cs.LG（2025-07-28）

📊 共 30 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) (12 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (9) 支柱八：物理动画 (Physics-based Animation) (4 🔗1) 支柱一：机器人控制 (Robot Control) (2) 支柱三：空间感知与语义 (Perception & Semantics) (1) 支柱四：生成式动作 (Generative Motion) (1) 支柱五：交互与反应 (Interaction & Reaction) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (12 篇)

#题目一句话要点标签🔗⭐
1 Advancing Compositional LLM Reasoning with Structured Task Relations in Interactive Multimodal Communications 提出ContextLoRA和ContextGear，利用单个组合LLM解决交互式多模态通信中的任务推理问题。 large language model multimodal
2 Adaptive Multimodal Protein Plug-and-Play with Diffusion-Based Priors Adam-PnP：自适应多模态蛋白质结构生成，利用扩散先验解决逆问题。 multimodal
3 When Brain Foundation Model Meets Cauchy-Schwarz Divergence: A New Framework for Cross-Subject Motor Imagery Decoding 提出基于脑基础模型和Cauchy-Schwarz散度的跨个体运动想象脑电解码框架 foundation model
4 SmallThinker: A Family of Efficient Large Language Models Natively Trained for Local Deployment SmallThinker：为本地部署原生设计的高效大语言模型家族 large language model
5 Prostate Cancer Classification Using Multimodal Feature Fusion and Explainable AI 提出基于BERT和随机森林的多模态融合可解释AI系统，用于前列腺癌分类。 multimodal
6 A Contrastive Diffusion-based Network (CDNet) for Time Series Classification 提出基于对比扩散的CDNet，提升时间序列分类在复杂数据条件下的性能。 multimodal
7 Predicting Cognition from fMRI:A Comparative Study of Graph, Transformer, and Kernel Models Across Task and Rest Conditions 对比图、Transformer和核模型，论文旨在使用fMRI预测认知能力，并探索任务态和静息态数据的影响。 multimodal
8 Filtering with Self-Attention and Storing with MLP: One-Layer Transformers Can Provably Acquire and Extract Knowledge 通过自注意力过滤和MLP存储：单层Transformer可证明地获取和提取知识 large language model
9 Repairing vulnerabilities without invisible hands. A differentiated replication study on LLMs 通过差异化复现研究揭示LLM自动修复漏洞的潜在因素 large language model
10 PROVCREATOR: Synthesizing Complex Heterogenous Graphs with Node and Edge Attributes ProvCreator：用于合成具有节点和边属性的复杂异构图 large language model
11 Dissecting Persona-Driven Reasoning in Language Models via Activation Patching 通过激活干预剖析语言模型中由角色驱动的推理机制 large language model
12 First Hallucination Tokens Are Different from Conditional Ones 揭示LLM幻觉Token分布特性：首个幻觉Token更易检测 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🔗⭐
13 Large Language Model-Enhanced Reinforcement Learning for Diverse and Novel Recommendations 提出LAAC：利用大语言模型增强强化学习，实现多样性和新颖性推荐 reinforcement learning large language model
14 Learning Treatment Policies From Multimodal Electronic Health Records 提出一种基于专家标注的多模态EHR治疗策略学习方法，提升治疗效果估计。 policy learning predictive model multimodal
15 Flow Matching Policy Gradients 提出Flow Policy Optimization (FPO)，将Flow Matching融入策略梯度，解决强化学习中策略优化问题。 reinforcement learning PPO flow matching
16 DmC: Nearest Neighbor Guidance Diffusion Model for Offline Cross-domain Reinforcement Learning DmC：基于近邻引导扩散模型的离线跨域强化学习 reinforcement learning policy learning offline RL
17 LoRA-PAR: A Flexible Dual-System LoRA Partitioning Approach to Efficient LLM Fine-Tuning LoRA-PAR：一种灵活的双系统LoRA划分方法，用于高效LLM微调 reinforcement learning chain-of-thought
18 Geometry of Neural Reinforcement Learning in Continuous State and Action Spaces 通过几何分析揭示连续状态-动作空间中神经强化学习的状态空间维度特性 reinforcement learning
19 Kimi K2: Open Agentic Intelligence Kimi K2：开源Agentic智能大模型，采用MuonClip优化器，实现卓越的Agent能力。 reinforcement learning large language model
20 Attributed Graph Clustering with Multi-Scale Weight-Based Pairwise Coarsening and Contrastive Learning 提出MPCCL模型，通过多尺度图粗化和对比学习提升属性图聚类性能 contrastive learning
21 RadioMamba: Breaking the Accuracy-Efficiency Trade-off in Radio Map Construction via a Hybrid Mamba-UNet RadioMamba：混合Mamba-UNet突破无线电地图构建的精度-效率权衡 Mamba

🔬 支柱八：物理动画 (Physics-based Animation) (4 篇)

#题目一句话要点标签🔗⭐
22 Multiscale geometrical and topological learning in the analysis of soft matter collective dynamics 提出基于多尺度几何与拓扑学习的软物质集体动力学分析方法 spatiotemporal
23 Bubbleformer: Forecasting Boiling with Transformers Bubbleformer：利用Transformer预测沸腾现象，实现自主长程动态预测。 spatiotemporal
24 Locally Adaptive Conformal Inference for Operator Models 提出局部自适应共形推断(LSCI)框架，用于算子模型的函数值预测不确定性量化。 spatiotemporal
25 PySHRED: A Python package for SHallow REcurrent Decoding for sparse sensing, model reduction and scientific discovery PySHRED：用于稀疏感知、模型降阶和科学发现的浅层循环解码Python包 spatiotemporal ✅

🔬 支柱一：机器人控制 (Robot Control) (2 篇)

#题目一句话要点标签🔗⭐
26 Exposing the Illusion of Fairness: Auditing Vulnerabilities to Distributional Manipulation Attacks 揭示公平性幻觉：审计算法对分布操纵攻击的脆弱性 manipulation
27 Rep-MTL: Unleashing the Power of Representation-level Task Saliency for Multi-Task Learning 提出Rep-MTL以解决多任务学习中的任务互补性问题 manipulation representation learning

🔬 支柱三：空间感知与语义 (Perception & Semantics) (1 篇)

#题目一句话要点标签🔗⭐
28 DEM-NeRF: A Neuro-Symbolic Method for Scientific Discovery through Physics-Informed Simulation 提出DEM-NeRF，通过物理信息模拟实现科学发现，从稀疏图像重建弹性物体。 NeRF neural radiance field spatiotemporal

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
29 Modular Delta Merging with Orthogonal Constraints: A Scalable Framework for Continual and Reversible Model Composition 提出MDM-OC框架，实现可扩展、无干扰、可逆的模型组合与持续学习 MDM

🔬 支柱五：交互与反应 (Interaction & Reaction) (1 篇)

#题目一句话要点标签🔗⭐
30 HIAL: A New Paradigm for Hypergraph Active Learning via Influence Maximization 提出HIAL：一种基于影响最大化的超图主动学习新范式 HOI

⬅️ 返回 cs.LG 首页 · 🏠 返回主页