大模型推理能力

Causal Sufficiency and Necessity Improves Chain-of-Thought Reasoning

Xiangning Yu, Zhuohan Wang, Linyi Yang, Haoxuan Li, Anjie Liu, Xiao Xue, Jun Wang, Mengyue Yang

机构 * Tianjin University（天津大学） ； City University of Hong Kong（香港城市大学） ； University College London（伦敦大学学院） ； Peking University（北京大学） ； The Hong Kong University of Science and Technology (Guangzhou)（香港科学与技术大学（广州）） ； University of Bristol（布里斯托大学） 纠错

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

Reasoning as an Adaptive Defense for Safety

Taeyoun Kim, Fahim Tajwar, Aditi Raghunathan, Aviral Kumar

机构 * Carnegie Mellon University（卡内基梅隆大学） 纠错

专题命中 数学推理 ：reasoning(title,abstract)；chain-of-thought(abstract)；test-time compute(abstract)；分类 cs.AI、cs.LG

Comments 44 pages, 10 Figures, 7 Tables

Thought Anchors: Which LLM Reasoning Steps Matter?

Paul C. Bogdan, Uzay Macar, Neel Nanda, Arthur Conmy

机构 * MATS 纠错

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

Comments Paul C. Bogdan and Uzay Macar contributed equally to this work, and their listed order was determined by coinflip. Neel Nanda and Arthur Conmy contributed equally to this work as senior authors, and their listed order was determined by coinflip

First SFT, Second RL, Third UPT: Continual Improving Multi-Modal LLM Reasoning via Unsupervised Post-Training

Lai Wei, Yuting Li, Chen Wang, Yue Wang, Linghe Kong, Weiran Huang, Lichao Sun

机构 * School of Computer Science, Shanghai Jiao Tong University（上海交通大学计算机科学学院） ； Zhongguancun Academy（中关村学院） ； Shanghai Innovation Institute（上海创新研究院） ； Lehigh University（莱特大学） 纠错

专题命中 数学推理 ：reasoning(title,abstract)；分类 cs.CL、cs.AI、cs.LG

Comments Accepted by NeurIPS 2025

Dipper: Diversity in Prompts for Producing Large Language Model Ensembles in Reasoning tasks

Gregory Kang Ruey Lau, Wenyang Hu, Diwen Liu, Jizhuo Chen, See-Kiong Ng, Bryan Kian Hsiang Low

机构 * National University of Singapore（新加坡国立大学） ； SAP（SAP公司） 纠错

专题命中 数学推理 ：reasoning(title,abstract)；分类 cs.CL、cs.AI、cs.LG

Comments Accepted to EMNLP 2025 Main Conference

What Makes a Good Curriculum? Disentangling the Effects of Data Ordering on LLM Mathematical Reasoning

Yaning Jia, Chunhui Zhang, Xingjian Diao, Xiangchi Yuan, Zhongyu Ouyang, Chiyu Ma, Soroush Vosoughi

机构 * Dartmouth College（达特茅斯学院） 纠错

专题命中 数学推理 ：reasoning(title,abstract)；分类 cs.AI、cs.LG

Comments 8 pages (main text) + 4 pages (appendix), 4 figures

The Surprising Effectiveness of Negative Reinforcement in LLM Reasoning

Xinyu Zhu, Mengzhou Xia, Zhepei Wei, Wei-Lin Chen, Danqi Chen, Yu Meng

机构 * Computer Science Department, University of Virginia（弗吉尼亚大学计算机科学系） ； Princeton Language and Intelligence (PLI), Princeton University（普林斯顿大学语言与智能研究所） 纠错

专题命中 数学推理 ：reasoning(title,abstract)；分类 cs.CL、cs.LG

Comments Accepted to NeurIPS 2025

Controllable Mathematical Reasoning via Self-Optimizing Thought Vectors

Xuying LI

机构 * Kepler（开普勒） 纠错

专题命中 数学推理 ：reasoning(title,abstract)；分类 cs.AI

Uni-MuMER: Unified Multi-Task Fine-Tuning of Vision-Language Model for Handwritten Mathematical Expression Recognition

Yu Li, Jin Jiang, Jianhua Zhu, Shuai Peng, Baole Wei, Yuxuan Zhou, Liangcai Gao

机构 * Wangxuan Institute of Computer Technology, Peking University（计算机技术研究院，北京大学） 纠错

专题命中 数学推理 ：reasoning(abstract)；chain-of-thought(abstract)；CoT(abstract)

Comments Accepted by NeurIPS 2025 as a spotlight

Encoder-Decoder Diffusion Language Models for Efficient Training and Inference

Marianne Arriola, Yair Schiff, Hao Phung, Aaron Gokaslan, Volodymyr Kuleshov

机构 * Department of Computer Science, Cornell University（计算机科学系，康奈尔大学） 纠错

专题命中 数学推理 ：reasoning(abstract)；分类 cs.AI、cs.LG

Comments NeurIPS 2025. We provide the code at https://github.com/kuleshov-group/e2d2

VEHME: A Vision-Language Model For Evaluating Handwritten Mathematics Expressions

Thu Phuong Nguyen, Duc M. Nguyen, Hyotaek Jeon, Hyunwook Lee, Hyunmin Song, Sungahn Ko, Taehwan Kim

专题命中 数学推理 ：reasoning(abstract)；分类 cs.CL、cs.LG

Comments EMNLP 2025. Project Website: https://vehme.github.io/

Evolving LLMs' Self-Refinement Capability via Synergistic Training-Inference Optimization

Yongcheng Zeng, Xinyu Cui, Xuanfa Jin, Qirui Mi, Guoqing Liu, Zexu Sun, Mengyue Yang, Dong Li, Weiyu Ma, Ning Yang, Jian Zhao, Jianye Hao, Haifeng Zhang, Jun Wang

机构 * Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所） ； School of Artificial Intelligence, University of Chinese Academy of Sciences（中国科学院大学人工智能学院） ； University of Science and Technology of China（中国科学技术大学） ； Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学耿家庄人工智能学院） ； University of Bristo, UK（英国布里斯托大学） ； Zhongguancun Academy, China（中关村学院） ； Huawei Noah’s Ark Lab, China（华为诺亚实验室） ； University College London, UK（伦敦大学学院） 纠错

专题命中 数学推理 ：reasoning(abstract)；分类 cs.CL、cs.LG

Foundation of Intelligence: Review of Math Word Problems from Human Cognition Perspective

Zhenya Huang, Jiayu Liu, Xin Lin, Zhiyuan Ma, Shangzi Xue, Tong Xiao, Qi Liu, Yee Whye Teh, Enhong Chen

机构 * Department of Statistics, University of Oxford（统计学系、牛津大学） 纠错

专题命中 数学推理 ：reasoning(abstract)；分类 cs.AI

LaX: Boosting Low-Rank Training of Foundation Models via Latent Crossing

Ruijie Zhang, Ziyue Liu, Zhengyang Wang, Zheng Zhang

机构 * University of California at Santa Barbara（加州大学圣芭芭拉分校） 纠错

专题命中 数学推理 ：reasoning(abstract)；分类 cs.LG

Once Upon an Input: Reasoning via Per-Instance Program Synthesis

Adam Stein, Neelay Velingker, Mayur Naik, Eric Wong

机构 * University of Pennsylvania（宾夕法尼亚大学） 纠错

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

Comments Accepted at NeurIPS 2025. 34 pages, 7 figures

MOOSE-Chem3: Toward Experiment-Guided Hypothesis Ranking via Simulated Experimental Feedback

Wanhao Liu, Zonglin Yang, Jue Wang, Lidong Bing, Di Zhang, Dongzhan Zhou, Yuqiang Li, Houqiang Li, Erik Cambria, Wanli Ouyang

机构 * University of Science and Technology of China（中国科学技术大学） ； Shanghai Artificial Intelligence Laboratory（上海人工智能实验室） ； Nanyang Technological University（南洋理工大学） ； MiroMind 纠错

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

Understanding Network Behaviors through Natural Language Question-Answering

Mingzhe Xing, Chang Tian, Jianan Zhang, Lichen Pan, Peipei Liu, Zhaoteng Yan, Yinliang Yue

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

Comments Large Language Models

AI-Enhanced Operator Assistance for UNICOS Applications

Bernard Tam, Jean-Charles Tournier, Fernando Varela Rodriguez

机构 * The University of Sydney（悉尼大学） 纠错

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

Comments Prepared as part of the CERN openlab programme 2025. Also available on Zenodo, a repository operated by CERN and co-funded by the European Union

QuestBench: Can LLMs ask the right question to acquire information in reasoning tasks?

Belinda Z. Li, Been Kim, Zi Wang

机构 * Massachusetts Institute of Technology（麻省理工学院） ； Google DeepMind（谷歌DeepMind） 纠错

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

Comments Code and dataset are available at \url{https://github.com/google-deepmind/questbench}

Journal ref 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Track on Datasets and Benchmarks

When No Paths Lead to Rome: Benchmarking Systematic Neural Relational Reasoning

Anirban Das, Irtaza Khalid, Rafael Peñaloza, Steven Schockaert

机构 * Cardiff University（卡迪夫大学） ； University of Milano-Bicocca（米兰-比科卡大学） 纠错

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

Comments accepted at NeurIPS 2025 D&B track

LLM-AR: LLM-powered Automated Reasoning Framework

Rick Chen, Joseph Ternasky, Aaron Ontoyin Yin, Xianling Mu, Fuat Alican, Yigit Ihlamur

机构 * University of Oxford（牛津大学） ； Vela Research（Vela研究公司） 纠错

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

LacMaterial: Large Language Models as Analogical Chemists for Materials Discovery

Hongyu Guo

机构 * National Research Council Canada（加拿大国家研究理事会） 纠错

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

Beyond Pipelines: A Survey of the Paradigm Shift toward Model-Native Agentic AI

Jitao Sang, Jinlin Xiao, Jiarun Han, Jilin Chen, Xiaoyi Chen, Shuyu Wei, Yongjie Sun, Yuhang Wang

机构 * Institute for Clarity in Documentation（清晰文档研究所） ； Inria Paris-Rocquencourt（巴黎- Rocquencourt 国家信息与自动化所） ； Rajiv Gandhi University（拉吉夫·甘地大学） ； Tsinghua University（清华大学） ； Palmer Research Laboratories（帕勒尔研究实验室） ； Beijing Jiaotong University（北京交通大学） 纠错

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

Grounding Methods for Neural-Symbolic AI

Rodrigo Castellano Ontiveros, Francesco Giannini, Marco Gori, Giuseppe Marra, Michelangelo Diligenti

机构 * University of Siena（锡耶纳大学） ； Scuola Normale Superiore（正规大学） ； KU Leuven（卢森堡大学） 纠错

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

Journal ref Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence (IJCAI-25), pp. 4806-4814, 2025

BTL-UI: Blink-Think-Link Reasoning Model for GUI Agent

Shaojie Zhang, Ruoceng Zhang, Pei Fu, Shaokang Wang, Jiahui Yang, Xin Du, Shiqi Cui, Bin Qin, Ying Huang, Zhenbo Luo, Jian Luan

机构 * Xiaomi Inc（小米公司） 纠错

专题命中 规划推理 ：reasoning(title,abstract)；planning(abstract)；分类 cs.AI

Comments Accepted at NeurIPS 2025

LLMs as Planning Formalizers: A Survey for Leveraging Large Language Models to Construct Automated Planning Models

Marcus Tantakoun, Xiaodan Zhu, Christian Muise

机构 * Ingenuity Labs Research Institute, Queen’s University（英杰实验室研究 institute，皇后大学） ； School of Computing, Queen’s University（计算学院，皇后大学） ； Department of Electrical and Computer Engineering, Queen’s University（电气与计算机工程学院，皇后大学） 纠错

专题命中 规划推理 ：planning(title,abstract)；reasoning(abstract)；分类 cs.AI

Comments 20 pages, 3 figures, 3 appendices

PACR: Progressively Ascending Confidence Reward for LLM Reasoning

Eunseop Yoon, Hee Suk Yoon, Jaehyun Jang, SooHwan Eom, Qi Dai, Chong Luo, Mark A. Hasegawa-Johnson, Chang D. Yoo

机构 * Korea Advanced Institute of Science and Technology (KAIST)（韩国科学技术院） ； Microsoft Research Asia (MSRA)（微软亚洲研究院） ； University of Illinois at Urbana-Champaign (UIUC)（伊利诺伊大学厄巴纳-香槟分校） 纠错

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

Comments 16 pages, 14 figures

Large language model-based task planning for service robots: A review

Shaohan Bian, Ying Zhang, Guohui Tian, Zhiqiang Miao, Edmond Q. Wu, Simon X. Yang, Changchun Hua

机构 * School of Electrical Engineering, Yanshan University（燕山大学电气工程学院） ； Engineering Research Center of Intelligent Control System and Intelligent Equipment, Ministry of Education, Yanshan University（燕山大学智能控制与设备工程研究中心） ； Hebei Key Laboratory of Intelligent Rehabilitation and Neuromodulation, Yanshan University（河北省智能康复与神经调制重点实验室） ； School of Control Science and Engineering, Shandong University（山东大学控制科学与工程学院） ； National Engineering Research Center of Robot Visual Perception and Control Technology, Hunan University（国家机器人视觉感知与控制技术工程研究中心） ； Department of Automation, Shanghai Jiao Tong University（上海交通大学自动化系） ； Advanced Robotics and Intelligent Systems Laboratory, School of Engineering, University of Guelph（圭尔夫大学工程学院先进机器人与智能系统实验室） 纠错

专题命中 规划推理 ：planning(title,abstract)

Comments Submitted to Biomimetic Intelligence and Robotics for possible publication

SceneDecorator: Towards Scene-Oriented Story Generation with Scene Planning and Scene Consistency

Quanjian Song, Donghao Zhou, Jingyu Lin, Fei Shen, Jiaze Wang, Xiaowei Hu, Cunjian Chen, Pheng-Ann Heng

机构 * Monash University（莫纳什大学） ； The Chinese University of Hong Kong（香港中文大学） ； National University of Singapore（新加坡国立大学） ； South China University of Technology（华南理工大学） 纠错

专题命中 规划推理 ：planning(title,abstract)

Comments Accepted by NeurIPS 2025; Project Page: https://lulupig12138.github.io/SceneDecorator

Learning to Better Search with Language Models via Guided Reinforced Self-Training

Seungyong Moon, Bumsoo Park, Hyun Oh Song

机构 * Seoul National University（首尔国立大学） ； KRAFTON 纠错

专题命中 规划推理 ：reasoning(abstract)；test-time compute(abstract)；分类 cs.CL、cs.AI

Comments Accepted at NeurIPS 2025