cs.AI（2026-05-11）

📊 共 39 篇论文 | 🔗 5 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (24 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (9 🔗1) 支柱一：机器人控制 (Robot Control) (3) 支柱八：物理动画 (Physics-based Animation) (2) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (24 篇)

#题目一句话要点标签🔗⭐
1 RW-Post: Auditable Evidence-Grounded Multimodal Fact-Checking in the Wild 提出RW-Post基准与AgentFact框架，实现真实场景下可审计的多模态事实核查 multimodal visual grounding ✅
2 AgentRx: A Benchmark Study of LLM Agents for Multimodal Clinical Prediction Tasks AgentRx：评估LLM Agent在多模态临床预测任务中的性能表现 large language model multimodal
3 Active Testing of Large Language Models via Approximate Neyman Allocation 提出基于近似Neyman分配的主动测试算法，显著降低大语言模型生成任务的评估成本 large language model multimodal
4 LLM4Branch: Large Language Model for Discovering Efficient Branching Policies of Integer Programs 提出LLM4Branch框架，利用大语言模型自动化发现高效的混合整数线性规划分支策略。 large language model ✅
5 CORTEG: Foundation Models Enable Cross-Modality Representation Transfer from Scalp to Intracranial Brain Recordings 提出CORTEG框架，利用预训练EEG基础模型实现跨模态脑电信号迁移与高效解码 foundation model
6 Hypothesis-Driven Deep Research with Large Language Models: A Structured Methodology for Automated Knowledge Discovery 提出假设驱动的深度研究方法（HDRI）与INFOMINER系统，实现自动化知识发现 large language model
7 GuardAD: Safeguarding Autonomous Driving MLLMs via Markovian Safety Logic 提出GuardAD以解决自动驾驶系统安全性问题 large language model multimodal
8 MaD Physics: Evaluating information seeking under constraints in physical environments 提出MaD Physics基准，旨在评估智能体在物理环境约束下的信息获取与科学发现能力 multimodal
9 Rethinking Agentic Search with Pi-Serini: Is Lexical Retrieval Sufficient? 提出Pi-Serini搜索代理框架，论证了在强推理LLM加持下词法检索（BM25）在深度研究任务中的有效性。 large language model ✅
10 The First Drop of Ink: Nonlinear Impact of Misleading Information in Long-Context Reasoning 揭示长文本推理中的“墨水滴入效应”：量化误导性信息对大模型性能的非线性影响 large language model
11 Probing Cross-modal Information Hubs in Audio-Visual LLMs 揭示音视频大模型中的跨模态信息枢纽，并提出无需训练的幻觉缓解策略 large language model ✅
12 Threat Modelling using Domain-Adapted Language Models: Empirical Evaluation and Insights 系统性评估领域适配语言模型在5G结构化威胁建模中的效能与局限 large language model
13 Reasoning Is Not Free: Robust Adaptive Cost-Efficient Routing for LLM-as-a-Judge 提出RACER路由框架，通过分布鲁棒优化实现LLM-as-a-Judge的成本效益平衡 large language model
14 Can You Keep a Secret? Involuntary Information Leakage in Language Model Writing 揭示语言模型写作中的无意信息泄露现象 chain-of-thought
15 The Bystander Effect in Multi-Agent Reasoning: Quantifying Cognitive Loafing in Collaborative Interactions 揭示多智能体推理中的“旁观者效应”：量化协作交互中的认知惰性与主权缺失 large language model
16 Re-Triggering Safeguards within LLMs for Jailbreak Detection 提出一种基于嵌入扰动的重触发机制，通过激活大模型内置安全防御来检测越狱提示词。 large language model
17 Budget-Efficient Automatic Algorithm Design via Code Graph 提出基于代码图（Code Graph）的自动算法设计框架，通过增量式修正实现预算高效的算法搜索。 large language model
18 An agentic framework for gravitational-wave counterpart association in the multi-messenger era 提出GW-Eyes智能体框架，利用大语言模型实现引力波与电磁对应体的自动化关联 large language model
19 Guaranteed Jailbreaking Defense via Disrupt-and-Rectify Smoothing 提出Disrupt-and-Rectify Smoothing防御框架，实现大语言模型越狱攻击的理论保证防御 large language model
20 LLM Jaggedness Unlocks Scientific Creativity 提出SciAidanBench基准并揭示LLM的“锯齿状”能力分布，通过模型集成提升科学创造力 large language model
21 A Reflective Storytelling Agent for Older Adults: Integrating Argumentation Schemes and Argument Mining in LLM-Based Personalised Narratives 提出一种基于论证挖掘与知识图谱的反射式叙事代理，以提升老年人数字陪伴的叙事质量与可信度。 large language model
22 ASIA: an Autonomous System Identification Agent 提出ASIA自主系统辨识智能体，通过大语言模型实现动力学模型构建的自动化闭环。 large language model
23 Beyond Autonomy: A Dynamic Tiered AgentRunner Framework for Governable and Resilient Enterprise AI Execution 提出动态分层AgentRunner框架，以实现企业级AI执行的可治理性与韧性 large language model
24 Benchmarking Safety Risks of Knowledge-Intensive Reasoning under Malicious Knowledge Editing 提出EditRisk-Bench基准，系统评估大模型在恶意知识编辑下的推理安全风险 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (9 篇)

#题目一句话要点标签🔗⭐
25 How Mobile World Model Guides GUI Agents? 提出多模态移动世界模型，通过对比四种表征方式揭示其在GUI智能体中的指导作用 world model world models multimodal
26 BEACON: A Multimodal Dataset for Learning Behavioral Fingerprints from Gameplay Data 提出BEACON多模态数据集，通过高保真电竞游戏数据实现行为指纹识别与持续身份认证 representation learning multimodal
27 CLEF: EEG Foundation Model for Learning Clinical Semantics 提出CLEF：一种面向临床语义理解的长时程脑电图（EEG）基础模型 representation learning foundation model
28 Towards Autonomous Railway Operations: A Semi-Hierarchical Deep Reinforcement Learning Approach to the Vehicle Rescheduling Problem 提出半层级深度强化学习方法，以解决复杂铁路网中的车辆重调度问题 reinforcement learning deep reinforcement learning
29 BenchCAD: A Comprehensive, Industry-Standard Benchmark for Programmatic CAD 提出BenchCAD基准测试，旨在评估并提升多模态大模型在工业级参数化CAD代码生成中的能力。 reinforcement learning large language model multimodal
30 Evolving-RL: End-to-End Optimization of Experience-Driven Self-Evolving Capability within Agents 提出Evolving-RL框架，通过经验提取与利用的协同进化提升大模型的自我演化能力 reinforcement learning large language model foundation model
31 TRACE: Distilling Where It Matters via Token-Routed Self On-Policy Alignment 提出TRACE框架：通过基于Token路由的自策略对齐，实现高效的强化学习蒸馏 reinforcement learning distillation privileged information
32 TMAS: Scaling Test-Time Compute via Multi-Agent Synergy 提出TMAS框架，通过多智能体协同与分层记忆机制实现推理阶段计算规模化 reinforcement learning large language model ✅
33 Verifiable Process Rewards for Agentic Reasoning 提出可验证过程奖励（VPR）框架，通过密集的中间监督解决长程智能体推理中的信用分配难题。 reinforcement learning large language model

🔬 支柱一：机器人控制 (Robot Control) (3 篇)

#题目一句话要点标签🔗⭐
34 LoopVLA: Learning Sufficiency in Recurrent Refinement for Vision-Language-Action Models 提出LoopVLA架构，通过循环细化与充分性估计提升视觉-语言-动作模型的推理效率与控制精度。 manipulation vision-language-action VLA
35 When Prompts Become Payloads: A Framework for Mitigating SQL Injection Attacks in Large Language Model-Driven Applications 提出一种多层安全框架，通过提示词清洗与语义异常检测防御LLM驱动的SQL注入攻击 manipulation large language model
36 Hierarchical Causal Abduction: A Foundation Framework for Explainable Model Predictive Control 提出分层因果溯因（HCA）框架，实现非线性模型预测控制（MPC）的可解释性决策 MPC model predictive control

🔬 支柱八：物理动画 (Physics-based Animation) (2 篇)

#题目一句话要点标签🔗⭐
37 Geospatial-Temporal Sensemaking of Remote Sensing Activity Detections with Multimodal Large Language Model 提出SMART-HC-VQA数据集与多模态大模型框架，实现遥感影像的时空活动推理 spatiotemporal large language model multimodal
38 A Cold Diffusion Approach for Percussive Dereverberation 提出基于冷扩散（Cold Diffusion）的打击乐去混响框架，有效处理瞬态信号的复杂混响问题。 PULSE

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
39 Cavity-Enhanced Collective Quantum Processing with Polarization-Encoded Qubits 提出一种基于腔增强光学架构的集体量子处理方案，利用偏振编码量子比特实现通用量子门。 physically plausible

⬅️ 返回 cs.AI 首页 · 🏠 返回主页