大模型推理能力

Modular Design Patterns for Hybrid Learning and Reasoning Systems: a taxonomy, patterns and use cases

Michael van Bekkum, Maaike de Boer, Frank van Harmelen, André Meyer-Vitali, Annette ten Teije

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

Comments 20 pages, 22 figures, accepted for publication in the International Journal of Applied Intelligence

DeceptionX: From Multimodal Evidence to Explainable Deception Detection

DeceptionX: 基于多模态大语言模型的可解释欺骗检测

Jiayu Zhang, Shuo Ye, Jiajian Huang, Yawen Cui, Taorui Wang, Wei Xia, Zeheng Wang, Haowen Tang, Yelin Wang, Hui Ma, Zitong Yu

机构 * Great Bay University（大湾区大学） ； Hong Kong Polytechnic University（香港理工大学） 纠错

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

AI总结 提出DeceptionX框架，将欺骗检测从黑箱分类转变为可解释的观察-思考-总结推理过程，通过构建DeceptChain数据集和三阶段训练管道，在标准基准上超越现有方法，同时提供专家级可解释推理路径。

Logic-Guided Socially-aware Robot Navigation World Model

逻辑引导的社会感知机器人导航世界模型

Weizheng Wang, Obi Ike, Soyun Choi, Sungeun Hong, Aniket Bera, Byung-Cheol Min

机构 * School of Applied and Creative Computing, Purdue University（应用与创意计算学院，普渡大学） ； Department of Computer Science, Purdue University（计算机科学系，普渡大学） ； Department of Applied Artificial Intelligence, Sungkyunkwan University（应用人工智能系，成均馆大学） ； Department of Computer Science and Department of Intelligent Systems Engineering, Indiana University Bloomington（计算机科学系和智能系统工程系，印第安纳大学布卢明顿分校） 纠错

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

AI总结 提出NaviWM，通过结合结构化世界模型和逻辑驱动推理链，增强大语言模型在动态人类空间中生成社交合规且物理安全的导航决策的能力。

Comments The authors have decided to withdraw this manuscript due to concerns regarding its current scope, framing, and presentation. Please do not cite this version

Video Streaming Thinking: VideoLLMs Can Watch and Think Simultaneously

视频流思考：VideoLLMs可以同时观看和思考

Yiran Guan, Liang Yin, Dingkang Liang, Jianzhong Ju, Zhenbo Luo, Jian Luan, Yuliang Liu, Xiang Bai

机构 * Huazhong University of Science and Technology（华中科技大学） ； MiLM Plus Xiaomi Inc.（小米公司） 纠错

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

AI总结 VST提出了一种视频流理解的新范式，通过在流式过程中激活推理，提高实时响应和理解能力，同时在多个基准测试中表现出更高的效率和泛化能力。

Comments Accepted by ECCV 2026, project page https://1ranguan.github.io/VST/

Semantic-based Internet of Embodied Intelligence: Visions and Frontiers

基于语义的具身智能物联网：愿景与前沿

Yaheng Wang, Rui Meng, Xiaodong Xu, Yiming Liu, Feiliang Song, Linyuan Hu, Huishi Song, Lexi Xu, Tony Q. S. Quek, Ping Zhang

专题命中 逻辑推理 ：reasoning(abstract)；planning(abstract)；logical reasoning(abstract)

AI总结 提出基于语义的具身智能物联网（SIoEI），通过语义信息统一度量解决多智能体网络中的多模态数据传输开销和逻辑推理与物理约束解耦问题，并验证其在信道鲁棒性和端到端延迟方面的优势。

Comments 7 pages, 4 figures

Decoupling Semantics and Logic: A Training-Free Coarse-to-Fine Pipeline for Video Retrieval-Augmented Generation

解耦语义与逻辑：一种无需训练的从粗到精的视频检索增强生成流水线

Jiaxin Dai, Zehang Wei, Jiamin Yan, Xiang Xiang

机构 * School of Computer Science & Tech, Huazhong University of Science and Technology（华中科技大学计算机科学与技术学院） ； School of AI and Automation, Huazhong University of Science and Technology（华中科技大学人工智能与自动化学院） 纠错

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

AI总结 提出一种无需训练的两阶段级联视频RAG流水线，通过解耦语义检索与逻辑推理，实现跨语言长视频理解、严格角色遵循和零幻觉时间定位。

Comments To be presented at ACL 2026 MAGMAR Workshop (Oral; Retrieval leaderboard No.1)

PowerFlow: Unlocking the Dual Nature of LLMs via Principled Distribution Matching

PowerFlow: 通过原则性分布匹配释放LLMs的双重特性

Ruishuo Chen, Yu Chen, Zhuoran Li, Longbo Huang

机构 * Institute for Interdisciplinary Information Sciences, Tsinghua University, Beijing, China（清华大学交叉信息研究院） 纠错

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

AI总结 提出PowerFlow框架，将无监督微调重构成分布匹配问题，利用GFlowNet和长度感知轨迹平衡目标，通过调整α-幂分布方向性激发LLMs的逻辑推理或创造性。

Comments Camera-ready version accepted at ICML 2026

SciAtlas: A Large-Scale Knowledge Graph for Automated Scientific Research

SciAtlas：面向自动化科学研究的大规模知识图谱

Shuofei Qiao, Yunxiang Wei, Jiazheng Fan, Bin Wu, Busheng Zhang, Mengru Wang, Yuqi Zhu, Ningyu Zhang, Keyan Ding, Qiang Zhang, Huajun Chen

机构 * Zhejiang University（浙江大学） ； University College London（伦敦大学学院） 纠错

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

AI总结 为应对学术信息爆炸和跨学科整合困难，提出包含4300万论文、1.57亿实体和30亿三元组的多学科知识图谱SciAtlas，并开发神经符号检索算法，实现从语义匹配到确定性关联发现的转变，降低推理成本。

Comments Ongoing Work

LLM-Enabled Low-Altitude UAV Natural Language Navigation via Signal Temporal Logic Specification Translation and Repair

基于大语言模型的低空无人机自然语言导航：通过信号时序逻辑规范翻译与修复

Yuqi Ping, Huahao Ding, Tianhao Liang, Longyu Zhou, Guangyu Lei, Xinglin Chen, Junwei Wu, Jieyu Zhou, Tingting Zhang

机构 * Guangdong Provincial Key Laboratory of Space-Aerial Networking and Intelligent Sensing, Harbin Institute of Technology, Shenzhen（广东省空天网络与智能感知重点实验室，哈尔滨工业大学（深圳）） ； Peng Cheng Laboratory (PCL), Shenzhen（鹏城实验室） ； Information Systems Technology and Design, Singapore University of Technology and Design（新加坡科技设计大学信息系统技术与设计系） ； School of Computer Science and Engineering, Central South University, Changsha（中南大学计算机科学与工程学院） 纠错

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

AI总结 本文提出一种统一框架，通过将自然语言指令转换为信号时序逻辑规范并利用混合整数线性规划生成轨迹，提升低空无人机在复杂环境中的安全导航能力。

Interpretable Logical Anomaly Classification via Constraint Decomposition and Instruction Fine-Tuning

通过约束分解和指令微调实现可解释的逻辑异常分类

Xufei Zhang, Xinjiao Zhou, Ziling Deng, Dongdong Geng, Jianxiong Wang

机构 * Beijing XingYun Digital Technology Co., Ltd.（北京星云数字技术有限公司） 纠错

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

AI总结 LogiCls通过约束分解和指令微调实现工业逻辑异常的可解释分类，结合视觉-语言框架和数据驱动的监督方法，提升异常检测的准确性和可解释性。

Comments 6 pages, 6 figures

Error Taxonomy-Guided Prompt Optimization

误差分类引导的提示优化

Mayank Singh, Vikas Yadav, Eduardo Blanco

机构 * University of Arizona（亚利桑那大学） 纠错

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

AI总结 ETGPO通过自顶向下方法，利用误差分类指导提示优化，提升模型性能并降低计算成本。

Towards Sample-Efficient and Stable Reinforcement Learning for LLM-based Recommendation

面向LLM推荐的高效稳定强化学习

Hongxun Ding, Keqin Bao, Jizhi Zhang, Yi Fang, Wenxin Xu, Fuli Feng, Xiangnan He

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

AI总结 本文提出RISER框架，通过强化学习提升LLM推荐的效率和稳定性，有效解决样本效率低和训练不稳定问题。

LogicGuard: Improving Embodied LLM agents through Temporal Logic based Critics

Anand Gokhale, Vaibhav Srivastava, Francesco Bullo

机构 * Department of Mechanical Engineering, University of California at Santa Barbara（加州大学圣芭芭拉分校机械工程系） ； Department of Electrical and Computer Engineering, Michigan State University（密歇根州立大学电气与计算机工程系） 纠错

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

Comments Modified version of prior LTLCrit work with new robotics dataset

Visual Programmability: A Guide for Code-as-Thought in Chart Understanding

Bohao Tang, Yan Ma, Fei Zhang, Jiadi Su, Ethan Chern, Zhulin Hu, Zhixin Wang, Pengfei Liu, Ya Zhang

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

R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimization

Jingyi Zhang, Jiaxing Huang, Huanjin Yao, Shunyu Liu, Xikun Zhang, Shijian Lu, Dacheng Tao

机构 * Nanyang Technological University（南洋理工大学） 纠错

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

Comments ICCV 2025 Camera Ready

It's Not That Simple. An Analysis of Simple Test-Time Scaling

Guojun Wu

机构 * Independent Researcher（独立研究者） 纠错

专题命中 逻辑推理 ：CoT(abstract)；test-time compute(abstract)；分类 cs.CL、cs.AI、cs.LG

VectorLLM: Human-like Extraction of Structured Building Contours vis Multimodal LLMs

Tao Zhang, Shiqing Wei, Shihao Chen, Wenling Yu, Muying Luo, Shunping Ji

机构 * School of Remote Sensing and Information Engineering（遥感与信息工程学院） ； College of Oceanography and Space Informatics（海洋学与空间信息学院） ； School of Surveying and Geoinformation Engineering（测绘与地理信息工程学院） 纠错

专题命中 逻辑推理 ：reasoning(abstract)；planning(abstract)；logical reasoning(abstract)

LAD-Reasoner: Tiny Multimodal Models are Good Reasoners for Logical Anomaly Detection

Weijia Li, Guanglei Chu, Jiong Chen, Guo-Sen Xie, Caifeng Shan, Fang Zhao

专题命中 逻辑推理 ：reasoning(abstract)；CoT(abstract)；logical reasoning(abstract)

Multimodal Long Video Modeling Based on Temporal Dynamic Context

Haoran Hao, Jiaming Han, Yiyuan Zhang, Xiangyu Yue

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

Logicbreaks: A Framework for Understanding Subversion of Rule-based Inference

Anton Xue, Avishree Khare, Rajeev Alur, Surbhi Goel, Eric Wong

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

Federated In-Context LLM Agent Learning

Panlong Wu, Kangshuo Li, Junbao Nan, Fangxin Wang

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

Unleash LLMs Potential for Recommendation by Coordinating Twin-Tower Dynamic Semantic Token Generator

Jun Yin, Zhengxin Zeng, Mingzheng Li, Hao Yan, Chaozhuo Li, Weihao Han, Jianjin Zhang, Ruochen Liu, Allen Sun, Denvy Deng, Feng Sun, Qi Zhang, Shirui Pan, Senzhang Wang

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

ThinkRepair: Self-Directed Automated Program Repair

Xin Yin, Chao Ni, Shaohua Wang, Zhenhao Li, Limin Zeng, Xiaohu Yang

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

Comments Accepted By ISSTA'24

SelfPiCo: Self-Guided Partial Code Execution with LLMs

Zhipeng Xue, Zhipeng Gao, Shaohua Wang, Xing Hu, Xin Xia, Shanping Li

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

Comments Accepted by ISSTA'24

Easy Problems That LLMs Get Wrong

Sean Williams, James Huckle

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

Comments AutogenAI Ltd. GitHub Repo: https://github.com/autogenai/easy-problems-that-llms-get-wrong

Prompting Is All You Need: Automated Android Bug Replay with Large Language Models

Sidong Feng, Chunyang Chen

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

Comments 46th International Conference on Software Engineering (ICSE 2024)

Large Language User Interfaces: Voice Interactive User Interfaces powered by LLMs

Syed Mekael Wasti, Ken Q. Pu, Ali Neshati

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

Comments Accepted as peer-reviewed publication

RePLan: Robotic Replanning with Perception and Language Models

Marta Skreta, Zihan Zhou, Jia Lin Yuan, Kourosh Darvish, Alán Aspuru-Guzik, Animesh Garg

专题命中 逻辑推理 ：reasoning(abstract)；planning(abstract)；logical reasoning(abstract)

A Comprehensive Survey on Pretrained Foundation Models: A History from BERT to ChatGPT

Ce Zhou, Qian Li, Chen Li, Jun Yu, Yixin Liu, Guangjing Wang, Kai Zhang, Cheng Ji, Qiben Yan, Lifang He, Hao Peng, Jianxin Li, Jia Wu, Ziwei Liu, Pengtao Xie, Caiming Xiong, Jian Pei, Philip S. Yu, Lichao Sun

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

Comments 99 pages, 16 figures

Advancing Natural Language Formalization to First Order Logic with Fine-tuned LLMs

通过微调LLMs推进自然语言形式化到一阶逻辑

Felix Vossel, Till Mossakowski, Björn Gehrke

机构 * Osnabrück University（奥斯纳布鲁克大学） ； University of Zurich（苏黎世大学） 纠错

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

AI总结 本文通过微调LLMs提升自然语言到一阶逻辑的转换性能，实验显示谓词可用性显著提升模型表现，T5模型优于大解码器-only模型，并能泛化至未见逻辑论证。

Comments 15 pages, 7 tables, accepted at the International Joint Conference on Learning & Reasoning (IJCLR 2025)