cs.AI（2026-06-01）

📊 共 43 篇论文 | 🔗 9 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (21 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (20 🔗5) 支柱一：机器人控制 (Robot Control) (1) 支柱六：视频提取与匹配 (Video Extraction) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (21 篇)

#题目一句话要点标签🔗⭐
1 Compliance-Scored Best-of-N Guardrail Orchestration for Multimodal Document Generation in Payments Dispute Defense 提出合规性评分的Best-of-N机制，用于支付争议防御中的多模态文档生成 multimodal
2 Not All Errors Are Equal: A Systematic Study of Error Propagation in Large Language Model Inference 提出LLMFI框架，系统研究大语言模型推理中的错误传播问题 large language model
3 RA-LWLM: Retrieval-Augmented In-Context Localization with Wireless Foundation Models 提出RA-LWLM，利用无线基础模型实现免训练的跨场景定位。 foundation model
4 Boosting Multimodal Federated Learning via Chained Modality Optimization 提出FedMChain，通过链式模态优化提升多模态联邦学习性能。 multimodal
5 Demystifying Multimodal Biomolecular Co-design With Intrinsic Geodesic Coupling 提出GeoCoupling框架以优化生物分子多模态协同设计 multimodal
6 eMoT: evolving Memory-of-Thought via Symbolic Anchoring and Memory Corrosion 提出eMoT框架，通过演进式记忆和符号锚定提升LLM多步推理的可靠性。 large language model chain-of-thought
7 MOSS-Audio Technical Report MOSS-Audio：面向语音、环境声和音乐理解的统一音视频语言模型 large language model instruction following
8 HLL: Can Agents Cross Humanity's Last Line of Verification? 提出HLL基准测试，评估多模态Agent在交互式验证码破解中的类人能力。 multimodal ✅
9 Iteris: Agentic Research Loops for Computational Mathematics Iteris：面向计算数学的Agentic研究环路系统 large language model
10 Evidence-Gated LLM Priors for Multi-Objective Bayesian Optimization 提出证据门控LLM先验的多目标贝叶斯优化方法，提升黑盒优化中LLM建议的可靠性。 large language model
11 Bridging the Last Mile of Time Series Forecasting with LLM Agents 提出基于LLM Agent的时间序列预测框架，弥合统计预测与业务应用间的差距 foundation model
12 MCP-Persona: Benchmarking LLM Agents on Real-World Personal Applications via Environment Simulation MCP-Persona：通过环境模拟评估LLM Agent在真实个人应用中的性能 large language model ✅
13 MOC: Multi-Order Communication in LLM-based Multi-Agent Systems 提出MOC多阶通信方案，提升LLM多智能体系统中的信息传递效率与任务性能。 large language model ✅
14 POIROT: Interrogating Agents for Failure Detection in Multi-Agent Systems 提出POIROT以解决多智能体系统中的故障检测问题 large language model
15 SMH-Bench: Benchmarking LLM Agents for Environment-Grounded Reasoning and Action in Smart Homes SMH-Bench：用于评估LLM智能家居环境推理与行动能力的综合基准 large language model
16 WorldCoder-Bench: Benchmarking Physically Grounded 3D World Synthesis 提出WorldCoder-Bench，用于评估LLM在物理规则3D世界合成中的能力。 large language model
17 RadioMaster: Multi-Agent System for Autonomous Radio Signal Generation RadioMaster：用于自主无线信号生成的多智能体系统 large language model
18 Does Compression Preserve Uncertainty? A Unified Benchmark for Quantized and Sparse LLMs via Conformal Prediction 提出基于Conformal Prediction的统一基准，评估压缩LLM的不确定性保持能力。 large language model
19 Consistency evaluation of benchmarks used for causal discovery 提出一种基于LLM的自动pipeline，用于评估因果发现benchmark的知识一致性。 large language model
20 Revisiting Ripple Effects in Knowledge Editing through Pressure-Aware Joint Neighborhood Optimization 提出JNO框架，通过压力感知联合邻域优化解决知识编辑中的涟漪效应。 large language model
21 RoleCDE:Benchmarking and Mitigating Role-Alignment Trade-offs in Role-Playing Agents RoleCDE：用于评估和缓解角色扮演智能体中角色对齐权衡的基准 large language model ✅

🔬 支柱二：RL算法与架构 (RL & Architecture) (20 篇)

#题目一句话要点标签🔗⭐
22 Spatial Representation Learning Beyond Pixels: Unifying Raster Data and Vector Semantics for Human-Centric Geospatial Foundation Models 提出统一空间表征学习框架，融合栅格数据与矢量语义，构建以人为中心的地理空间基础模型。 representation learning foundation model multimodal
23 TrafficRAG: A Multimodal RAG Framework for Traffic Accident Liability Determination TrafficRAG：多模态检索增强框架，用于交通责任事故判定 MAE large language model multimodal
24 Explainable Data-driven Deep Reinforcement Learning Methods for Optimal Energy Management in Buildings 提出可解释深度强化学习框架，优化建筑能源管理并提升用户信任 reinforcement learning deep reinforcement learning DRL
25 COMAP: Co-Evolving World Models and Agent Policies for LLM Agents COMAP：面向LLM Agent的协同进化世界模型与策略，提升交互环境决策能力 world model world models distillation ✅
26 Echo: A Joint-Embedding Predictive Architecture for Speaker Diarization and Speech Recognition in a Shared Latent Space Echo：基于共享隐空间的联合嵌入预测架构，用于说话人分离和语音识别 JEPA Joint-Embedding Predictive Architecture joint-embedding predictive architecture
27 EvoBrain: Continual Learning of EEG Foundation Models Across Heterogeneous BCI Tasks EvoBrain：面向异构BCI任务的脑电基础模型持续学习框架 distillation foundation model
28 SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment SafeSteer：面向安全对齐的局部化On-Policy蒸馏方法 distillation large language model ✅
29 Learning When Not to Act: Mitigating Tool Abuse in Agentic Reinforcement Learning EAPO：通过学习何时不行动来缓解Agentic强化学习中的工具滥用问题 reinforcement learning policy learning reward shaping
30 SafeMCP: Proactive Power Regulation for LLM Agent Defense via Environment-Grounded Look-Ahead Reasoning SafeMCP：通过环境感知的前瞻推理实现LLM Agent的主动式能力管控 reinforcement learning world model world models
31 Community-Aware Assessment of Social Textual Engagement and Resonance: A Human-Centric Perspective on User-Generated Content Evaluation 提出MEDEA模型，通过模拟社群共鸣评估用户生成内容质量，超越传统视觉保真度指标。 reinforcement learning multimodal chain-of-thought
32 SIRI: Self-Internalizing Reinforcement Learning with Intrinsic Skills for LLM Agent Training SIRI：通过自内部化强化学习与内在技能训练LLM Agent reinforcement learning distillation ✅
33 S-SPPO: Semantic-Calibrated Self-Play Preference Optimization S-SPPO：通过语义校准的自博弈偏好优化，解决LLM对齐中的策略退化问题 DPO direct preference optimization large language model ✅
34 Coordination Graphs for Constrained Multi-Agent Reinforcement Learning 提出CG-CMARL框架，通过协调图和拉格朗日对偶解决约束多智能体强化学习问题 reinforcement learning reward shaping
35 Physically-Constrained Mamba-SDE for Remaining Useful Life Prediction under Irregular Observations 提出PC-MambaSDE，解决不规则观测下剩余寿命预测的物理约束问题 latent dynamics Mamba
36 JenBridge: Adaptive Long-Form Video Soundtracking across Scene Transitions 提出JenBridge，解决长视频场景过渡中配乐连贯性问题 flow matching large language model
37 Harness-1: Reinforcement Learning for Search Agents with State-Externalizing Harnesses Harness-1：利用强化学习和外部状态管理提升搜索Agent性能 reinforcement learning ✅
38 EVA-Net: Subject-Independent EEG Motor Decoding with Video-Derived Motor Priors EVA-Net：利用视频运动先验实现与受试者无关的脑电运动解码 distillation multimodal
39 TriAlign: Towards Universal Truth Consistency in Personalized LLM Alignment TriAlign：面向个性化LLM对齐的通用真值一致性方法 reinforcement learning large language model
40 TRON: Targeted Rule-Verifiable Online Environments for Visual Reasoning RL TRON：面向视觉推理强化学习的可控规则验证在线环境 reinforcement learning multimodal
41 ReSkill: Reconciling Skill Creation with Policy Optimization in Agentic RL ReSkill：在Agentic RL中协调技能创建与策略优化 reinforcement learning policy learning

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
42 Bridging the Sim-to-Real Gap in Semiconductor Visual Program Synthesis via Input Binarization 提出基于输入二值化的视觉程序合成方法，弥合半导体图像Sim-to-Real差距 manipulation sim-to-real

🔬 支柱六：视频提取与匹配 (Video Extraction) (1 篇)

#题目一句话要点标签🔗⭐
43 Token Predictors Are Not Planners: Building Physically Grounded Causal Reasoners 提出Causal-Plan-Bench和Causal-Plan-1M，提升具身智能体物理因果推理能力。 egocentric

⬅️ 返回 cs.AI 首页 · 🏠 返回主页