大模型推理能力

SafeThinker: Reasoning about Risk to Deepen Safety Beyond Shallow Alignment

SafeThinker: 通过风险推理深化安全对齐

Xianya Fang, Xianying Luo, Yadong Wang, Xiang Chen, Yu Tian, Zequn Sun, Rui Liu, Jun Fang, Naiqiang Tan, Yuanning Cui, Sheng-Jun Huang

机构 * Nanjing University of Aeronautics and Astronautics（南京航空航天大学） ； Didi International Business Group（滴滴国际商务集团） ； Institute for AI, Tsinghua University（清华大学人工智能研究院） ； Nanjing University of Information Science & Technology（南京信息科学技术大学） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

AI总结 SafeThinker通过动态分配防御资源，有效提升模型对伪装攻击的防御能力，同时保持实用性。

Activation-Space Anchored Access Control for Multi-Class Permission Reasoning in Large Language Models

基于激活空间锚定的多类权限推理访问控制

Zhaopeng Zhang, Pengcheng Sun, Lan Zhang, Chen Tang, Jiewei Lai, Yunhao Wang, Hui Jin

机构 * University of Science and Technology of China（中国科学技术大学） ； Lenovo Research（联想研究院） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

AI总结 本文提出AAAC框架，通过激活空间锚点实现多类权限控制，有效降低权限违规和攻击成功率。

GRASP: Generic Reasoning And SPARQL Generation across Knowledge Graphs

GRASP：跨知识图谱的通用推理与SPARQL生成

Sebastian Walter, Hannah Bast

机构 * University of Freiburg（弗赖堡大学） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

AI总结 GRASP通过大型语言模型实现跨知识图谱的通用推理与SPARQL生成，无需微调，在多个基准中达到先进水平。

Comments Accepted for publication at ISWC 2025. This version of the contribution has been accepted for publication, after peer review but is not the Version of Record. The Version of Record is available online at: https://doi.org/10.1007/978-3-032-09527-5_15

Journal ref The Semantic Web - ISWC 2025, LNCS 16140, pp. 271-289 (2026)

Data Augmented Pipeline for Legal Information Extraction and Reasoning

法律信息抽取与推理的数据增强管道

Nguyen Minh Phuong, Ha-Thanh Nguyen, May Myo Zin, Ken Satoh

机构 * Center for Juris-Informatics, ROIS-DS（司法信息中心，ROIS-DS） ； Japan Advanced Institute of Science and Technology（日本先进科学研究所） ； Research and Development Center for LLMs, NII（大语言模型研发中心，NII） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

AI总结 本文提出了一种基于大型语言模型的数据增强管道，用于提升法律领域信息抽取与推理的效率和鲁棒性，同时具备跨领域应用的通用性。

Comments Accepted in the Demonstration Track at ICAIL 2025

Sharp Eyes and Memory for VideoLLMs: Information-Aware Visual Token Pruning for Efficient and Reliable VideoLLM Reasoning

锐眼与记忆：视频大语言模型的信息感知视觉标记修剪方法

Jialong Qin, Xin Zou, Di Lu, Yibo Yan, Xuming Hu

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

AI总结 SharpV通过自适应视觉标记修剪和键值缓存修剪，提升视频大语言模型的效率与可靠性，是首个无需暴露注意力分数的两阶段修剪框架。

Comments The 40th Annual AAAI Conference on Artificial Intelligence (AAAI-26) Poster

Advancing site-specific disease and pest management in precision agriculture: From reasoning-driven foundation models to adaptive, feedback-based learning

Nitin Rai, Daeun, Choi, Nathan S. Boyd, Arnold W. Schumann

机构 * Department of Horticultural Sciences（园艺科学系） ； Gulf Coast Research and Education Center（墨西哥湾沿岸研究与教育中心） ； University of Florida（佛罗里达大学） ； Department of Agricultural and Biological Engineering（农业与生物工程系） ； Department of Soil, Water, and Ecosystem Sciences（土壤、水与生态系统科学系） ； Citrus Research and Education Center（柑橘研究与教育中心） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Comments 26 pages, 8 figures, and 2 tables

Cross-Question Method Reuse in Large Language Models: From Word-Level Prediction to Rational Logical-Layer Reasoning

Hong Su

机构 * School of Computer Science, Chengdu University of Information Technology（信息工程大学计算机科学学院） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Empowering LLM Agents with Geospatial Awareness: Toward Grounded Reasoning for Wildfire Response

Yiheng Chen, Lingyao Li, Zihui Ma, Qikai Hu, Yilun Zhu, Min Deng, Runlong Yu

机构 * University of Alabama（阿拉巴马大学） ； University of South Florida（佛罗里达州立大学） ； New York University（纽约大学） ； University of Michigan（密歇根大学） ； Texas Tech University（得克萨斯科技大学） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Multi-Agent Collaborative Intelligence: Dual-Dial Control for Reliable LLM Reasoning

Edward Y. Chang, Ethan Y. Chang

机构 * Stanford University（斯坦福大学） ； UIUC（伊利诺伊大学香槟分校） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Comments 27 pages, 5 figures, 21 tables

D$^2$HScore: Reasoning-Aware Hallucination Detection via Semantic Breadth and Depth Analysis in LLMs

Yue Ding, Xiaofang Zhu, Tianze Xia, Junfei Wu, Xinlong Chen, Qiang Liu, Liang Wang

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Comments under review

LLM world models are mental: Output layer evidence of brittle world model use in LLM mechanical reasoning

Cole Robertson, Philip Wolff

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Comments Manuscript comprises 14 pages, 4 figures, 4 tables in the Technical Appendix and Supplementary Material, and is under review at NeurIPS 2025

Decoupled Seg Tokens Make Stronger Reasoning Video Segmenter and Grounder

Dang Jisheng, Wu Xudong, Wang Bimei, Lv Ning, Chen Jiayu, Jingwen Zhao, Yichu liu, Jizhao Liu, Juncheng Li, Teng Wang

机构 * Lanzhou University（兰州大学） ； National University of Singapore（新加坡国立大学） ； Sun Yat-sen University（中山大学） ； Jinan University（济南大学） ； South China University of Technology（华南理工大学） ； Zhejiang University（浙江大学） ； The University of Hong Kong（香港大学） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Causal Explanations Over Time: Articulated Reasoning for Interactive Environments

Sebastian Rödling, Matej Zečević, Devendra Singh Dhami, Kristian Kersting

机构 * Technical University of Darmstadt（德累斯顿技术大学） ； Eindhoven University of Technology（埃因霍温理工大学） ； Hessian Center for Artificial Intelligence（黑森人工智能中心） ； German Research Center for Artificial Intelligence（德国人工智能研究中心） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Comments Main paper: 9 pages, References: 2 pages, Supplementary: 9 pages. Number of figures: 10, number of tables: 3

ASPO: Adaptive Sentence-Level Preference Optimization for Fine-Grained Multimodal Reasoning

Yeyuan Wang, Dehong Gao, Rujiao Long, Lei Yi, Linbo Jin, Libin Yang, Xiaoyan Cai

机构 * Northwestern Polytechnical University, School of Automation, Xi’an, China（西北工业大学自动化学院，西安，中国） ； Northwestern Polytechnical University, School of Cybersecurity, Xi’an, China（西北工业大学网络安全学院，西安，中国） ； Binjiang Institute of Artificial Intelligence, ZJUT, Hangzhou, China（浙大宁波理工学院滨海人工智能研究院，杭州，中国） ； Alibaba Group, Hangzhou, China（阿里巴巴集团，杭州，中国） 纠错

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Comments Accepted by ACL 2025 findings

RECIPE-TKG: From Sparse History to Structured Reasoning for LLM-based Temporal Knowledge Graph Completion

Ömer Faruk Akgül, Feiyu Zhu, Yuxin Yang, Rajgopal Kannan, Viktor Prasanna

专题命中 其他推理 ：reasoning(title)；分类 cs.LG

LLM-Assisted Proactive Threat Intelligence for Automated Reasoning

Shuva Paul, Farhad Alemi, Richard Macwan

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Comments 10 Pages, 1 Figure

Rule-Guided Feedback: Enhancing Reasoning by Enforcing Rule Adherence in Large Language Models

Aissatou Diallo, Antonis Bikakis, Luke Dickens, Anthony Hunter, Rob Miller

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Role-Play Paradox in Large Language Models: Reasoning Performance Gains and Ethical Dilemmas

Jinman Zhao, Zifan Qian, Linbo Cao, Yining Wang, Yitian Ding, Yulan Hu, Zeyu Zhang, Zeyong Jin

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Comments 9 pages, 7 figures, 3 tables, submitted to CogSci 2025

Rules, Cases, and Reasoning: Positivist Legal Theory as a Framework for Pluralistic AI Alignment

Nicholas A. Caputo

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Comments NeurIPS Pluralistic Alignment Workshop 2024

Large Language Models are In-context Teachers for Knowledge Reasoning

Jiachen Zhao, Zonghai Yao, Zhichao Yang, Hong Yu

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Comments EMNLP 24 Findings

Communicate to Play: Pragmatic Reasoning for Efficient Cross-Cultural Communication in Codenames

Isadora White, Sashrika Pandey, Michelle Pan

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Towards an Analysis of Discourse and Interactional Pragmatic Reasoning Capabilities of Large Language Models

Amelie Robrecht, Judith Sieker, Clara Lachenmaier, Sina Zarieß, Stefan Kopp

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

MORE: Multi-mOdal REtrieval Augmented Generative Commonsense Reasoning

Wanqing Cui, Keping Bi, Jiafeng Guo, Xueqi Cheng

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Comments Published as a conference paper at ACL Findings 2024

Fast Chain-of-Thought: A Glance of Future from Parallel Decoding Leads to Answers Faster

Hongxuan Zhang, Zhining Liu, Yao Zhao, Jiaqi Zheng, Chenyi Zhuang, Jinjie Gu, Guihai Chen

专题命中 其他推理 ：chain-of-thought(title)；分类 cs.CL

Advancing Legal Reasoning: The Integration of AI to Navigate Complexities and Biases in Global Jurisprudence with Semi-Automated Arbitration Processes (SAAPs)

Michael De'Shazer

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Sasha: Creative Goal-Oriented Reasoning in Smart Homes with Large Language Models

Evan King, Haoxiang Yu, Sangsu Lee, Christine Julien

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Comments To appear in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (March 2024)

An Approach to Solving the Abstraction and Reasoning Corpus (ARC) Challenge

Tan John Chong Min

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Comments 14 pages

Transformer Working Memory Enables Regular Language Reasoning and Natural Language Length Extrapolation

Ta-Chung Chi, Ting-Han Fan, Alexander I. Rudnicky, Peter J. Ramadge

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Learning and Reasoning Multifaceted and Longitudinal Data for Poverty Estimates and Livelihood Capabilities of Lagged Regions in Rural India

Atharva Kulkarni, Raya Das, Ravi S. Srivastava, Tanmoy Chakraborty

专题命中 其他推理 ：reasoning(title)；分类 cs.CL

Comments Accepted to IJCAI 2023 Main Conference (AI for Social Good Track)

Advances in apparent conceptual physics reasoning in GPT-4

Colin G. West

专题命中 其他推理 ：reasoning(title)；分类 cs.AI

Comments 5 pages, one figure one table. arXiv admin note: text overlap with arXiv:2303.01067 (longer, prior version of this project)