cs.LG（2025-05-07）

📊 共 27 篇论文 | 🔗 5 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (16 🔗3) 支柱九：具身大模型 (Embodied Foundation Models) (8 🔗1) 支柱一：机器人控制 (Robot Control) (3 🔗1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (16 篇)

#题目一句话要点标签🔗⭐
1 ALFEE: Adaptive Large Foundation Model for EEG Representation ALFEE：用于脑电信号表征的自适应大规模预训练模型 representation learning foundation model ✅
2 A Heuristic-Integrated DRL Approach for Phase Optimization in Large-Scale RISs 提出一种启发式集成的DRL方法，用于大规模RIS的相位优化 reinforcement learning deep reinforcement learning DRL
3 On-Device LLM for Context-Aware Wi-Fi Roaming 提出一种基于设备端LLM的上下文感知Wi-Fi漫游方法，优化无线网络连接。 DRL large language model chain-of-thought
4 InfoNCE is a Free Lunch for Semantically guided Graph Contrastive Learning IFL-GCL：利用InfoNCE进行语义引导的图对比学习，解决负样本偏差问题 contrastive learning foundation model ✅
5 ARDNS-FN-Quantum: A Quantum-Enhanced Reinforcement Learning Framework with Cognitive-Inspired Adaptive Exploration for Dynamic Environments 提出ARDNS-FN-Quantum框架以解决动态环境中的强化学习探索问题 reinforcement learning PPO
6 ABKD: Pursuing a Proper Allocation of the Probability Mass in Knowledge Distillation via $α$-$β$-Divergence ABKD：通过α-β散度实现知识蒸馏中概率质量的合理分配 teacher-student distillation ✅
7 Adaptive and Robust DBSCAN with Multi-agent Reinforcement Learning 提出AR-DBSCAN，利用多智能体强化学习自适应解决DBSCAN在多密度数据集上的聚类问题。 reinforcement learning deep reinforcement learning
8 DMRL: Data- and Model-aware Reward Learning for Data Extraction 提出DMRL：一种数据与模型感知的奖励学习方法，用于从大型语言模型中提取敏感数据。 reinforcement learning inverse reinforcement learning large language model
9 Fight Fire with Fire: Defending Against Malicious RL Fine-Tuning via Reward Neutralization 提出奖励中和方法，防御恶意RL微调对语言模型的安全攻击。 reinforcement learning large language model
10 A Two-Timescale Primal-Dual Framework for Reinforcement Learning via Online Dual Variable Guidance 提出PGDA-RL以解决强化学习中的数据利用与探索问题 reinforcement learning
11 Distillation-Enabled Knowledge Alignment Protocol for Semantic Communication in AI Agent Networks 提出DeKAP协议，通过知识蒸馏对齐AI Agent网络中的语义通信知识。 distillation
12 A Survey on Temporal Interaction Graph Representation Learning: Progress, Challenges, and Opportunities 时序交互图表示学习综述：进展、挑战与机遇 representation learning
13 Extending a Quantum Reinforcement Learning Exploration Policy with Flags to Connect Four 扩展量子强化学习探索策略，应用于四子棋游戏 reinforcement learning
14 Optimization of Infectious Disease Intervention Measures Based on Reinforcement Learning -- Empirical analysis based on UK COVID-19 epidemic data 基于强化学习优化传染病干预措施，以英国COVID-19疫情数据为实证分析 reinforcement learning
15 Putting the Value Back in RL: Better Test-Time Scaling by Unifying LLM Reasoners With Verifiers RL$^V$: 统一LLM推理器与验证器，提升强化学习测试时可扩展性 reinforcement learning PPO
16 Deep Learning Innovations for Energy Efficiency: Advances in Non-Intrusive Load Monitoring and EV Charging Optimization for a Sustainable Grid 利用深度学习创新，通过非侵入式负荷监测和电动汽车充电优化，提升能源效率。 reinforcement learning deep reinforcement learning

🔬 支柱九：具身大模型 (Embodied Foundation Models) (8 篇)

#题目一句话要点标签🔗⭐
17 ORBIT-2: Scaling Exascale Vision Foundation Models for Weather and Climate Downscaling ORBIT-2：通过可扩展的视觉基础模型实现天气和气候的降尺度 foundation model
18 Lossless Compression of Large Language Model-Generated Text via Next-Token Prediction 利用下一token预测，实现对大语言模型生成文本的无损压缩 large language model
19 Nature's Insight: A Novel Framework and Comprehensive Analysis of Agentic Reasoning Through the Lens of Neuroscience 提出神经科学启发的智能体推理框架，提升AI自主性和泛化能力 foundation model multimodal chain-of-thought ✅
20 A Large Language Model for Feasible and Diverse Population Synthesis 提出基于LLM和贝叶斯网络的混合方法，用于生成可行且多样化的人口合成数据。 large language model
21 When Bad Data Leads to Good Models 通过预训练阶段引入不良数据，提升后训练阶段大语言模型的可控性与安全性。 large language model
22 Towards Effectively Leveraging Execution Traces for Program Repair with Code LLMs 利用程序执行轨迹增强代码大语言模型以提升程序修复效果 large language model
23 STRGCN: Capturing Asynchronous Spatio-Temporal Dependencies for Irregular Multivariate Time Series Forecasting 提出STRGCN，用于解决不规则多元时间序列预测中异步时空依赖建模难题。 TAMP
24 Rethinking LLM Advancement: Compute-Dependent and Independent Paths to Progress 区分计算依赖与独立创新，评估算法进步对大语言模型能力的影响 large language model

🔬 支柱一：机器人控制 (Robot Control) (3 篇)

#题目一句话要点标签🔗⭐
25 Merging and Disentangling Views in Visual Reinforcement Learning for Robotic Manipulation 提出MAD算法，通过融合与解耦多视角信息，提升机器人操作视觉强化学习的效率与鲁棒性。 manipulation reinforcement learning ✅
26 Trajectory Entropy Reinforcement Learning for Predictable and Robust Control 提出轨迹熵强化学习，提升控制策略的预测性与鲁棒性 locomotion reinforcement learning deep reinforcement learning
27 Robust ML Auditing using Prior Knowledge 提出一种利用先验知识的鲁棒机器学习审计方法，防止平台操纵审计结果。 manipulation

⬅️ 返回 cs.LG 首页 · 🏠 返回主页