大模型推理能力 - arXivDaily 专题

2505.03335 2025-10-17 cs.LG cs.AI cs.CL 82%

Absolute Zero: Reinforced Self-play Reasoning with Zero Data

Andrew Zhao, Yiran Wu, Yang Yue, Tong Wu, Quentin Xu, Yang Yue, Matthieu Lin, Shenzhi Wang, Qingyun Wu, Zilong Zheng, Gao Huang

机构 * Tsinghua University（清华大学）； Beijing Institute for General Artificial Intelligence（北京通用人工智能研究院）； Penn State University（宾夕法尼亚州立大学）

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

2510.10104 2025-10-14 cs.CV 82%

Answer-Consistent Chain-of-thought Reinforcement Learning For Multi-modal Large Langauge Models

Minbin Huang, Runhui Huang, Chuanyang Zheng, Jingyao Li, Guoxuan Chen, Han Shi, Hong Cheng

机构 * The Chinese University of Hong Kong（香港中文大学）； The University of Hong Kong（香港大学）； Huawei Noah’s Ark Lab（华为诺亚实验室）

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

2510.07358 2025-10-10 cs.LG cs.AI cs.CL 82%

Encode, Think, Decode: Scaling test-time reasoning with recursive latent thoughts

Yeskendir Koishekenov, Aldo Lipani, Nicola Cancedda

机构 * FAIR at Meta（Meta的FAIR）； University College London（伦敦大学学院）

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

2505.22203 2025-10-08 cs.LG cs.AI cs.CL 82%

From Accuracy to Robustness: A Study of Rule- and Model-based Verifiers in Mathematical Reasoning

Yuzhen Huang, Weihao Zeng, Xingshan Zeng, Qi Zhu, Junxian He

机构 * The Hong Kong University of Science and Technology（香港科学与技术大学）； The Chinese University of Hong Kong（香港中文大学）； Tsinghua University（清华大学）

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

2510.01165 2025-10-02 cs.CL cs.AI cs.LG 82%

GRAD: Generative Retrieval-Aligned Demonstration Sampler for Efficient Few-Shot Reasoning

Oussama Gabouj, Kamel Charaf, Ivan Zakazov, Nicolas Baldwin, Robert West

机构 * EPFL（苏黎世联邦理工学院）

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

Comments EMNLP 2025 (findings)

2509.25133 2025-09-30 cs.LG cs.AI cs.CL 82%

Rethinking Entropy Regularization in Large Reasoning Models

Yuxian Jiang, Yafu Li, Guanxu Chen, Dongrui Liu, Yu Cheng, Jing Shao

机构 * Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）； Fudan University（复旦大学）； Shanghai Jiao Tong University（上海交通大学）； Chinese University of Hong Kong（香港中文大学）

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

2509.24147 2025-09-30 cs.CL cs.AI cs.LG 82%

Your thoughts tell who you are: Characterize the reasoning patterns of LRMs

Yida Chen, Yuning Mao, Xianjun Yang, Suyu Ge, Shengjie Bi, Lijuan Liu, Saghar Hosseini, Liang Tan, Yixin Nie, Shaoliang Nie

机构 * Harvard University（哈佛大学）； Meta Superintelligence Labs（Meta超智能实验室）

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

Comments 32 pages, 28 figures

2503.18866 2025-09-30 cs.LG cs.AI cs.CL 82%

Reasoning to Learn from Latent Thoughts

Yangjun Ruan, Neil Band, Chris J. Maddison, Tatsunori Hashimoto

机构 * Stanford University（斯坦福大学）； University of Toronto（多伦多大学）； Vector Institute（向量研究所）

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

2509.21070 2025-09-26 cs.LG cs.AI cs.CL 82%

ScaleDiff: Scaling Difficult Problems for Advanced Mathematical Reasoning

Qizhi Pei, Zhuoshi Pan, Honglin Lin, Xin Gao, Yu Li, Zinan Tang, Conghui He, Rui Yan, Lijun Wu

机构 * Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学耿丽学校）； OpenDataLab, Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）； Tsinghua University（清华大学）； School of Artificial Intelligence, Wuhan University（武汉大学人工智能学院）

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

Comments 15 pages

2509.17318 2025-09-25 cs.AI cs.CL cs.LG 82%

CogAtom: From Cognitive Atoms to Olympiad-level Mathematical Reasoning in Large Language Models

Zhuofan Chen, Jiyuan He, Yichi Zhang, Xing Hu, Haoxing Wen, Jun Bai, Wenge Rong

机构 * School of Computer Science and Engineering, Beihang University, China（北京航空航天大学计算机科学与工程学院）； Meituan Inc.（美团公司）； Beijing Institute for General Artificial Intelligence, China（北京通用人工智能研究院）

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

2502.10454 2025-08-25 cs.LG cs.AI cs.CL 82%

One Example Shown, Many Concepts Known! Counterexample-Driven Conceptual Reasoning in Mathematical LLMs

Yinghui Li, Jiayi Kuang, Haojing Huang, Zhikun Xu, Xinnian Liang, Yi Yu, Wenlian Lu, Yangning Li, Xiaoyu Tan, Chao Qu, Ying Shen, Hai-Tao Zheng, Philip S. Yu

机构 * Tsinghua University（清华大学）； Sun-Yat Sen University（孙中山大学）； School of Mathematical Science, Fudan University（复旦大学数学学院）； Bytedance Inc.（字节跳动公司）； INFLY TECH (Shanghai) Co., Ltd.（INFLY TECH（上海）有限公司）； ARC Lab, Arizona State University（ARC实验室，亚利桑那州立大学）； University of Illinois Chicago（伊利诺伊大学芝加哥分校）； Peng Cheng Laboratory（鹏城实验室）； Guangdong Provincial Key Laboratory of Fire Science（广东省消防科学与智能应急技术重点实验室）

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

Comments ICML 2025

2504.19483 2025-07-28 cs.LG cs.AI cs.CL 82%

Improving Reasoning Performance in Large Language Models via Representation Engineering

Bertram Højer, Oliver Jarvis, Stefan Heinrich

机构 * Department of Computer Science, IT University of Copenhagen（计算机科学系，丹麦技术大学）

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

Comments Has been accepted at "The Thirteenth International Conference on Learning Representations (ICLR 2025)" Link to publication: https://openreview.net/forum?id=IssPhpUsKt

Journal ref 13th International Conference on Learning Representations (ICLR 2025)

2507.08806 2025-07-15 cs.AI cs.CL cs.LG 82%

Think Clearly: Improving Reasoning via Redundant Token Pruning

Daewon Choi, Jimin Lee, Jihoon Tack, Woomin Song, Saket Dingliwal, Sai Muralidhar Jayanthi, Bhavana Ganesh, Jinwoo Shin, Aram Galstyan, Sravan Babu Bodapati

机构 * KAIST（韩国科学技术院）； Korea University（韩国大学）； Amazon AGI（亚马逊人工智能实验室）

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

2506.18330 2025-06-26 cs.LG cs.AI cs.CL 82%

Confucius3-Math: A Lightweight High-Performance Reasoning LLM for Chinese K-12 Mathematics Learning

Lixin Wu, Na Cai, Qiao Cheng, Jiachen Wang, Yitao Duan

机构 * NetEase Youdao（网易有道）

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

2506.19767 2025-06-25 cs.CL cs.AI cs.LG 82%

SRFT: A Single-Stage Method with Supervised and Reinforcement Fine-Tuning for Reasoning

Yuqian Fu, Tinghong Chen, Jiajun Chai, Xihuai Wang, Songjun Tu, Guojun Yin, Wei Lin, Qichao Zhang, Yuanheng Zhu, Dongbin Zhao

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）； Meituan（美团）； Shanghai Jiao Tong University（上海交通大学）

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

2410.13857 2025-06-24 cs.LG cs.AI cs.CL stat.ML 82%

How Numerical Precision Affects Arithmetical Reasoning Capabilities of LLMs

Guhao Feng, Kai Yang, Yuntian Gu, Xinyue Ai, Shengjie Luo, Jiacheng Sun, Di He, Zhenguo Li, Liwei Wang

机构 * State Key Laboratory of General Artificial Intelligence, School of Intelligence Science and Technology, Peking University（通用人工智能国家重点实验室，智能科学与技术学院，北京大学）； School of EECS, Peking University（电子工程系，北京大学）； Huawei Noah’s Ark Lab（华为诺亚实验室）； Center for Machine Learning Research, Peking University（机器学习研究中心，北京大学）

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

Comments 40 pages, 4 figures, ACL 2025 Findings

2506.13923 2025-06-23 cs.LG cs.AI cs.CL 82%

Adaptive Guidance Accelerates Reinforcement Learning of Reasoning Models

Vaskar Nath, Elaine Lau, Anisha Gunjal, Manasi Sharma, Nikhil Baharte, Sean Hendryx

机构 * Scale AI

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

2506.14965 2025-06-19 cs.LG cs.AI cs.CL 82%

Revisiting Reinforcement Learning for LLM Reasoning from A Cross-Domain Perspective

Zhoujun Cheng, Shibo Hao, Tianyang Liu, Fan Zhou, Yutao Xie, Feng Yao, Yuexin Bian, Yonghao Zhuang, Nilabjo Dey, Yuheng Zha, Yi Gu, Kun Zhou, Yuqi Wang, Yuan Li, Richard Fan, Jianshu She, Chengqian Gao, Abulhair Saparov, Haonan Li, Taylor W. Killian, Mikhail Yurochkin, Zhengzhong Liu, Eric P. Xing, Zhiting Hu

机构 * Carnegie Mellon University（卡内基梅隆大学）； Purdue University（普渡大学）

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

Comments 38 pages, 9 figures. Under review

2506.13284 2025-06-17 cs.CL cs.AI cs.LG 82%

AceReason-Nemotron 1.1: Advancing Math and Code Reasoning through SFT and RL Synergy

Zihan Liu, Zhuolin Yang, Yang Chen, Chankyu Lee, Mohammad Shoeybi, Bryan Catanzaro, Wei Ping

机构 * NVIDIA

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

Comments The AceReason-Nemotron collection: https://huggingface.co/collections/nvidia/acereason-682f4e1261dc22f697fd1485

2506.09967 2025-06-17 cs.CL cs.AI cs.LG 82%

Resa: Transparent Reasoning Models via SAEs

Shangshang Wang, Julian Asilis, Ömer Faruk Akgül, Enes Burak Bilgin, Ollie Liu, Deqing Fu, Willie Neiswanger

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

2506.10987 2025-06-16 cs.SE 82%

Chain of Draft for Software Engineering: Challenges in Applying Concise Reasoning to Code Tasks

Shaoyi Yang

专题命中数学推理：reasoning(title,abstract)；CoT(abstract)

2506.08935 2025-06-11 cs.CL cs.AI cs.LG 82%

Can A Gamer Train A Mathematical Reasoning Model?

Andrew Shin

机构 * Keio University（庆应大学）

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

2501.18858 2025-06-10 cs.LG cs.AI cs.CL 82%

BRiTE: Bootstrapping Reinforced Thinking Process to Enhance Language Model Reasoning

Han Zhong, Yutong Yin, Shenao Zhang, Xiaojun Xu, Yuanxin Liu, Yifei Zuo, Zhihan Liu, Boyi Liu, Sirui Zheng, Hongyi Guo, Liwei Wang, Mingyi Hong, Zhaoran Wang

机构 * Peking University（北京大学）； Northwestern University（西北大学）； Bytedance Inc（字节跳动公司）； University of Minnesota（明尼苏达大学）

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

Comments ICML 2025

2505.16400 2025-06-06 cs.LG cs.AI cs.CL 82%

AceReason-Nemotron: Advancing Math and Code Reasoning through Reinforcement Learning

Yang Chen, Zhuolin Yang, Zihan Liu, Chankyu Lee, Peng Xu, Mohammad Shoeybi, Bryan Catanzaro, Wei Ping

机构 * NVIDIA

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

Comments Add pass@1024 evaluation results for LiveCodeBench v6. We release the models at: https://huggingface.co/collections/nvidia/acereason-682f4e1261dc22f697fd1485

2505.22591 2025-05-29 cs.CL cs.AI cs.LG 82%

Self-Error-Instruct: Generalizing from Errors for LLMs Mathematical Reasoning

Erxin Yu, Jing Li, Ming Liao, Qi Zhu, Boyang Xue, Minghui Xu, Baojun Wang, Lanqing Hong, Fei Mi, Lifeng Shang

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

Comments 16 pages, 9 figures

2412.06559 2025-05-27 cs.AI cs.CL cs.LG 82%

ProcessBench: Identifying Process Errors in Mathematical Reasoning

Chujie Zheng, Zhenru Zhang, Beichen Zhang, Runji Lin, Keming Lu, Bowen Yu, Dayiheng Liu, Jingren Zhou, Junyang Lin

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

Comments ACL 2025

2505.17250 2025-05-26 cs.CL cs.AI cs.LG 82%

ConciseRL: Conciseness-Guided Reinforcement Learning for Efficient Reasoning Models

Razvan-Gabriel Dumitru, Darius Peteleaza, Vikas Yadav, Liangming Pan

机构 * First Author Affiliation（第一作者机构）； Second Author Affiliation（第二作者机构）

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

Comments 25 pages, 18 figures, and 6 tables

2505.13417 2025-05-20 cs.CL cs.AI cs.LG 82%

AdaptThink: Reasoning Models Can Learn When to Think

Jiajie Zhang, Nianyi Lin, Lei Hou, Ling Feng, Juanzi Li

机构 * Tsinghua University（清华大学）

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

2505.03981 2025-05-09 cs.AI cs.CL cs.LG 82%

X-Reasoner: Towards Generalizable Reasoning Across Modalities and Domains

Qianchu Liu, Sheng Zhang, Guanghui Qin, Timothy Ossowski, Yu Gu, Ying Jin, Sid Kiblawi, Sam Preston, Mu Wei, Paul Vozila, Tristan Naumann, Hoifung Poon

机构 * Microsoft Research（微软研究院）

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

2504.20708 2025-04-30 cs.CL cs.AI cs.LG 82%

Beyond the Last Answer: Your Reasoning Trace Uncovers More than You Think

Hasan Abed Al Kader Hammoud, Hani Itani, Bernard Ghanem

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

Comments Preprint

1. 数学推理 3208 篇

Absolute Zero: Reinforced Self-play Reasoning with Zero Data

Answer-Consistent Chain-of-thought Reinforcement Learning For Multi-modal Large Langauge Models

Encode, Think, Decode: Scaling test-time reasoning with recursive latent thoughts

From Accuracy to Robustness: A Study of Rule- and Model-based Verifiers in Mathematical Reasoning

GRAD: Generative Retrieval-Aligned Demonstration Sampler for Efficient Few-Shot Reasoning

Rethinking Entropy Regularization in Large Reasoning Models

Your thoughts tell who you are: Characterize the reasoning patterns of LRMs

Reasoning to Learn from Latent Thoughts

ScaleDiff: Scaling Difficult Problems for Advanced Mathematical Reasoning

CogAtom: From Cognitive Atoms to Olympiad-level Mathematical Reasoning in Large Language Models

One Example Shown, Many Concepts Known! Counterexample-Driven Conceptual Reasoning in Mathematical LLMs

Improving Reasoning Performance in Large Language Models via Representation Engineering

Think Clearly: Improving Reasoning via Redundant Token Pruning

Confucius3-Math: A Lightweight High-Performance Reasoning LLM for Chinese K-12 Mathematics Learning

SRFT: A Single-Stage Method with Supervised and Reinforcement Fine-Tuning for Reasoning

How Numerical Precision Affects Arithmetical Reasoning Capabilities of LLMs

Adaptive Guidance Accelerates Reinforcement Learning of Reasoning Models

Revisiting Reinforcement Learning for LLM Reasoning from A Cross-Domain Perspective

AceReason-Nemotron 1.1: Advancing Math and Code Reasoning through SFT and RL Synergy

Resa: Transparent Reasoning Models via SAEs

Chain of Draft for Software Engineering: Challenges in Applying Concise Reasoning to Code Tasks

Can A Gamer Train A Mathematical Reasoning Model?

BRiTE: Bootstrapping Reinforced Thinking Process to Enhance Language Model Reasoning

AceReason-Nemotron: Advancing Math and Code Reasoning through Reinforcement Learning

Self-Error-Instruct: Generalizing from Errors for LLMs Mathematical Reasoning

ProcessBench: Identifying Process Errors in Mathematical Reasoning

ConciseRL: Conciseness-Guided Reinforcement Learning for Efficient Reasoning Models

AdaptThink: Reasoning Models Can Learn When to Think

X-Reasoner: Towards Generalizable Reasoning Across Modalities and Domains

Beyond the Last Answer: Your Reasoning Trace Uncovers More than You Think