo, Ali Ayub

机构 * Concordia University（康考迪亚大学） 纠错

专题命中 代码与定理证明 ：reasoning(title,abstract)

AI总结 提出GLOBE框架，结合n-gram马尔可夫模型与不确定性引导的大语言模型推理，在家庭环境中实现高效鲁棒的主动机器人辅助，并在HOMER-Noise数据集上验证了其性能与效率。

Comments Accepted to the 2026 IEEE 35th International Conference on Robot and Human Interactive Communication (RO-MAN)

Proceedings 2nd International Workshop on Causal Reasoning for Embedded and safety-critical Systems Technologies

第二届嵌入式和安全关键系统技术因果推理国际研讨会论文集

Alex Groce, Stefan Leue

专题命中 代码与定理证明 ：reasoning(title,abstract)

AI总结 本文探讨了复杂嵌入式和安全关键系统中因果推理的方法，旨在促进不同领域研究者之间的交流，以提升关键系统故障根本原因分析的科学性。

Journal ref EPTCS 259, 2017

Measuring Information Burden: From Coalition-Based Reasoning to the Price System

衡量信息负担：从基于联盟的推理到价格体系

Shuige Liu

专题命中 代码与定理证明 ：reasoning(title,abstract)

AI总结 本文通过形式化框架衡量价格体系节省的信息量，证明在可转移效用博弈中，核心与一致同意等价所需的最小信息结构是每个联盟至少被其一个成员所知，且平均每代理信息负担随经济规模指数增长。

LAD-VF: LLM-Automatic Differentiation Enables Fine-Tuning-Free Robot Planning from Formal Methods Feedback

LAD-VF：LLM自动微分实现基于形式化方法反馈的无微调机器人规划

Yunhao Yang, Junyuan Hong, Gabriel Jacob Perin, Zhiwen Fan, Li Yin, Zhangyang Wang, Ufuk Topcu

机构 * The University of Texas at Austin（德克萨斯大学奥斯汀分校） ； University of São Paulo（圣保罗大学） ； Texas A&M University（德克萨斯A&M大学） ； SylphAI 纠错

专题命中 代码与定理证明 ：planning(title,abstract)

AI总结 提出LAD-VF框架，利用形式化验证反馈和LLM自动微分自动优化提示词，无需微调即可提升机器人规划任务中规范符合率，成功率从60%提升至90%以上。

Comments Presented at ICRA 2026

Guiding LLM-based Loop Invariant Synthesis via Feedback on Local Reasoning Errors

通过反馈局部推理错误引导基于LLM的循环不变式合成

Tianchi Li, Zhenyu Yan, Junhao Liu, Peng Di, Xin Zhang

专题命中 代码与定理证明 ：reasoning(title,abstract)

AI总结 本文提出了一种框架，通过形式验证LLM的思考过程并检测局部推理错误，为基于LLM的循环不变式合成提供建设性反馈，从而提高合成效果。

Comments Accepted by ACM Transactions on Programming Languages and Systems (TOPLAS). DOI: 10.1145/3806652

Journal ref ACM Trans. Program. Lang. Syst. 48, 2, Article 8 (May 2026)

Evaluating LLMs Code Reasoning Under Real-World Context

评估大语言模型在现实世界上下文中的代码推理能力

Changshu Liu

专题命中 代码与定理证明 ：reasoning(title,abstract)

AI总结 本文提出R2Eval1基准测试，包含135个来自十个广泛使用的Python项目的代码推理问题，通过序列化复合和自定义类型，更真实地评估LLM的通用性。

Comments Accepted by ICES SRC (ACM Student Research Competition)

Saliency-R1: Enforcing Interpretable and Faithful Vision-language Reasoning via Saliency-map Alignment Reward

Saliency-R1: 通过显著图对齐奖励增强可解释性和忠实的视觉-语言推理

Shizhan Gong, Minda Hu, Qiyuan Zhang, Chen Ma, Qi Dou

机构 * The Chinese University of Hong Kong（香港中文大学） ； City University of Hong Kong（香港城市大学） 纠错

专题命中 代码与定理证明 ：reasoning(title,abstract)

AI总结 Saliency-R1通过显著图对齐奖励提升视觉-语言模型的推理可解释性和忠实性，利用显著图技术高效突出关键图像区域，结合GRPO优化策略增强模型对相关区域的关注，实验显示提升推理忠实性和任务性能。

Comments CVPR 2026