cs.LG（2026-04-08）

📊 共 26 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 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (10) 支柱一：机器人控制 (Robot Control) (3) 支柱八：物理动画 (Physics-based Animation) (1)

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

#题目一句话要点标签🔗⭐
1 OmniTabBench: Mapping the Empirical Frontiers of GBDTs, Neural Networks, and Foundation Models for Tabular Data at Scale OmniTabBench：大规模表格数据上GBDT、神经网络和基础模型的经验前沿探索 large language model foundation model
2 Frailty Estimation in Elderly Oncology Patients Using Multimodal Wearable Data and Multi-Instance Learning 提出基于多模态可穿戴数据和多示例学习的老年肿瘤患者虚弱程度评估框架 multimodal
3 STQuant: Spatio-Temporal Adaptive Framework for Optimizer Quantization in Large Multimodal Model Training 提出STQuant框架，通过时空自适应优化器量化降低大模型训练内存占用。 multimodal
4 Geometric Properties of the Voronoi Tessellation in Latent Semantic Manifolds of Large Language Models 研究大型语言模型潜在语义空间中的Voronoi tessellation几何特性，提出margin refinement procedures优化模型。 large language model
5 Bi-level Heterogeneous Learning for Time Series Foundation Models: A Federated Learning Approach 提出双层异构联邦学习方法，用于训练时间序列基础模型，提升异构环境下的泛化能力。 foundation model
6 On the Price of Privacy for Language Identification and Generation 研究差分隐私对语言识别与生成任务的影响，量化隐私保护的代价。 large language model
7 Beyond the Mean: Modelling Annotation Distributions in Continuous Affect Prediction 提出基于Beta分布的连续情感预测模型，建模标注分布以提升性能。 multimodal
8 Selective Neuron Amplification for Training-Free Task Enhancement 提出选择性神经元放大（SNA）方法，无需训练即可提升大语言模型在特定任务上的表现。 large language model
9 MoE Routing Testbed: Studying Expert Specialization and Routing Behavior at Small Scale 提出MoE路由测试平台，用于小规模研究专家特化和路由行为 large language model
10 ConceptTracer: Interactive Analysis of Concept Saliency and Selectivity in Neural Representations ConceptTracer：交互式分析神经表征中概念显著性和选择性的工具 foundation model ✅
11 MoBiE: Efficient Inference of Mixture of Binary Experts under Post-Training Quantization 提出MoBiE以解决MoE模型量化效率问题 large language model ✅
12 ExplainFuzz: Explainable and Constraint-Conditioned Test Generation with Probabilistic Circuits ExplainFuzz：利用概率电路实现可解释和约束条件的测试用例生成 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (10 篇)

#题目一句话要点标签🔗⭐
13 Equivariant Multi-agent Reinforcement Learning for Multimodal Vehicle-to-Infrastructure Systems 提出一种基于等变多智能体强化学习的V2I系统资源优化方法。 reinforcement learning multimodal
14 Epistemic Robust Offline Reinforcement Learning 提出基于不确定性集合的离线强化学习框架，提升策略鲁棒性和泛化性 reinforcement learning SAC offline RL
15 Smart Commander: A Hierarchical Reinforcement Learning Framework for Fleet-Level PHM Decision Optimization 提出Smart Commander以优化军用航空舰队的PHM决策 reinforcement learning deep reinforcement learning DRL
16 Multi-Turn Reasoning LLMs for Task Offloading in Mobile Edge Computing 提出COMLLM框架，利用多轮推理LLM解决移动边缘计算中的任务卸载问题 reinforcement learning deep reinforcement learning DRL
17 The Theorems of Dr. David Blackwell and Their Contributions to Artificial Intelligence 总结David Blackwell理论对AI的贡献，涵盖信息压缩、序贯决策和信息比较。 reinforcement learning RLHF large language model
18 Predictive Representations for Skill Transfer in Reinforcement Learning 提出基于结果预测状态表示的技能迁移强化学习方法 reinforcement learning
19 Extraction of linearized models from pre-trained networks via knowledge distillation 提出基于知识蒸馏的线性化模型提取框架，提升线性模型分类精度。 distillation
20 TwinLoop: Simulation-in-the-Loop Digital Twins for Online Multi-Agent Reinforcement Learning TwinLoop：面向在线多智能体强化学习的仿真环数字孪生，提升适应效率 reinforcement learning
21 Android Coach: Improve Online Agentic Training Efficiency with Single State Multiple Actions Android Coach：通过单状态多动作提升在线Agent训练效率 reinforcement learning PPO
22 A First Guess is Rarely the Final Answer: Learning to Search in the Travelling Salesperson Problem NICO-TSP：学习TSP问题的搜索策略，提升求解效率与泛化性 reinforcement learning imitation learning

🔬 支柱一：机器人控制 (Robot Control) (3 篇)

#题目一句话要点标签🔗⭐
23 Graph Neural ODE Digital Twins for Control-Oriented Reactor Thermal-Hydraulic Forecasting Under Partial Observability 提出GNN-ODE数字孪生模型，用于部分可观测下反应堆热工水力预测与控制 sim-to-real latent dynamics MAE
24 The Rhetoric of Machine Learning 揭示机器学习的修辞本质：从客观建模到说服艺术 manipulation world model world models
25 DDP-SA: Scalable Privacy-Preserving Federated Learning via Distributed Differential Privacy and Secure Aggregation 提出DDP-SA框架，通过分布式差分隐私和安全聚合实现可扩展的隐私保护联邦学习。 MPC

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

#题目一句话要点标签🔗⭐
26 ELC: Evidential Lifelong Classifier for Uncertainty Aware Radar Pulse Classification 提出ELC：一种基于证据理论的终身分类器，用于不确定性感知的雷达脉冲分类 PULSE

⬅️ 返回 cs.LG 首页 · 🏠 返回主页