大模型推理能力

Visual Access Boundaries in Vision-Language Model Reasoning

视觉语言模型推理中的视觉访问边界

Hiroto Osaka, Shohei Taniguchi, Gouki Minegishi, Kai Yamashita, Masahiro Suzuki, Yutaka Matsuo

机构 * The University of Tokyo（东京大学） 纠错

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

AI总结 研究视觉语言模型推理中CoT提示扩展的内容，通过视觉访问扫描定义VAB，发现CoT不主要靠延长图像令牌访问提效，而是扩展语言侧计算，且CoT收益受感知读出限制，瓶颈在读出非计数。

VeryTrace: Verifying Reasoning Traces through Compilable Formalism and Structured Verification

VeryTrace: 通过可编译形式化与结构化验证验证推理轨迹

Ninghan Zhong, Ahmet Ege Tanriverdi, Kaan Kale, Sriram Vishwanath

机构 * School of Electrical and Computer Engineering, Georgia Institute of Technology, USA（佐治亚理工学院电子与计算机工程学院） ； Department of Electrical and Computer Engineering, Bogazici University, Turkey（博亚奇大学电子与计算机工程系） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；CoT(abstract,abstract_cn)；chain-of-thought(abstract)；planning(abstract)

AI总结 提出VeryTrace框架，将自然语言推理轨迹形式化为可编译的领域特定语言（DSL），结合确定性检查与LLM审计进行混合验证，实现步骤级错误定位与修复，在数学、机器人规划、亲属推理任务上提升准确率。

Comments Accepted at LM4Plan Workshop @ ICML 2026

Beyond Symbolic Solving: Multi Chain-of-Thought Voting for Geometric Reasoning in Large Language Models

超越符号求解：用于大语言模型几何推理的多链式思维投票

Md. Abu Bakor Siddique, Shahrin Hossain, Sadman Ahmed Siam, Syed Rifat Raiyan, Hasan Mahmud, Md Kamrul Hasan

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

AI总结 本文提出MARS-GPS方法，通过生成多个并行推理流程并结合Python代码执行，利用熵值进行排序并多阶段投票，提升大语言模型在几何推理中的性能，实验显示其在Geometry3K数据集上达到88.8%的准确率，比现有最佳水平提升11%。

Comments Under review, 4 figures, 7 tables

The Kinetics of Reasoning: How Chain-of-Thought Shapes Learning in Transformers?

Zihan Pengmei, Costas Mavromatis, Zhengyuan Shen, Yunyi Zhang, Vassilis N. Ioannidis, Huzefa Rangwala

机构 * The University of Chicago（芝加哥大学） ； Amazon Web Services（亚马逊网络服务） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract)；分类 cs.AI、cs.LG

Comments 10 pages, 7 figures, with appendix

Rethinking Reasoning in LLMs: Neuro-Symbolic Local RetoMaton Beyond ICL and CoT

Rushitha Santhoshi Mamidala, Anshuman Chhabra, Ankur Mali

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

Temporalizing Confidence: Evaluation of Chain-of-Thought Reasoning with Signal Temporal Logic

Zhenjiang Mao, Artem Bisliouk, Rohith Reddy Nama, Ivan Ruchkin

机构 * University of Florida（佛罗里达大学） ； University of Mannheim（曼海姆大学） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；chain-of-thought(title,abstract)；CoT(abstract)；分类 cs.AI、cs.LG

Enigmata: Scaling Logical Reasoning in Large Language Models with Synthetic Verifiable Puzzles

Jiangjie Chen, Qianyu He, Siyu Yuan, Aili Chen, Zhicheng Cai, Weinan Dai, Hongli Yu, Qiying Yu, Xuefeng Li, Jiaze Chen, Hao Zhou, Mingxuan Wang

机构 * ByteDance Seed（字节跳动种子） ； Fudan University（复旦大学） ； Institute for AI Industry Research (AIR), Tsinghua University（人工智能产业研究院（AIR），清华大学） ； Nanjing University（南京大学） ； Shanghai Jiao Tong University（上海交通大学） ； SIA-Lab of Tsinghua AIR and ByteDance Seed（清华大学AIR实验室和字节跳动种子） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；verifier(abstract)；分类 cs.CL、cs.AI

CoT is Not True Reasoning, It Is Just a Tight Constraint to Imitate: A Theory Perspective

Jintian Shao, Yiming Cheng

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

Comments Last page bad picture, lacking some sufficient experiments

Multi-LogiEval: Towards Evaluating Multi-Step Logical Reasoning Ability of Large Language Models

Nisarg Patel, Mohith Kulkarni, Mihir Parmar, Aashna Budhiraja, Mutsumi Nakamura, Neeraj Varshney, Chitta Baral

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

Comments Accepted at EMNLP 2024 Main

Deciphering the Factors Influencing the Efficacy of Chain-of-Thought: Probability, Memorization, and Noisy Reasoning

Akshara Prabhakar, Thomas L. Griffiths, R. Thomas McCoy

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

Comments EMNLP 2024 Findings; 9 pages plus references and appendices

Thought-Like-Pro: Enhancing Reasoning of Large Language Models through Self-Driven Prolog-based Chain-of-Thought

Xiaoyu Tan, Yongxin Deng, Xihe Qiu, Weidi Xu, Chao Qu, Wei Chu, Yinghui Xu, Yuan Qi

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

Assessing Logical Reasoning Capabilities of Encoder-Only Transformer Models

Paulo Pirozelli, Marcos M. José, Paulo de Tarso P. Filho, Anarosa A. F. Brandão, Fabio G. Cozman

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

LogicBench: Towards Systematic Evaluation of Logical Reasoning Ability of Large Language Models

Mihir Parmar, Nisarg Patel, Neeraj Varshney, Mutsumi Nakamura, Man Luo, Santosh Mashetty, Arindam Mitra, Chitta Baral

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

Comments Accepted at ACL(Main) 2024 | First version available @ https://openreview.net/forum?id=7NR2ZVzZxx

Towards LogiGLUE: A Brief Survey and A Benchmark for Analyzing Logical Reasoning Capabilities of Language Models

Man Luo, Shrinidhi Kumbhar, Ming shen, Mihir Parmar, Neeraj Varshney, Pratyay Banerjee, Somak Aditya, Chitta Baral

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

Comments Work in progress

Logic-LM: Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning

Liangming Pan, Alon Albalak, Xinyi Wang, William Yang Wang

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL、cs.AI

Comments EMNLP 2023 (Findings, long paper)

The Reasoning Trap -- Logical Reasoning as a Mechanistic Pathway to Situational Awareness

推理陷阱——逻辑推理作为情境意识的机制路径

Subramanyam Sahoo, Aman Chadha, Vinija Jain, Divya Chaudhary

机构 * MARS 4.0 Fellowship, Cambridge AI Safety Hub(CAISH), University of Cambridge（MARS 4.0 Fellow，剑桥人工智能安全中心（CAISH），剑桥大学） ； AWS Generative AI Innovation Center, Amazon Web Services, USA（亚马逊生成AI创新中心，亚马逊网络服务，美国） ； Google, USA（谷歌，美国） ； Stanford University（斯坦福大学） ； Northeastern University, Seattle, WA, USA（东北大学，西雅图，华盛顿州，美国） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；分类 cs.CL、cs.AI、cs.LG

AI总结 本文提出RAISE框架，揭示逻辑推理能力提升与情境意识升级的机制路径，并提出安全原则与测试方法以应对潜在风险。

Comments Accepted at ICLR 2026 Workshop on Logical Reasoning of Large Language Models. 21 Pages. Position Paper

Mamba-SSM with LLM Reasoning for Feature Selection: Faithfulness-Aware Biomarker Discovery

Mamba-SSM结合LLM推理进行特征选择：面向生物标志物发现的可信度意识方法

Pushpa Kumar Balan, Aijing Feng

机构 * Department of Computer Science and Cybersecurity（计算机科学与网络安全系） 纠错

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

AI总结 本文研究LLM推理在过滤特征选择中的作用，通过Mamba-SSM提取基因并结合CoT提升预测性能，发现特征筛选能提高AUC并验证部分已知生物标志物。

Comments 9 pages, 4 figures. Accepted at ICLR 2026 Workshop on Logical Reasoning of Large Language Models

Agentified Assessment of Logical Reasoning Agents

具有代理特性的逻辑推理代理评估

Zhiyu Ni, Yifeng Xiao, Zheng Liang

机构 * University of California, Berkeley（加州大学伯克利分校） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.AI

AI总结 本文提出一种评估和基准测试逻辑推理代理的框架，通过代理评估实现可重复、可审计和抗执行失败的评估。案例研究中，自动形式化代理在FOLIO验证集上以86.70%的准确率超越了基线方法。

Comments Accepted at ICLR 2026 Agents in the Wild (AIWILD) Workshop. 5 pages, 2 figures, 1 table

Improving Chain-of-Thought Reasoning via Quasi-Symbolic Abstractions

通过准符号抽象改进链式推理

Leonardo Ranaldi, Marco Valentino, Andrè Freitas

机构 * Idiap Research Institute（Idiap 研究所） ； School of Informatics, University of Edinburgh（爱丁堡大学信息学院） ； School of Computer Science, University of Sheffield（谢菲尔德大学计算机科学学院） ； Department of Computer Science, University of Manchester（曼彻斯特大学计算机科学系） ； National Biomarker Centre (NBC), CRUK Manchester Institute（英国癌症研究中心曼彻斯特研究所国家生物标志物中心） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；chain-of-thought(title)；CoT(abstract)；logical reasoning(abstract)

AI总结 本文提出QuaSAR方法，通过准符号解释引导LLM在更高抽象层面推理，提升小模型的推理能力，实验显示在自然语言和符号推理任务中准确率提升8%。

Journal ref 2025.acl-long.843

$π$-CoT: Prolog-Initialized Chain-of-Thought Prompting for Multi-Hop Question-Answering

$π$-CoT：基于Prolog初始化的多跳问答链式思维提示法

Chao Wan, Albert Gong, Mihir Mishra, Carl-Leander Henneking, Claas Beger, Kilian Q. Weinberger

机构 * cornell（康奈尔大学） 纠错

专题命中 逻辑推理 ：chain-of-thought(title,abstract)；CoT(title,abstract)；reasoning(abstract)；分类 cs.CL

AI总结 $π$-CoT通过结合逻辑编程与语言模型，提升多跳问答性能，优于传统RAG和CoT方法。

REASON: Accelerating Probabilistic Logical Reasoning for Scalable Neuro-Symbolic Intelligence

REASON: 加速概率逻辑推理以实现可扩展的神经符号智能

Zishen Wan, Che-Kai Liu, Jiayi Qian, Hanchen Yang, Arijit Raychowdhury, Tushar Krishna

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；planning(abstract)；分类 cs.AI

AI总结 REASON通过统一图表示和自适应剪枝技术，实现概率逻辑推理的加速，提升神经符号AI的效率与可扩展性。

Comments 16 pages, 13 figures, 5 tables, 2026 IEEE International Symposium on High-Performance Computer Architecture (HPCA)

Can Pruning Improve Reasoning? Revisiting Long-CoT Compression with Capability in Mind for Better Reasoning

剪枝能否提升推理？基于能力对齐的Long-Cot压缩以获得更好的推理

Shangziqi Zhao, Jiahao Yuan, Jinyang Wu, Zhenglin Wang, Guisong Yang, Usman Naseem

机构 * XJTU（西安交通大学） ； ECNU（华东师范大学） ； THU（清华大学） ； SEU（上海师范大学） ； USST（上海师范大学） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；CoT(title,abstract)；chain-of-thought(abstract)；分类 cs.CL

AI总结 本文提出Prune-on-Logic框架，通过结构感知剪枝提升Long-CoT推理效率，发现验证剪枝在保持准确性的同时减少令牌使用，揭示剪枝与模型容量对齐的重要性。

Comments 19 pages,6 figures

A Large Language Model Based Method for Complex Logical Reasoning over Knowledge Graphs

基于大型语言模型的复杂知识图谱逻辑推理方法

Ziyan Zhang, Chao Wang, Zhuo Chen, Lei Chen, Chiyi Li, Kai Song

机构 * School of Information Science and Engineering, Chongqing Jiaotong University（信息科学与工程学院，重庆交通大学） ； State Grid Chongqing Electric Power Company（国网重庆市电力公司） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL

AI总结 本文提出ROG方法，结合KG邻域检索与LLM推理，有效提升复杂知识图谱逻辑推理性能。

The Curse of CoT: On the Limitations of Chain-of-Thought in In-Context Learning

Tianshi Zheng, Yixiang Chen, Chengxi Li, Chunyang Li, Qing Zong, Haochen Shi, Baixuan Xu, Yangqiu Song, Ginny Y. Wong, Simon See

机构 * The Hong Kong University of Science and Technology（香港科技大学） ； NVIDIA（英伟达） 纠错

专题命中 逻辑推理 ：chain-of-thought(title,abstract)；CoT(title,abstract)；reasoning(abstract)；分类 cs.CL

Comments Accepted by TMLR

Enhancing Logical Reasoning in Language Models via Symbolically-Guided Monte Carlo Process Supervision

Xingwei Tan, Marco Valentino, Mahmud Akhter, Maria Liakata, Nikolaos Aletras

机构 * School of Computer Science, University of Sheffield（谢菲尔德大学计算机科学学院） ； School of Electronic Engineering and Computer Science, Queen Mary University of London（伦敦大学玛丽女王学院电子工程与计算机科学学院） ； The Alan Turing Institute（艾伦·图灵研究所） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；planning(abstract)；分类 cs.CL

Comments EMNLP 2025 (Main), 9+6 pages

GreenMind: A Next-Generation Vietnamese Large Language Model for Structured and Logical Reasoning

Luu Quy Tung, Hoang Quoc Viet, Pham Bao Loc, Vo Trong Thu

机构 * John Von Neumann Institute（约翰·冯·诺依曼研究所） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title)；chain-of-thought(abstract)；CoT(abstract)

Divide and Translate: Compositional First-Order Logic Translation and Verification for Complex Logical Reasoning

Hyun Ryu, Gyeongman Kim, Hyemin S. Lee, Eunho Yang

机构 * KAIST（韩国科学技术院） ； MIT（麻省理工学院） 纠错

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL

Comments ICLR 2025 camera-ready version

Journal ref The Thirteenth International Conference on Learning Representations (ICLR 2025)

Learning to Reason via Mixture-of-Thought for Logical Reasoning

Tong Zheng, Lichang Chen, Simeng Han, R. Thomas McCoy, Heng Huang

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL

Comments 38 pages

Large Language Models Meet Symbolic Provers for Logical Reasoning Evaluation

Chengwen Qi, Ren Ma, Bowen Li, He Du, Binyuan Hui, Jinwang Wu, Yuanjun Laili, Conghui He

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title)；chain-of-thought(abstract)；CoT(abstract)

Comments Accepted by ICLR 2025

ARN: Analogical Reasoning on Narratives

Zhivar Sourati, Filip Ilievski, Pia Sommerauer, Yifan Jiang

专题命中 逻辑推理 ：reasoning(title,abstract)；logical reasoning(title,abstract)；chain-of-thought(abstract)；分类 cs.CL