大模型推理能力

Chain-of-Thought for Large Language Model-empowered Wireless Communications

Xudong Wang, Jian Zhu, Ruichen Zhang, Lei Feng, Dusit Niyato, Jiacheng Wang, Hongyang Du, Shiwen Mao, Zhu Han

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

Comments 7 pages, 5 figures

FCoReBench: Can Large Language Models Solve Challenging First-Order Combinatorial Reasoning Problems?

Chinmay Mittal, Krishna Kartik, Mausam, Parag Singla

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

Nash CoT: Multi-Path Inference with Preference Equilibrium

Ziqi Zhang, Cunxiang Wang, Xiong Xiao, Yue Zhang, Donglin Wang

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

Journal ref EMNLP 2024

Unveiling Narrative Reasoning Limits of Large Language Models with Trope in Movie Synopses

Hung-Ting Su, Ya-Ching Hsu, Xudong Lin, Xiang-Qian Shi, Yulei Niu, Han-Yuan Hsu, Hung-yi Lee, Winston H. Hsu

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

Comments EMNLP 2024 Findings. The first two authors contributed equally. Code: https://github.com/Shelley1214/Trope

Language models show human-like content effects on reasoning tasks

Ishita Dasgupta, Andrew K. Lampinen, Stephanie C. Y. Chan, Hannah R. Sheahan, Antonia Creswell, Dharshan Kumaran, James L. McClelland, Felix Hill

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

Comments Published version of record: https://academic.oup.com/pnasnexus/article/3/7/pgae233/7712372

Disentangling Logic: The Role of Context in Large Language Model Reasoning Capabilities

Wenyue Hua, Kaijie Zhu, Lingyao Li, Lizhou Fan, Shuhang Lin, Mingyu Jin, Haochen Xue, Zelong Li, JinDong Wang, Yongfeng Zhang

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

Comments 22 pages, 9 figures

Learning Deductive Reasoning from Synthetic Corpus based on Formal Logic

Terufumi Morishita, Gaku Morio, Atsuki Yamaguchi, Yasuhiro Sogawa

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

Journal ref Proceedings of the 40th International Conference on Machine Learning, PMLR 202:25254-25274, 2023

Calc-X and Calcformers: Empowering Arithmetical Chain-of-Thought through Interaction with Symbolic Systems

Marek Kadlčík, Michal Štefánik, Ondřej Sotolář, Vlastimil Martinek

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

Comments Published in EMNLP 2023: Main track

Structured, flexible, and robust: benchmarking and improving large language models towards more human-like behavior in out-of-distribution reasoning tasks

Katherine M. Collins, Catherine Wong, Jiahai Feng, Megan Wei, Joshua B. Tenenbaum

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

Comments Originally accepted to the 2022 Cognitive Science (CogSci) conference

Attention as Binding: A Vector-Symbolic Perspective on Transformer Reasoning

注意力作为绑定：一种向量符号视角下的Transformer推理

Sahil Rajesh Dhayalkar

机构 * Sahil Rajesh Dhayalkar（独立研究者） 纠错

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

AI总结 本文提出将Transformer的注意力机制视为软向量符号计算，通过向量符号架构视角解释其推理行为，并提出改进架构和训练目标以提升逻辑可靠性和可解释性。

Comments 12 pages with references. Submitted to 'Logical and Symbolic Reasoning in Language Models @ AAAI 2026' conference and is under review

Towards a Neurosymbolic Reasoning System Grounded in Schematic Representations

François Olivier, Zied Bouraoui

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

Comments To appear in Proceedings of Machine Learning Research, 19th Conference on Neurosymbolic Learning and Reasoning, 2025

JARVIS: A Neuro-Symbolic Commonsense Reasoning Framework for Conversational Embodied Agents

Kaizhi Zheng, Kaiwen Zhou, Jing Gu, Yue Fan, Jialu Wang, Zonglin Di, Xuehai He, Xin Eric Wang

机构 * University of California（加州大学） 纠错

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

Comments 19th International Conference on Neurosymbolic Learning and Reasoning

CBR with Commonsense Reasoning and Structure Mapping: An Application to Mediation

Atilim Gunes Baydin, Ramon Lopez de Mantaras, Simeon Simoff, Carles Sierra

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

Comments 15 pages, 3 figures; updated copyright notice

Journal ref Case-Based Reasoning Research and Development, LNCS (LNAI), Volume 6880 (2011), 378-392, Springer Berlin / Heidelberg, ISBN: 978-3-642-23290-9

Long SKILL Compliance as Logical Reasoning: Closure-Grounded Detection with Scaling-Guided On-Policy Distillation

长SKILL合规性检测作为逻辑推理：基于闭包的检测与尺度引导的在线策略蒸馏

Shuaitao Zhao, Feng Ni, Lichao Ma, Jiaye Lin, Fei Han, Yang Wei, Lu Pan

机构 * Meituan（美团） 纠错

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

AI总结 针对长SKILL合规性检测的成本与精度矛盾，提出SkillCDG框架，结合两级检索与依赖闭包实现检测，在多数据集上优于基线，还发现尺度趋势以优化小模型性能。

Think Deep, Speak Once: Relit, A Recursive Latent Implicit Transformer Framework

深度思考，一次输出：ReLIT，一种递归隐式隐式Transformer框架

Abhishek Panwar, Maheep Singh, Saksham Bansal

机构 * Indian Institute of Technology Roorkee（鲁尔基印度理工学院） 纠错

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

AI总结 该研究针对思维链提示计算开销大、微型递归模型语义连贯性差的问题，提出ReLIT框架，结合TinyLlama-1.1B骨干与可训练递归块，在逻辑推理基准上高效实现推理，性能优于或媲美更大模型。

Comments 14 pages, 6 figures

ChaosBench-Logic: A Benchmark for Logical and Symbolic Reasoning on Chaotic Dynamical Systems

ChaosBench-Logic：逻辑和符号推理在混沌动力系统上的基准

Noel Thomas

机构 * Noel Thomas 纠错

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

AI总结 ChaosBench-Logic通过30种混沌动力系统评估LLM逻辑和符号推理能力，揭示其在复杂推理任务中的局限性。

Comments 7 pages, 0 figures , Accepted to AAAI-26 Bridge Program: Logical and Symbolic Reasoning in Language Models (camera-ready)

Journal ref AAAI 2026 Bridge Program on Logical and Symbolic Reasoning in Language Models, Singapore, Jan 2026

Reasoning in Neurosymbolic AI

Son Tran, Edjard Mota, Artur d'Avila Garcez

机构 * School of Information Technology, Deakin University（信息科技学院，德堪大学） ； Instituto de Computação, Universidade Federal do Amazonas（计算学院，亚马逊联邦大学） 纠错

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

Comments 50 pages, 13 figures, 56 references. Keywords: Neurosymbolic AI, Restricted Boltzmann Machines, Logical Reasoning, SAT solving, MaxSAT, Energy-based Learning, Constrained Optimization, Modular Deep Learning

Logic Haystacks: Probing LLMs Long-Context Logical Reasoning (Without Easily Identifiable Unrelated Padding)

Damien Sileo

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

STAIR: Semantic-Temporal Automaton for Interpretable Reasoning in Temporal Question Answering

STAIR：用于时间问答中可解释推理的语义-时间自动机

Xinlong Dai, Jinchuan Zhang, Lei Gao, Xinzhe Hu, Yuefeng He, Hui Gao

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

AI总结 本文提出STAIR语义-时间自动机，将语义解读与时间推理分离，规则优先设计减少自由推理，在多个时间问答数据集上优于基线，提升了F1值且可解释性更强。

Attend to Your Own Thoughts: Breaking the Barrier for Post-Training Quantization of Reasoning LLMs through the Lens of 1.58-Bit Quantization

关注自身思维：通过1.58比特量化视角打破推理型大语言模型的后训练量化障碍

Shigeng Wang, Chao Li, Yangyuxuan Kang, Jiawei Fan, Anbang Yao

机构 * Intel Labs China（英特尔中国实验室） 纠错

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

AI总结 该研究提出ScaleQ-1.58三值后训练量化框架，通过集成AYOT校准方法，提升推理型LLM量化性能，该框架可扩展且泛化性强，仅需少量校准token即可实现优异效果。

Comments This research work was completed and submitted for publication in early May 2026. The project page: https://github.com/IntelChina-AI/BitTern

Do Diagrams Help Large Language Models Reason? Evidence from Syllogistic Reasoning

图表有助于大语言模型推理吗？来自三段论推理的证据

Risako Ando, Koji Mineshima

机构 * Keio University（庆应义塾大学） 纠错

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

AI总结 研究三段论推理中图表对大语言模型的影响，比较自然语言、逻辑符号、线性图表和欧拉图四种表示条件，利用285个问题评估两个模型，发现图表不能始终提升性能，模型在中性问题上表现不佳，从图表中获益有限。

Comments To appear in the Proceedings of the 15th International Conference on the Theory and Application of Diagrams (Diagrams 2026)

Bridging Symbolic Control and Neural Reasoning in LLM Agents -- The Structured Cognitive Loop

连接符号控制与神经推理的LLM代理——结构认知循环

Myung Ho Kim

机构 * JEI University（JEI大学） 纠错

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

AI总结 本文提出结构认知循环(SCL)架构，通过分离检索、认知、控制、行动和记忆模块，解决LLM代理的推理与执行纠缠、记忆波动和行动序列失控问题，实现零策略违规、防止冗余调用并保持决策可追溯性。

Comments This update clarifies the theoretical architecture by separating Regulation as the Soft Symbolic Control layer from Control as a deterministic runtime engine, while adding explicit discussion of how the current implementation should be interpreted in light of that distinction

Evaluating Implicit Biases in LLM Reasoning through Logic Grid Puzzles

通过逻辑网格谜题评估LLM推理中的隐性偏见

Fatima Jahara, Mark Dredze, Sharon Levy

机构 * Rutgers University（罗格斯大学） ； Johns Hopkins University（约翰霍普金斯大学） 纠错

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

AI总结 提出PRIME框架，利用逻辑网格谜题系统探测LLM在复杂推理中受社会刻板印象的影响，发现模型在解与刻板印象一致时推理更准确。

Comments 26 pages (including appendix)

Position: Reasoning After Perception Means Reasoning Without Vision

立场：感知之后推理意味着无视觉推理

Hongcheng Gao, Zihao Huang, Jingyi Tang, Lin Xu, Xinhao Li, Haoyang Li, Yue Liu, Minhua Lin, Xinlong Yang, Taihang Hu, Ge Wu, Balong Bi, Hongyu Chen, Olive Huang, Wentao Zhang

机构 * Tsinghua University（清华大学） ； Peking University（北京大学） ； National University of Singapore（新加坡国立大学） ； Zhongguancun Academy（中关村学院） ； Nanjing University（南京大学） ； Nankai University（南开大学） 纠错

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

AI总结 本文质疑多模态研究中语言推理能弥补视觉缺陷的假设，提出“感知-推理”范式导致视觉推理退化为文本空间推理，并通过图灵眼测试验证文本推理无法修复感知失败，主张转向感知内的推理。

Multi-Granularity Reasoning for Natural Language Inference

自然语言推理的多粒度推理

Chunling Xi, Di Liang

机构 * University of Science and Technology of China（中国科学技术大学） 纠错

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

AI总结 提出多粒度推理网络（MGRN），通过分层语义特征交互模拟人类认知过程，在多个基准上超越强基线模型。

Reinforcement-aware Knowledge Distillation for LLM Reasoning

面向LLM推理的强化学习感知知识蒸馏

Zhaoyang Zhang, Shuli Jiang, Yantao Shen, Yuting Zhang, Dhananjay Ram, Shuo Yang, Zhuowen Tu, Wei Xia, Stefano Soatto

机构 * Meta ； Guo et al. ； Lin et al. ； Xu et al. ； Shao et al. ； Schulman et al. ； Xie et al. 纠错

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

AI总结 提出RL感知蒸馏（RLAD），通过信任区域比率蒸馏（TRRD）在强化学习后训练中实现选择性模仿，解决分布不匹配和目标干扰问题，在逻辑推理和数学基准上优于现有方法。

Decoupling Thought from Speech: Knowledge-Grounded Counterfactual Reasoning for Resilient Multi-Agent Argumentation

思想与言语解耦：基于知识反事实推理的鲁棒多智能体辩论

Jakub Masłowski, Jarosław A. Chudziak

机构 * Institute of Computer Science, Warsaw University of Technology（华沙技术大学计算机科学学院） 纠错

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

AI总结 提出知识反事实推理（KG-CFR）双阶段架构，通过私有规划缓冲与公共执行层分离，在动态资源分配环境下将扰动后论证质量从0.694提升至0.822，并减少语义循环。

Comments Accepted for publication in the Proceedings of the 30th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems (KES 2026)

Pramana: Fine-Tuning Large Language Models for Epistemic Reasoning through Navya-Nyaya

Pramana: 通过 Navya-Nyaya 微调大型语言模型进行认知推理

Sharath Sathish

机构 * University of York（约克大学） 纠错

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

AI总结 提出 Pramana 方法，利用 2500 年历史的印度 Navya-Nyaya 逻辑框架微调 LLM，通过结构化六阶段推理解决认知差距，提升可溯源性。

Comments 52 pages + appendices, comprehensive treatment of Navya-Nyaya computational formalization

Domain Adaptation and Reasoning Frameworks in Language Models: A Controlled Experiment with Historical Cosmology

语言模型中的领域适应与推理框架：以历史宇宙学为受控实验

Francesco De Bernardis

机构 * Independent Researcher（独立研究者） 纠错

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

AI总结 通过历史宇宙学受控实验，研究领域适应如何重塑语言模型的解释行为，发现适应主要改变解释框架而非直接改变立场。

Comments 17 pages, 3 figures

Symbolic Intermediaries as a Linguistic-Numerical Interface for LLM-Driven Geometric Reasoning

符号中介作为LLM驱动几何推理的语言-数值接口

João Pedro Gandarela, Thiago Rios, Stefan Menzel, André Freitas

机构 * Idiap Research Institute（Idiap研究 institute） ； École Polytechnique Fédérale de Lausanne（瑞士联邦理工学院） ； Honda Research Institute Europe（本田欧洲研究院） ； Department of Computer Science, University of Manchester（曼彻斯特大学计算机科学系） ； National Biomarker Centre, CRUK-MI, University of Manchester（曼彻斯特大学国家生物标记中心） 纠错

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

AI总结 提出符号中介作为连接物理模拟器数值输出与语言模型推理的接口，通过符号回归将连续数值转化为符号表达式，并在协同优化循环中提升几何推理性能。

Comments 33 pages, 18 figures