cs.LG（2026-06-15）

📊 共 30 篇论文 | 🔗 4 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (14 🔗1) 支柱九：具身大模型 (Embodied Foundation Models) (9 🔗1) 支柱一：机器人控制 (Robot Control) (5 🔗1) 支柱八：物理动画 (Physics-based Animation) (1 🔗1) 支柱五：交互与反应 (Interaction & Reaction) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (14 篇)

#题目一句话要点标签🔗⭐
1 Diffusion Offline Reinforcement Learning for Fair and Energy-Efficient UAV-Assisted Wireless Networks 提出扩散离线强化学习以优化无人机辅助无线网络的能效与公平性 reinforcement learning policy learning SAC
2 Phys-JEPA: Physics-Informed Latent World Models for Multivariate Time-Series Forecasting 提出Phys-JEPA以解决多变量时间序列预测中的物理一致性问题 world model world models JEPA
3 BRICKS-WM: Building Reusability via Interface Composition Kinetics for Structured World Models 提出BRICKS-WM以解决模型重用性不足问题 reinforcement learning world model world models
4 Maximum Entropy Inverse Reinforcement Learning for Mean-Field Games with Average Reward 提出最大熵逆强化学习解决平均奖励均场博弈问题 reinforcement learning inverse reinforcement learning
5 Taylor-Calibrate: Principled Initialization for Hybrid Linear Attention Distillation 提出Taylor-Calibrate以解决混合线性注意力蒸馏初始化问题 linear attention distillation
6 How Post-Training Shapes Biological Reasoning Models 研究后训练对生物推理模型的影响 reinforcement learning foundation model multimodal
7 A Unified Causal-Origin Taxonomy of Distributional Shifts in Reinforcement Learning 提出统一因果起源分类法以解决强化学习中的分布转移问题 reinforcement learning
8 SPICE: Synergy and Partial Information Based Curriculum Evolution 提出SPICE以解决多模态学习中的样本复杂性适应问题 curriculum learning multimodal
9 Diffusion Flow Matching: Dimension-Improved KL Bounds and Wasserstein Guarantees 提出扩展的KL界限与Wasserstein保证以改进扩散流匹配 flow matching
10 Autonomous End-to-End SOH Prediction Services for Battery Systems via Temporal-Contrastive Representation Learning 提出TC-SOH以解决锂离子电池健康状态预测问题 representation learning
11 From Tokens to Regions: CUDA-Sensitive Instruction Tuning for GPU Kernel Generation 提出CuSeT以解决CUDA内核生成中的敏感性问题 reinforcement learning large language model
12 Graphical conditional generative modeling for digital twin modeling 提出图形条件生成建模以解决数字双胞胎建模中的复杂性问题 reinforcement learning multimodal
13 ExpRL: Exploratory RL for LLM Mid-Training 提出ExpRL以解决稀疏奖励强化学习的覆盖不足问题 reinforcement learning distillation
14 Infant Spontaneous Movement Noise Improves Exploration in Deep RL 提出基于婴儿自发运动噪声的探索机制以提升深度强化学习效率 reinforcement learning deep reinforcement learning ✅

🔬 支柱九：具身大模型 (Embodied Foundation Models) (9 篇)

#题目一句话要点标签🔗⭐
15 Scalable Circuit Learning for Interpreting Large Language Models 提出CircuitLasso以解决大语言模型的可解释性问题 large language model
16 Multimodal Evaluator Preference Collapse: Cross-Modal Contagion in Self-Evolving Agents 提出多模态评估偏好崩溃模型以解决自我演化代理中的偏见问题 multimodal
17 MUNI: Multimodal Unified Latent Diffusion for Coherent Any-to-Any Generation 提出MUNI以解决多模态生成中的一致性问题 multimodal ✅
18 MIRAGE: Auditing Anti-Muslim Bias in Frontier LLMs Across Reasoning, Agentic, and Time-Coupled Conditions 提出MIRAGE基准以审计前沿LLM中的反穆斯林偏见 large language model chain-of-thought
19 Multi-Modal Spatio-Temporal Graph Neural Network with Mixture of Experts for Soil Organic Carbon Prediction 提出多模态时空图神经网络以解决土壤有机碳预测问题 foundation model
20 Phantoms and Disclosures: a Causal Framework for Auditing Synthetic Data 提出可定制的审计框架以检测合成数据中的隐私泄露 large language model
21 Privacy from Symmetry: Orthogonally Equivariant Transformers for LLM Inference 提出正交混淆方法以保护大型语言模型推理隐私 large language model
22 Communication-Efficient Verifiable Attention for LLM Inference 提出通信高效的可验证注意力机制以加速LLM推理 large language model
23 Auditing Machine Unlearning: A Systematic Research on Whether Models Truly Forget 提出实用审计框架以解决机器遗忘算法的有效性问题 large language model

🔬 支柱一：机器人控制 (Robot Control) (5 篇)

#题目一句话要点标签🔗⭐
24 FlowMPC: Improving Flow Matching policies with World Models 提出FlowMPC以提升多模态行为克隆策略的表现 manipulation behavior cloning flow matching
25 Evolutionary Bilevel Reward Shaping for Generalization in Reinforcement Learning 提出进化双层奖励塑形以解决强化学习泛化问题 domain randomization reinforcement learning reward shaping
26 Direction-Conditioned Policies via Compositional Subgoal Scoring for Online Goal-Conditioned Reinforcement Learning 提出方向条件策略以解决在线目标条件强化学习问题 manipulation reinforcement learning
27 PhysGuard: Fisher-Guided Gradient Projection for Sim-to-Real Neural PDE Surrogates 提出PhysGuard以解决神经算子在仿真到现实中的适应问题 sim-to-real ✅
28 TCHG: Tri-Trust Conditioned Heterogeneous Graph Learning for Reliable Dynamic Trust Prediction 提出TCHG框架以解决动态信任预测中的异构信任证据问题 manipulation

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

#题目一句话要点标签🔗⭐
29 HawkesNest: A Multi-Axis Synthetic Benchmark for Spatiotemporal Pattern Complexity 提出HawkesNest以解决时空点过程模型评估问题 spatiotemporal ✅

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

#题目一句话要点标签🔗⭐
30 FEnc$^2$: Unifying Data Packing for Efficient Private Inference via Convolution and Architecture-Aware Fragment Encoding 提出FEnc²以解决同态加密推理中的数据打包效率问题 OMOMO

⬅️ 返回 cs.LG 首页 · 🏠 返回主页