大模型推理能力

Deductive Algorithmic Knowledge

Riccardo Pucella

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

Comments 28 pages. A preliminary version of this paper appeared in the Proceedings of the 8th International Symposium on Artificial Intelligence and Mathematics, AI&M 22-2004, 2004

Journal ref Journal of Logic and Computation 16 (2), pp. 287-309, 2006

Knowledge And The Action Description Language A

Jorge Lobo, Gisela Mendez, Stuart R. Taylor

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

Comments Appeared in Theory and Practice of Logic Programming, vol. 1, no. 2, 2001

Journal ref Theory and Practice of Logic Programming, vol. 1, no. 2, 2001

Set-Theoretic Completeness for Epistemic and Conditional Logic

Joseph Y. Halpern

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

Comments This is an expanded version of a paper that appeared in AI and Mathematics, 1998

Journal ref Annals of Mathematics and Artificial Intelligence, vol. 26, 1999, pp. 1-27

Qualitative and Quantitative Models of Speech Translation

Hiyan Alshawi

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

Comments Appeared in proceedings of the ACL workshop "The Balancing Act, Combining Symbolic and Statistical Approaches to Language", Las Cruces NM, July 1994. LaTeX, 24 pages

Representation Theory for Default Logic

Victor Marek, Jan Treur, Miroslaw Truszczynski

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

Comments Annals of Mathematics and Artificial Intelligence, 21 (1997), pp. 343-358

Constraint-based analysis of composite solvers

Evgueni Petrov, Eric Monfroy

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

Comments submitted to AI SAC 2004

Defeasible Logic Programming: An Argumentative Approach

Alejandro Javier Garcia, Guillermo Ricardo Simari

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

Comments 43 pages, to appear in the journal "Theory and Practice of Logic Programming"

A Modal Logic Framework for Multi-agent Belief Fusion

Churn-Jung Liau

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

Comments LaTex, 29 pages

Multi-Agent Only Knowing

Joseph Y. Halpern, Gerhard Lakemeyer

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

Comments To appear, Journal of Logic and Computation

GPT in Game Theory Experiments

Fulin Guo

专题命中 逻辑推理 ：reasoning(abstract,comments)

Comments updated to use GPT-4 instead of GPT-3.5 and added reasoning analysis

Finding Small Proofs for Description Logic Entailments: Theory and Practice (Extended Technical Report)

Christian Alrabbaa, Franz Baader, Stefan Borgwardt, Patrick Koopmann, Alisa Kovtunova

专题命中 逻辑推理 ：reasoning(abstract,journal_ref)

Comments Extended version of a paper accepted at LPAR23

Journal ref LPAR-23: 23rd International Conference on Logic for Programming, Artificial Intelligence and Reasoning, vol 73, 2020, pages 32--67

RT-NeuS: Towards Real-Time Neuro-Symbolic Video Understanding via Adaptive Temporal Verification

LE-NeuS: 通过自适应时间验证实现低延迟的神经符号视频理解

Shawn Liang, Sahil Shah, Chengwei Zhou, S P Sharan, Harsh Goel, Arnab Sanyal, Sandeep Chinchali, Gourav Datta

机构 * Case Western Reserve University（凯斯西储大学） ； The University of Texas at Austin（德克萨斯大学奥斯汀分校） 纠错

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 LE-NeuS通过自适应时间验证优化，实现低延迟的神经符号视频理解，在保持高准确率的同时大幅减少推理延迟。

Comments Camera-ready version, accepted at NeuS 2026

PowerSlider: Exploiting Phase Asymmetry for LLM Serving under Demand Response

PowerSlider：利用相位不对称性实现需求响应下的大语言模型服务

Yueying Li, Jiayang Chen, Yuanfan Chen, Leo Han, Haoran Qiu, Esha Choukse, Rodrigo Fonseca, Udit Gupta

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 PowerSlider通过分解大语言模型服务阶段并结合KKT在线求解器，在电网需求响应的时变功率上限下，显著提升了服务吞吐量与延迟性能。

Multi-Tier Mentorship with AI-Assisted Development: Authentic Engineering for K-12 and Undergraduates

结合AI辅助开发的多层级指导模式：面向K-12学生与本科生的真实工程实践

Kelly Yuan, Ronald Liu, Daniel Crawford, Weihao Qu

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 本文提出结合AI辅助开发的多层级指导框架，让K-12学生与本科生协作开发LuckyTag系统，试点显示该模式可降低技术障碍、提升架构理解，为跨群体计算协作提供了新方案。

Comments 8 pages, 8 figures, 2 tables. Accepted to IEEE ISEC 2026

TrustMee: Self-Verifying Remote Attestation Evidence

TrustMee: 自验证远程证明证据

Parsa Sadri Sinaki, Zainab Ahmad, Wentao Xie, Merlijn Sebrechts, Jimmy Kjällman, Lachlan J. Gunn

专题命中 逻辑推理 ：verifier(abstract)

AI总结 TrustMee通过自验证远程证明证据机制，实现跨平台的证明验证，无需特定平台代码，利用WebAssembly组件实现验证逻辑的通用性。

Comments 16 pages, 7 figures

QRS: A Rule-Synthesizing Neuro-Symbolic Triad for Autonomous Vulnerability Discovery

QRS：一种用于自主漏洞发现的神经符号三元组规则生成器

George Tsigkourakos, Constantinos Patsakis

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 QRS通过神经符号方法生成代码查询，发现超出预定义模式的漏洞，显著降低假阳性，并在低开销下发现高严重性漏洞。

Autonomous Cyber Defense in Connected Vehicles: A Multi-Agent Approach to V2X Security

网联车辆中的自主网络防御：一种面向V2X安全的多智能体方法

Krishna Teja Medam

专题命中 逻辑推理 ：planning(abstract)

AI总结 针对网联车辆V2X安全，提出三层多智能体架构，以SAE标准时序为硬性约束，分层处理消息分类、冲突解决与模型优化，解决现有入侵检测系统的安全-安保冲突问题。

Automating Parent Selection Configuration in Genetic Programming with Agentic AI

用智能体人工智能实现遗传编程中父代选择配置的自动化

Jose Guadalupe Hernandez, Jui-Hsuan Chang, Anil Kumar Saini, Xi Li, Jason H. Moore

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 该研究提出智能体AI框架，结合LLM推理与RAG实现遗传编程父代选择配置自动化，经符号回归测试，5 mini--AR配置表现优于锦标赛选择，为进化系统自动化设计提供了新路径。

Astrobiology and the Transformation of Scientific Epistemology

天体生物学与科学认识论的变革

Kristina Šekrst

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 该研究探讨天体生物学因“n=1问题”成为缺乏基础确定性的“暂态科学”，其与AI等新兴领域共同推动科学知识构建方式的变革。

Comments In: Steven J. Dick, Astronomy and Philosophy: Conceptual and Methodological Foundations and Challenges, 2026. Cambridge: Cambridge University Press

Journal ref In: Steven J. Dick, Astronomy and Philosophy: Conceptual and Methodological Foundations and Challenges, 2026. Cambridge: Cambridge University Press

MPC for underactuated spacecraft control with a Lyapunov supervised physics-informed neural network correction layer

基于李雅普诺夫监督的物理信息神经网络校正层的欠驱动航天器MPC控制

Amirhossein Ayanmanesh Motlaghmofrad, Carlo Cena, Mauro Martini, Marcello Chiaberge

机构 * Politecnico di Torino（托斯纳理工学院） ； Argotec S.R.L.（Argotec公司） 纠错

专题命中 逻辑推理 ：planning(abstract)

AI总结 针对欠驱动航天器姿态控制，提出一种分层架构，结合非线性模型预测控制、物理信息神经网络和李雅普诺夫监督机制，在不确定性下降低稳态误差并保持鲁棒性。

Comments Accepted at SPAICE (AI in and for Space) 2026

KC-BFPRL: Knowledge-Guided Multi-UAV Collaboration for Grassland Restoration via Bilevel Formerpointer-Based Reinforcement Learning

KC-BFPRL：基于知识引导的多无人机草原修复协作方法——通过 bilevel Formerpointer 强化学习实现

Dongbin Jiao, Xianyi Wang, Yuchen Yuan, Weibo Yang, Peng Yang, Peng Zhao, Zhanhuan Shang, Shi Yan

专题命中 逻辑推理 ：planning(abstract)

AI总结 KC-BFPRL 是一种知识引导的 bilevel Formerpointer 强化学习框架，可解决多无人机草原修复的 RAMP 问题，性能优于现有方法，适用于大规模自动化生态修复。

When Truth Is Distributed: Misinformation Derails Collective Fact Recovery in LLM-Based Multi-Agent Systems

当真相被分散时：错误信息会破坏基于大语言模型（LLM）的多智能体系统中的集体事实恢复

Chenfei Yan, Zeyang Yue, Feifei Zhao, Erliang Lin, Lu Jia, Haibo Tong, Mingyang Lyu, Chengyi Sun, Yi Zeng

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 本研究提出Hi-Agreement评估框架，发现基于LLM的多智能体系统中，关键证据持有者的虚假证词会破坏集体事实恢复，使恢复率从72.50%降至14.17%，且虚假证据会持续传播。

From Guessing to Seeing: Enhancing LLM-Based Program Repair via Trace-Guided Multi-strategy Debate

通过多智能体辩论引导的运行时执行轨迹指导的自动程序修复

Jiaqing Wu, Tong Wu, Manqing Zhang, Yunwei Dong, Bo Shen

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 本文提出TraceRepair框架，利用运行时事实作为共享约束进行补丁验证，通过多智能体协作提高修复效率和泛化能力，实验显示其在修复缺陷和处理新 bug 数据集上表现优异。

Comments 13 pages, 4 figures, 10 tables. Accepted at the 41st IEEE/ACM International Conference on Automated Software Engineering (ASE 2026)

Mimir: A Neuro-Symbolic Memory System with Dynamic Grounding for Embodied Agents in Interactive Environments

Mimir：面向交互环境中具身智能体的、具备动态 grounding 的神经符号记忆系统

Haoming Xu, Zhenlin He, Hengyi Wang, Jiafeng Xu, Hao Dong

专题命中 逻辑推理 ：planning(abstract)

AI总结 本文提出Mimir，一种分离世界与任务记忆并具备动态grounding的神经符号记忆系统，在EB-ALFRED、EB-Habitat等具身任务中显著提升了智能体的长程执行成功率。

Comments 9 pages, 4 figures

DAC-Pose: Dual-Agent Collaborative Framework for Pose-Guided Human Generation

DAC-Pose：用于姿态引导人体生成的双智能体协作框架

Haotian Yang, Zhile Yang, Huiyu Zhou, Xin Sun

机构 * Faculty of Data Science, City University of Macau（澳门城市大学数据科学学院） ； Shenzhen University of Advanced Technology（深圳理工大学） ； Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences（中国科学院深圳先进技术研究院） ； School of Computing and Mathematical Sciences, University of Leicester（莱斯特大学计算与数学科学学院） 纠错

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 针对姿态引导人体生成在剧烈视角变化下的视觉伪影问题，提出双智能体协作框架DAC-Pose，通过PSR与DAVE智能体的反馈循环实现高保真合成，在DeepFashion和Market-1501基准上验证了其优越性。

Comments Code is available at DAC-Pose" target="_blank" rel="noopener">https://github.com/AIVRC/DAC-Pose

Neuro-Symbolic Proof-of-Vulnerability Generation with Open-Weight Models

基于开放权重模型的神经符号漏洞证明生成

Yu Nong, Haipeng Cai

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 研究人员提出POVGEN神经符号框架，利用开放权重模型实现低成本漏洞证明生成，在基准测试和真实CVE中表现优于模糊测试与符号执行，还发现了有缺陷补丁及未报告漏洞。

CROSS: Cascaded Distillation and Dual-Constraint Grounding for Remote Sensing Referring Segmentation

CROSS：用于遥感指称分割的级联蒸馏与双约束 grounding

Tingzhang Luo, Ruizhong Liu, Yichao Liu, Cheng Fan, Yu Liu, Jianyuan Guo

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 针对遥感指称分割中架构弱耦合与语义偏差问题，本文提出CROSS范式，通过LGCD与PSCL实现性能突破，成为RRSIS的稳健新方案。

Comments Accepted at the European Conference on Computer Vision (ECCV) 2026. 20 pages, 6 figures, and 5 tables. Tingzhang Luo and Ruizhong Liu contributed equally. Jianyuan Guo is the corresponding author. Project page: https://clarence-cv.github.io/CROSS/

EffiHolmes: Differential Profiling-Guided Repository Level Time Inefficiency Fix Localization

EffiHolmes：基于差分分析的仓库级时间低效修复定位

Haowen Yang, Yun Peng, Zishuo Ding

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 EffiHolmes是基于LLM的仓库级时间低效修复定位框架，通过差分分析等技术提升定位精度，在RepoEffi-Bench上优于多种基线方法，且跨模型规模稳健。

Comments Accepted at the 41st IEEE/ACM International Conference on Automated Software Engineering (ASE 2026). 13 pages, 3 figures, and 4 tables

Rethinking Agentic Kernel Generation for Emerging Accelerators

面向新兴加速器的智能体内核生成方法反思

Ruijie Gao, Jirong Yang, Barry Lyu, Haoran Jin, Nathan Bleier

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 针对新兴加速器内核生成的现有方法反复重建无关语义的问题，提出编译器介导的Zomboss框架，将语义编译为可复用接口，在20个Gemmini和36个PLENA工作负载上实现全实例正确，且性能优于基线方法、推理成本更低。

Comments 12 pages, 7 figures

Breaking the Horizontal Prior: From Long-Tailed Orientation Bias to Roll-Robust Monocular Depth Estimation

打破水平先验：从长尾方向偏差到抗翻滚单目深度估计

Kaihua Tang, Ziqing Xia, Xiaoxu Zheng, Xiaoxue Zhang, Michael Bi Mi, Zhan Xu, Dave Zhenyu Chen

专题命中 逻辑推理 ：reasoning(abstract)

AI总结 该研究针对单目深度估计中受水平先验（长尾方向偏差）导致的相机翻滚鲁棒性差问题，提出训练时监督策略ID-Constraint，经多基准数据集实验验证了方法的有效性。

Comments The code is publicly available on GitHub: https://github.com/KaihuaTang/Horizontal-Prior