cs.AI（2025-10-21）

📊 共 34 篇论文 | 🔗 2 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (23 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (9) 支柱四：生成式动作 (Generative Motion) (1 🔗1) 支柱八：物理动画 (Physics-based Animation) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (23 篇)

#题目一句话要点标签🔗⭐
1 Automated urban waterlogging assessment and early warning through a mixture of foundation models 提出UWAssess，利用混合基础模型自动评估城市内涝并预警 foundation model chain-of-thought
2 CytoNet: A Foundation Model for the Human Cerebral Cortex CytoNet：用于人脑皮层分析的基础模型，实现细胞级微观结构理解 foundation model
3 Counterfactual Reasoning for Steerable Pluralistic Value Alignment of Large Language Models 提出COUPLE框架，利用反事实推理实现大语言模型对多元价值的可控对齐 large language model
4 Unifying Inductive, Cross-Domain, and Multimodal Learning for Robust and Generalizable Recommendation MICRec：融合归纳、跨域和多模态学习的鲁棒通用推荐框架 multimodal
5 Illusions of reflection: open-ended task reveals systematic failures in Large Language Models' reflective reasoning 揭示大语言模型反思推理的系统性缺陷：开放任务下约束违反 large language model ✅
6 The MUSE Benchmark: Probing Music Perception and Auditory Relational Reasoning in Audio LLMS MUSE基准测试：用于评估音频LLM音乐感知和听觉关系推理能力 large language model multimodal chain-of-thought
7 Can Reasoning Models Obfuscate Reasoning? Stress-Testing Chain-of-Thought Monitorability 提出CoT混淆压力测试方法，评估推理模型在对抗环境下的可监控性 chain-of-thought
8 HarmNet: A Framework for Adaptive Multi-Turn Jailbreak Attacks on Large Language Models HarmNet：一种用于大语言模型的多轮自适应越狱攻击框架 large language model
9 Exploring Membership Inference Vulnerabilities in Clinical Large Language Models 探索临床大语言模型中的成员推理漏洞，评估患者隐私泄露风险 large language model
10 StarBench: A Turn-Based RPG Benchmark for Agentic Multimodal Decision-Making and Information Seeking StarBench：一个用于智能体多模态决策与信息寻求的回合制RPG基准 multimodal
11 PlanU: Large Language Model Reasoning through Planning under Uncertainty 提出PlanU：通过不确定性下的规划增强大语言模型推理能力 large language model
12 Earth AI: Unlocking Geospatial Insights with Foundation Models and Cross-Modal Reasoning Earth AI：利用基础模型和跨模态推理解锁地理空间洞察 foundation model
13 A Justice Lens on Fairness and Ethics Courses in Computing Education: LLM-Assisted Multi-Perspective and Thematic Evaluation 利用LLM多视角评估，提升计算教育中公平与伦理课程的教学设计。 large language model
14 Cultural Alien Sampler: Open-ended art generation balancing originality and coherence 提出文化异类采样器(CAS)，在开放式艺术生成中平衡原创性和连贯性。 large language model
15 Test-time Verification via Optimal Transport: Coverage, ROC, & Sub-optimality 基于最优传输的测试时验证：揭示覆盖率、ROC与次优性之间的关系 large language model
16 Prompt Decorators: A Declarative and Composable Syntax for Reasoning, Formatting, and Control in LLMs 提出Prompt Decorators以解决LLMs控制不足问题 large language model
17 LAFA: Agentic LLM-Driven Federated Analytics over Decentralized Data Sources LAFA：基于Agentic LLM的去中心化数据联邦分析框架 large language model
18 Probabilistic Modeling of Intentions in Socially Intelligent LLM Agents 提出基于概率意图建模的LLM Agent框架，提升社交对话中的智能水平。 large language model
19 CircuitSeer: Mining High-Quality Data by Probing Mathematical Reasoning Circuits in LLMs CircuitSeer：通过探查LLM数学推理电路挖掘高质量数据 large language model
20 Genesis: Evolving Attack Strategies for LLM Web Agent Red-Teaming Genesis：演化攻击策略，用于LLM Web Agent的红队测试 large language model
21 Prospects for Using Artificial Intelligence to Understand Intrinsic Kinetics of Heterogeneous Catalytic Reactions 利用人工智能理解非均相催化反应的本征动力学 multimodal
22 EdgeReasoning: Characterizing Reasoning LLM Deployment on Edge GPUs EdgeReasoning：表征边缘GPU上推理LLM的部署，优化延迟-精度权衡 large language model
23 ssToken: Self-modulated and Semantic-aware Token Selection for LLM Fine-tuning 提出ssToken，通过自调制和语义感知选择token，提升LLM微调效果。 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🔗⭐
24 Visual Attention Reasoning via Hierarchical Search and Self-Verification 提出Visual Attention Reasoning框架，解决多模态大模型中的幻觉问题。 reinforcement learning large language model multimodal
25 Lyapunov-Aware Quantum-Inspired Reinforcement Learning for Continuous-Time Vehicle Control: A Feasibility Study 提出Lyapunov感知的量子启发强化学习框架，用于连续时间车辆控制。 reinforcement learning policy learning
26 CodeRL+: Improving Code Generation via Reinforcement with Execution Semantics Alignment CodeRL+：通过执行语义对齐强化学习提升代码生成能力 reinforcement learning distillation large language model
27 Rectifying Shortcut Behaviors in Preference-based Reward Learning 提出PRISM，缓解基于偏好的奖励学习中的捷径行为，提升泛化性。 reinforcement learning large language model
28 Extracting alignment data in open models 提出一种从后训练模型中提取对齐训练数据的方法，用于提升模型能力。 distillation instruction following
29 DeLoad: Demand-Driven Short-Video Preloading with Scalable Watch-Time Estimation DeLoad：通过可扩展观看时长估计实现需求驱动的短视频预加载 reinforcement learning deep reinforcement learning DRL
30 On AI Verification in Open RAN 提出基于决策树的轻量级AI验证方法，保障Open RAN中DRL智能体的可靠性。 reinforcement learning deep reinforcement learning DRL
31 REPAIR Approach for Social-based City Reconstruction Planning in case of natural disasters 提出REPAIR方法，利用深度强化学习进行自然灾害后的城市重建规划，最大化社会效益。 reinforcement learning deep reinforcement learning
32 Heterogeneous Adversarial Play in Interactive Environments 提出异构对抗博弈(HAP)框架，解决交互环境中非对称自学习问题。 curriculum learning teacher-student

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
33 DiffGRM: Diffusion-based Generative Recommendation Model 提出DiffGRM，一种基于扩散模型的生成式推荐模型，解决语义ID的结构性问题。 MDM ✅

🔬 支柱八：物理动画 (Physics-based Animation) (1 篇)

#题目一句话要点标签🔗⭐
34 Local Guidance for Configuration-Based Multi-Agent Pathfinding 提出基于局部引导的多智能体路径规划方法，提升配置空间搜索效率。 spatiotemporal

⬅️ 返回 cs.AI 首页 · 🏠 返回主页