cs.LG（2025-10-29）
📊 共 9 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) (5 篇)
| # | 题目 | 一句话要点 | 标签 | 🔗 | ⭐ |
|---|---|---|---|---|---|
| 1 | Don't Blind Your VLA: Aligning Visual Representations for OOD Generalization | 提出视觉表征对齐方法，提升VLA模型在OOD泛化中的性能 | vision-language-action VLA | ||
| 2 | Application and Validation of Geospatial Foundation Model Data for the Prediction of Health Facility Programmatic Outputs -- A Case Study in Malawi | 利用地理空间基础模型数据预测卫生设施项目产出：以马拉维为例 | foundation model | ||
| 3 | MemEIC: A Step Toward Continual and Compositional Knowledge Editing | MemEIC：面向视觉-语言模型的持续组合式知识编辑方法 | multimodal | ||
| 4 | FaCT: Faithful Concept Traces for Explaining Neural Network Decisions | FaCT：提出可信的概念追踪方法，用于解释神经网络决策过程 | foundation model | ||
| 5 | Generalizing Test-time Compute-optimal Scaling as an Optimizable Graph | 提出Agent-REINFORCE框架，解决测试时计算量约束下多LLM组合与架构的优化问题。 | large language model |
🔬 支柱二：RL算法与架构 (RL & Architecture) (4 篇)
| # | 题目 | 一句话要点 | 标签 | 🔗 | ⭐ |
|---|---|---|---|---|---|
| 6 | Metis-SPECS: Decoupling Multimodal Learning via Self-distilled Preference-based Cold Start | 提出SPECS框架以解决多模态学习中的冷启动问题 | reinforcement learning DPO distillation | ✅ | |
| 7 | Retrieval-Augmented Multimodal Depression Detection | 提出检索增强的多模态抑郁症检测框架，提升情感理解能力。 | MAE large language model multimodal | ||
| 8 | MaGNet: A Mamba Dual-Hypergraph Network for Stock Prediction via Temporal-Causal and Global Relational Learning | MaGNet：一种用于股票预测的Mamba双超图网络，通过时序因果和全局关系学习。 | Mamba spatiotemporal | ✅ | |
| 9 | Learning Fair Graph Representations with Multi-view Information Bottleneck | 提出FairMIB，通过多视角信息瓶颈学习公平的图表示 | representation learning contrastive learning |