cs.AI（2026-10-02）

📊 共 43 篇论文 | 🔗 8 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (25 🔗6) 支柱二：RL算法与架构 (RL & Architecture) (15 🔗2) 支柱七：动作重定向 (Motion Retargeting) (1) 支柱四：生成式动作 (Generative Motion) (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) (25 篇)

#题目一句话要点标签🔗⭐
1 HyperBrowseComp: A Multilingual and Multimodal Stress Test for Web-Browsing Agents 提出HyperBrowseComp以解决多语言网页浏览代理的挑战 multimodal
2 Do Large Language Models Know Colombian Law? A Reliability Benchmark for the Colombian Legal System 提出针对哥伦比亚法律系统的LLM可靠性基准 large language model
3 CVE2AP: Automated Generation of PDDL-Encoded Attack Paths via Large Language Models 提出CVE2AP以自动生成PDDL编码攻击路径 large language model
4 Multimodal reasoning for broadly neutralizing antibody discovery from label-free human B cell repertoires across virus families 提出ImmuneAgent以解决广谱中和抗体发现难题 multimodal
5 On the Chain-of-Thought Monitorability of Looped Language Models 系统评估循环语言模型的链式思维可监控性 chain-of-thought
6 Large language models exhibit unreliable updating of clinical judgment as patient evidence evolves 提出EVLU方法以解决LLM在临床判断更新中的不可靠性问题 large language model
7 Efficient Reasoning Training Does Not Always Harm CoT Faithfulness and Monitorability 提出高效推理训练方法以提升CoT的可监控性与可信度 large language model chain-of-thought
8 HazardWeaver: Scientific Route Selection for Hazard Analysis Agents 提出HazardWeaver以解决自然灾害分析中的科学路线选择问题 large language model ✅
9 Jumping the Line: Exploiting Length Predictions in LLM Scheduling 提出JIL以优化LLM调度中的长度预测问题 large language model
10 Multi-Task Evolution for Zero-Shot Cross-Problem Generalization using LLMs 提出MECo框架以解决零-shot跨问题泛化挑战 large language model
11 MOF-VERIFY: A Failure-Aware Agentic Harness for MOF Hypothesis Verification 提出MOF-VERIFY以解决MOF假设验证中的失败问题 large language model ✅
12 When Numbers Start Talking: Numerical Signalling and Strategic Behaviour Among LLMs 研究LLM在多智能体系统中数值信号与战略行为的影响 large language model
13 Beyond Predefined Sinks: Security-Aware Dependency Analysis for LLM Agents 提出AgentSecGraph框架以解决LLM代理的安全依赖分析问题 large language model
14 PLCWorld: Benchmarking LLM-Generated PLC Programs in Closed-Loop Plant Simulation 提出PLCWorld以评估LLM生成PLC程序的有效性与安全性 large language model ✅
15 Relevant Evidence Decoding for Audio-Visual Hallucination Mitigation 提出相关证据解码以缓解音视频幻觉问题 large language model
16 Reliable Self-Evolution with Imperfect Proxy Rewards 提出CISE以解决不完美代理奖励下的自我进化问题 large language model ✅
17 CreateScore: Domain-Theory-Informed Bayesian Routing for LLM-Based CV Screening 提出CreateScore以优化LLM在简历筛选中的应用 large language model
18 Reasoning with Evidence, Not Merely Rationales: Verifiable Preference Proofs for LLM-Based Recommendation 提出PROVE-REC以解决LLM推荐中的证据与偏好不一致问题 large language model
19 LUMOS: Tracing Parametric Knowledge from Training Data to Behavioral Outputs in LLMs 提出LUMOS框架以解决LLMs知识评估中的不确定性问题 chain-of-thought
20 PsyEvo: A Personalized Counseling Agent That Self-Evolves at Test Time 提出PsyEvo以解决心理健康咨询个性化不足问题 large language model ✅
21 DNAlign: Dynamic Null-Space Safe Alignment for LLMs 提出DNAlign以解决大型语言模型安全对齐问题 large language model
22 AMBER: Multi-View Adaptive Budget Allocation for Listwise Vision-Language Reranking 提出AMBER以解决多模态检索中的高推理成本问题 multimodal ✅
23 MLCommons Jailbreak Benchmark v1.0 提出MLCommons Jailbreak Benchmark v1.0以评估语言模型的安全性 large language model
24 iS-KV: Online Low-Rank KV Cache Compression via Block-Incremental SVD 提出iS-KV以解决长链推理中的KV缓存压缩问题 chain-of-thought
25 Time Series Forecasting Benchmarks Need Scenario-Grounded Stress Testing 提出场景基础的压力测试以解决时间序列预测评估不足问题 foundation model

🔬 支柱二：RL算法与架构 (RL & Architecture) (15 篇)

#题目一句话要点标签🔗⭐
26 TACD: Distilling Efficient Text-to-Motion Models via Terminal Amplification Control 提出TACD以解决文本到运动模型的效率问题 distillation text-to-motion motion retrieval ✅
27 Keeping JEPA World Models Plannable When Little of the Frame Moves 提出SLIM基准以解决语言目标规划中的帧移动问题 world model world models JEPA
28 ReFract: Benchmarking Perspective Awareness in Language Model Agents with Text World Models 提出ReFract基准以解决语言模型代理的视角意识问题 world model world models large language model
29 Bounded Reachability & Jailbreak Detection via Contraction-Constrained State Space Models 提出收缩约束状态空间模型以解决安全头的可达性与监测问题 Mamba SSM state space model
30 Gains and Collapse in On-Policy Distillation:A Reinforcement Learning Perspective 提出基于强化学习的OPD机制分析以解决生成崩溃问题 reinforcement learning distillation ✅
31 VIGOR: Zero-Shot Visual Generalization via Latent-Space Consistency in Model-Based Reinforcement Learning 提出VIGOR以解决模型基础强化学习中的视觉泛化问题 reinforcement learning latent dynamics model-based RL
32 S2S-JEPA: Predicting the Predictable at Subseasonal-to-Seasonal Timescales 提出S2S-JEPA以解决季节性预测的可预测性问题 JEPA Joint-Embedding Predictive Architecture joint-embedding predictive architecture
33 When a Correct Reward Is Not Enough: Diagnosing and Guiding PPO in an Analytically Solved Broker-Trader Game 在解析解决的经纪人-交易者游戏中诊断与指导PPO reinforcement learning PPO reward shaping
34 DriftTTS: Few-Step Text-to-Speech Without Distillation via Distribution-Matching Drift 提出DriftTTS以解决少步文本到语音合成中的依赖问题 flow matching masked autoencoder distillation
35 MetaRubric: Learning to Reward for Rubric-Based Reinforcement Learning 提出MetaRubric以解决评估标准基础强化学习中的空洞信用问题 reinforcement learning multimodal
36 Learning to Revise Reasoning with Segment-wise On-Policy Distillation 提出段落级蒸馏以改进推理过程中的修正问题 distillation large language model
37 Equivariant Flow Matching for Electron Density Prediction 提出OrbFlow以解决电子密度预测中的计算成本问题 flow matching zero-shot transfer
38 Spend Teacher Tokens Where They Matter: Success-Referenced On-Policy Distillation 提出成功参考的在线策略蒸馏方法以降低教师计算成本 teacher-student distillation
39 CuBEs: Culturally-Situated Behavioral Evaluations and the Limitations of Culture-Blind LLM Judges 提出CuBEs以解决文化盲目评估LLM行为的问题 world model world models large language model
40 Harness-Aware Distillation for Small Language Model Agents 提出Harness-Aware Distillation以提升小型语言模型代理的性能 distillation

🔬 支柱七：动作重定向 (Motion Retargeting) (1 篇)

#题目一句话要点标签🔗⭐
41 LiBRA: Detection-Aware Image Watermark Removal via Bidirectional Latent Optimization 提出LiBRA以解决数字水印去除攻击问题 latent optimization

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
42 AgentTrap: Stateful Feedback Deception against Autonomous Penetration Testing Agents 提出AgentTrap以应对自主渗透测试代理的动态攻击 penetration

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

#题目一句话要点标签🔗⭐
43 Ego2World: Compiling Egocentric Cooking Videos into Executable Worlds for Belief-State Planning 提出Ego2World以解决执行计划环境中的部分观察问题 egocentric

⬅️ 返回 cs.AI 首页 · 🏠 返回主页