大模型推理能力

Learning Neuro-Symbolic Relational Transition Models for Bilevel Planning

Rohan Chitnis, Tom Silver, Joshua B. Tenenbaum, Tomas Lozano-Perez, Leslie Pack Kaelbling

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

Comments IROS 2022 final version

MRKL Systems: A modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoning

Ehud Karpas, Omri Abend, Yonatan Belinkov, Barak Lenz, Opher Lieber, Nir Ratner, Yoav Shoham, Hofit Bata, Yoav Levine, Kevin Leyton-Brown, Dor Muhlgay, Noam Rozen, Erez Schwartz, Gal Shachaf, Shai Shalev-Shwartz, Amnon Shashua, Moshe Tenenholtz

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

The Argument Reasoning Comprehension Task: Identification and Reconstruction of Implicit Warrants

Ivan Habernal, Henning Wachsmuth, Iryna Gurevych, Benno Stein

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

Comments Accepted as NAACL 2018 Long Paper; see details on the front page

Neuro-Symbolic Forward Reasoning

Hikaru Shindo, Devendra Singh Dhami, Kristian Kersting

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

Comments Preprint

Leveraging the Inductive Bias of Large Language Models for Abstract Textual Reasoning

Christopher Michael Rytting, David Wingate

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

Fixed Point Semantics for Stream Reasoning

Christian Antić

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

Neural-Symbolic Relational Reasoning on Graph Models: Effective Link Inference and Computation from Knowledge Bases

Henrique Lemos, Pedro Avelar, Marcelo Prates, Luís Lamb, Artur Garcez

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

Comments Under review: ICANN 2020

Neural Assistant: Joint Action Prediction, Response Generation, and Latent Knowledge Reasoning

Arvind Neelakantan, Semih Yavuz, Sharan Narang, Vishaal Prasad, Ben Goodrich, Daniel Duckworth, Chinnadhurai Sankar, Xifeng Yan

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

AI Reasoning Systems: PAC and Applied Methods

Jeffrey Cheng

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

Comments 26 pages

Polynomial-time probabilistic reasoning with partial observations via implicit learning in probability logics

Brendan Juba

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

Comments Presented in Eighth International Workshop on Statistical Relational AI (STARAI 2018)

Grounding LLM Reasoning under Incomplete Graph Evidence

在不完全图证据下 grounding LLM 推理

Jiaqi Li, Fanghui Song

机构 * Tianjin Normal University, College of Computer and Information Engineering（天津师范大学计算机与信息工程学院） ； Harbin Institute of Technology, School of Mathematics（哈尔滨工业大学数学学院） 纠错

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

AI总结 本文提出在不完全知识图谱证据下，通过KL正则化变形LLM先验实现软grounding，并给出稳定性界限，适用于GraphRAG、KGQA等场景。

Comments A theoretical perspective about Grounding LLM Reasoning

TS-Agent: Understanding and Reasoning Over Raw Time Series via Iterative Insight Gathering

TS-Agent：通过迭代洞察获取理解并推理原始时间序列

Penghang Liu, Elizabeth Fons, Annita Vapsi, Mohsen Ghassemi, Svitlana Vyetrenko, Daniel Borrajo, Vamsi K. Potluru, Manuela Veloso

机构 * JPMorgan AI Research, NY, USA（摩根大通人工智能研究院，纽约，美国） 纠错

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

AI总结 TS-Agent通过迭代洞察获取机制，在时间序列理解和推理任务中匹配或超越文本、视觉及时间序列语言模型基线，尤其在零样本设置下表现更优。

Comments NeurIPS 2025 Workshop on Foundations of Reasoning in Language Models

Generative Logic: A New Computer Architecture for Deterministic Reasoning and Knowledge Generation

生成逻辑：一种新的计算机架构用于确定性推理和知识生成

Nikolai Sergeev

机构 * Generative Logic UG 纠错

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

AI总结 生成逻辑（GL）是一种确定性架构，通过用户提供的公理定义生成可配置的演绎邻域，通过逻辑块和统一的哈希推理引擎实现证明生成与知识生成，实验验证其在数论中的应用及CAS系统潜力。

Comments v4: Incubator, Compressor, Verifier (34,320 checks, 0 failures). New CAS chapter. Pipeline diagram. Branching outlook, FTA campaign, CAS roadmap, LLM demo in Future Work. Updated MPL listing and runtimes. 24pp, 8 figs. Zenodo DOI: 10.5281/zenodo.17206386

Imandra CodeLogician: Neuro-Symbolic Reasoning for Precise Analysis of Software Logic

Imandra CodeLogician: 用于软件逻辑精确分析的神经符号推理

Hongyu Lin, Samer Abdallah, Makar Valentinov, Paul Brennan, Elijah Kagan, Christoph M. Wintersteiger, Denis Ignatovich, Grant Passmore

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

AI总结 Imandra CodeLogician通过神经符号推理提升软件逻辑分析的精确性，结合形式方法与LLM，实现更严谨的程序验证与分析。

Comments 52 pages, 23 figures. Includes a new benchmark dataset (code-logic-bench) and evaluation of neurosymbolic reasoning for software analysis

DESIGNER: Design-Logic-Guided Multidisciplinary Data Synthesis for LLM Reasoning

DESIGNER: 多学科数据合成用于LLM推理的设计逻辑引导

Weize Liu, Yongchi Zhao, Yijia Luo, Mingyu Xu, Jiaheng Liu, Yanan Li, Xiguo Hu, Zhiqi Bai, Yuchi Xu, Wenbo Su, Bo Zheng

机构 * Alibaba Group（阿里巴巴集团） ； Nanjing University（南京大学） 纠错

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

AI总结 DESIGNER通过设计逻辑引导生成多学科推理数据集，提升LLM跨学科推理能力。

Comments Accepted to ICLR 2026. Project page: https://attention-is-all-i-need.github.io/Design-Logic-Reasoning

Proceedings of the Second International Workshop on Next-Generation Language Models for Knowledge Representation and Reasoning (NeLaMKRR 2025)

Ha-Thanh Nguyen, Ken Satoh, Francesca Toni, Randy Goebel, Kostas Stathis

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

Comments Associated with the 22nd International Conference on Principles of Knowledge Representation and Reasoning (KR 2025) in Melbourne, Australia

SQLForge: Synthesizing Reliable and Diverse Data to Enhance Text-to-SQL Reasoning in LLMs

Yu Guo, Dong Jin, Shenghao Ye, Shuangwu Chen, Jian Yang, Xiaobin Tan

机构 * University of Science and Technology of China（中国科学技术大学） ； Institute of Artificial Intelligence, Hefei Comprehensive National Science Center（合肥综合性国家科学中心人工智能研究院） 纠错

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

Comments 12 pages, 7 figures, accepted to ACL Findings 2025

Journal ref SQLForge: Synthesizing Reliable and Diverse Data to Enhance Text-to-SQL Reasoning in LLMs (Guo et al., Findings 2025)

Model-Grounded Symbolic Artificial Intelligence Systems Learning and Reasoning with Model-Grounded Symbolic Artificial Intelligence Systems

Aniruddha Chattopadhyay, Raj Dandekar, Kaushik Roy

机构 * Massachusetts Institute of Technology（麻省理工学院） ； University of Alabama（阿拉巴马大学） 纠错

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

Comments Accepted as paper in 19th International Conference on Neurosymbolic Learning and Reasoning,NeSy 2025

Enhancing Large Language Models with Neurosymbolic Reasoning for Multilingual Tasks

Sina Bagheri Nezhad, Ameeta Agrawal

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

Comments Accepted at 19th Conference on Neurosymbolic Learning and Reasoning (NeSy 2025)

Proceedings of the First International Workshop on Next-Generation Language Models for Knowledge Representation and Reasoning (NeLaMKRR 2024)

Ken Satoh, Ha-Thanh Nguyen, Francesca Toni, Randy Goebel, Kostas Stathis

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

Comments Associated with the 21st International Conference on Principles of Knowledge Representation and Reasoning (KR 2024) in Hanoi, Vietnam

MonitorVLM-v2: A Deployed Vision-Language Framework for Real-Time Safety Violation Detection

MonitorVLM-v2：用于实时安全违规检测的已部署视觉-语言框架

Jiang Wu, Sichao Wu, Yinsong Ma, Lifang Zheng, Jingliang Duan

机构 * School of Mechanical Engineering, University of Science and Technology Beijing（北京科技大学机械工程学院） ； The Laboratory for Computational Sensing and Robotics, Johns Hopkins University（约翰霍普金斯大学计算传感与机器人实验室） 纠错

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

AI总结 本文提出MonitorVLM-v2框架，将VLM安全评估转化为有限符号空间概率推理，结合SymPO算法与熵驱动分类机制，在地下矿部署中实现19.45倍提速，确认违规量达人工检查的2.78倍，满足实时可审计工业监控需求。

Where Do Large Language Models Fail on Competitive Programming? A Taxonomy of Failures by Algorithm Type and Difficulty Rating

大型语言模型在竞赛编程中哪里失败？基于算法类型和难度评级的失败分类

Ayush Kumar Jha, Shalini Jha

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

AI总结 通过系统实证研究，发现链式思维推理会显著降低GPT-4o的通过率并增加Claude的编译错误，且错误答案主导了两种模型的失败模式。

Comments 12 pages, 4 figures

Unleashing the Power of Tree-of-Thoughts for Edge-Enabled AIGC Service Provisioning

释放树状思维的力量以实现边缘增强的AIGC服务 provisioning

Zhang Liu, Shanhao Zhan, Shaowei Shen, Lianfen Huang, Qiao Xiang, Ying-Jun Angela Zhang, Dusit Niyato

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

AI总结 本文研究了利用树状思维提示的移动边缘计算增强的AIGC服务 provisioning，通过建立DAG模型和提出DSAC算法来优化生成延迟和质量约束。

Comments 15 pages, 11 figures, journal paper

Entropy Centroids as Intrinsic Rewards for Test-Time Scaling

熵中心作为测试时缩放的内在奖励

Wenshuo Zhao, Qi Zhu, Xingshan Zeng, Fei Mi, Lifeng Shang, Yi R., Fung

机构 * HKUST（香港科技大学） ； ZJU（浙江大学） ； Huawei（华为） 纠错

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

AI总结 本文提出熵中心作为测试时缩放的内在奖励，通过分析不确定性的时间结构提升模型响应质量，在不同规模模型上均优于现有基线。

Comments Under Review, 39 pages

LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving

LaST-VLA: 在自动驾驶的视觉-语言-动作中思考于潜在的空间-时间空间

Yuechen Luo, Fang Li, Shaoqing Xu, Yang Ji, Zehan Zhang, Bing Wang, Yuannan Shen, Jianwei Cui, Long Chen, Guang Chen, Hangjun Ye, Zhi-Xin Yang, Fuxi Wen

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

AI总结 LaST-VLA通过潜在空间-时间推理框架，结合双特征对齐机制和渐进式SFT训练策略，提升自动驾驶中的视觉-语言-动作处理能力，实现更安全的物理基础推理。

MovSAM: A Single-image Moving Object Segmentation Framework Based on Deep Thinking

Chang Nie, Yiqing Xu, Guangming Wang, Zhe Liu, Yanzi Miao, Hesheng Wang

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

ReAcTable: Enhancing ReAct for Table Question Answering

Yunjia Zhang, Jordan Henkel, Avrilia Floratou, Joyce Cahoon, Shaleen Deep, Jignesh M. Patel

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

COLLIE: Systematic Construction of Constrained Text Generation Tasks

Shunyu Yao, Howard Chen, Austin W. Hanjie, Runzhe Yang, Karthik Narasimhan

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

Comments 18 pages, 12 figures

CircuitReason-1k: Benchmarking Long-Horizon Visual-to-Symbolic Reasoning inElectrical Circuits

CircuitReason-1k：电路中的长程视觉到符号推理基准测试

Xinqi Yang, Kang An, Tengyue Wang, Zhongyu Yang, Chenxu Du, Yuanchi Zhu, Hebao Zhu, Ziliang Wang, Faqiang Qian, Yunli Yang, Qibing Ren

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

AI总结 CircuitReason-1k是含1000个真实教材电路问题的基准，用于评估多模态模型的长程视觉到符号推理，现有模型在长程问题上表现差，该基准为相关测试提供了聚焦平台。

JigShape: Evaluating Visual-Geometric Reasoning in VLMs through Jigsaw Puzzles

JigShape：通过拼图任务评估视觉语言模型的视觉几何推理能力

Shawn Li, Wei Yang, Jike Zhong, Jiate Li, Jiawei Yang, You Qin, Ryan Rossi, Franck Dernoncourt, Roger Zimmermann, Yue Wang, Zhengzhong Tu, Vicente Ordonez, Mohit Bansal, Yue Zhao

机构 * University of Southern California（南加州大学） ； National University of Singapore（新加坡国立大学） ； Adobe Research（奥多比研究院） ； Texas A&M University（德克萨斯农工大学） ； Rice University（莱斯大学） ； The University of North Carolina at Chapel Hill（北卡罗来纳大学教堂山分校） 纠错

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

AI总结 本研究提出JigShape拼图基准，发现零样本VLM大多缺乏几何推理能力，所有模型在大尺寸拼图上均出现性能崩塌，将可扩展几何推理确立为VLM的开放性挑战。