cs.LG（2026-06-12）

📊 共 93 篇论文 | 🔗 1 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (44) 支柱九：具身大模型 (Embodied Foundation Models) (36 🔗1) 支柱四：生成式动作 (Generative Motion) (4) 支柱一：机器人控制 (Robot Control) (4) 支柱八：物理动画 (Physics-based Animation) (3) 支柱五：交互与反应 (Interaction & Reaction) (1) 支柱七：动作重定向 (Motion Retargeting) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (44 篇)

#题目一句话要点标签🔗⭐
1 M*: A Modular, Extensible, Serving System for Multimodal Models 提出M*以解决多模态模型服务效率问题 world model world models JEPA
2 LLM-ODDR: A Large Language Model Framework for Joint Order Dispatching and Driver Repositioning 提出LLM-ODDR框架以解决网约车订单调度与司机重定位问题 reinforcement learning distillation spatiotemporal
3 WOMBET: World Model-Based Experience Transfer for Robust and Sample-efficient Reinforcement Learning 提出WOMBET以解决强化学习中的经验转移问题 reinforcement learning world model world models
4 BrainDINO: A Brain MRI Foundation Model for Generalizable Clinical Representation Learning 提出BrainDINO以解决脑MRI任务特定性与数据稀缺问题 representation learning foundation model
5 EPM-JEPA: Operator-Side Experience Modulation in JEPA-Family World Models 提出EPM-JEPA以解决JEPA模型在动态变化中的适应性问题 world model world models JEPA
6 ARROW: Augmented Replay for RObust World models 提出ARROW以解决持续强化学习中的遗忘问题 reinforcement learning world model world models
7 Scale Buys Interpolation, Structure Buys a Horizon: Certified Predictability for Equivariant World Models 提出可验证的预测地平线以解决世界模型的信任问题 world model world models JEPA
8 Extracting Governing Equations from Latent Dynamics via Multi-View Contrastive Learning 提出DYSCO算法以从噪声数据中提取潜在动力学方程 latent dynamics representation learning contrastive learning
9 When to Align, When to Predict: A Phase Diagram for Multimodal Learning 提出统一框架以优化多模态学习中的对齐与预测问题 representation learning multimodal
10 ProPlay: Procedural World Models for Self-Evolving LLM Agents 提出ProPlay以解决自我进化智能体在部分可观察环境中的学习问题 world model world models
11 Boosting Direct Preference Optimization with Penalization 提出DPOP以提升离线偏好优化的效果 reinforcement learning DPO direct preference optimization
12 Demystifying Hidden-State Recurrence: Switchable Latent Reasoning with On-Policy Reinforcement Learning 提出SWITCH框架以解决隐状态递归的优化与可解释性问题 reinforcement learning chain-of-thought
13 Equivariant Flow Matching for Symmetry-Breaking Bifurcation Problems 提出流匹配方法以解决对称破缺分岔问题 flow matching multimodal
14 BrainPro: Towards Large-scale Brain State-aware EEG Representation Learning 提出BrainPro以解决EEG信号跨布局对齐问题 representation learning foundation model
15 QoS Improvement in Multi User Cellular-Symbiotic Radio Network Assisted by Active-STAR-RIS 提出ASRIS以提升多用户蜂窝网络的服务质量 reinforcement learning deep reinforcement learning DRL
16 Individual Control Barrier Functions-Guided Diffusion Model for Safe Offline Multi-Agent Reinforcement Learning 提出个体控制屏障函数引导的扩散模型以解决安全离线多智能体强化学习问题 reinforcement learning offline reinforcement learning
17 Keep Policy Gradient in Charge: Sibling-Guided Credit Distillation for Long-Horizon Tool-Use Agents 提出Sibling-Guided Credit Distillation以解决长时间工具使用的强化学习问题 reinforcement learning distillation
18 LongSpike: Fractional Order Spiking State Space Models for Efficient Long Sequence Learning 提出LongSpike以解决长序列学习中的记忆瓶颈问题 SSM state space model
19 $α$-fair heterogeneous agent reinforcement learning 提出α-公平异构智能体强化学习以解决公平性与效率问题 reinforcement learning reward shaping
20 A Unifying Lens on Reward Uncertainty in RLHF 提出分布式奖励模型以缓解RLHF中的奖励不确定性问题 reinforcement learning RLHF
21 One Step Closer to Ground Truth: A Multi-Scale Residual-Aware Representation Learning Pipeline for Predicting Time Series Data 提出多尺度残差感知表示学习管道以改进时间序列预测 representation learning MAE
22 PolyFlow: Safe and Efficient Polytope-Constrained Flow Matching with Constraint Embedding and Projection-free Update 提出PolyFlow以解决安全关键系统中的流匹配问题 flow matching
23 Reinforcement Learning for Neural Model Editing 提出强化学习框架以实现神经模型编辑 reinforcement learning
24 ReCal: Reward Calibration for RL-based LLM Routing 提出ReCal框架以解决RL基础LLM路由中的奖励校准问题 reinforcement learning large language model
25 PolicyGuard: Towards Test-time and Step-level Adversary Defense for Reinforcement Learning Agent 提出PolicyGuard以解决强化学习代理的后门攻击问题 reinforcement learning
26 From Uncertain Judgments to Calibrated Rankings: Conformal Elo Estimation for LLM Evaluation 提出符合性Elo估计以解决LLM评估中的系统性误差问题 MAE large language model
27 Rarity-Gated Context Conditioning for Offline Imitation Learning-Based Maritime Anomaly Detection 提出Rarity-Gated特征调制以解决海事异常检测中的频率偏差问题 imitation learning
28 Dense Supervision, Sparse Updates: On the Sparsity and Geometry of On-Policy Distillation 提出稀疏更新机制以优化在政策蒸馏的效果 distillation
29 Graph Reinforcement Learning for Calibration-Aware Quantum Circuit Routing 提出基于图强化学习的量子电路路由方法以提高校准精度 reinforcement learning
30 PlaceRep: Geospatial Place Representation Learning from Large-Scale Point-of-Interest Data 提出PlaceRep以解决城市环境表示学习中的POI聚合问题 representation learning
31 One Token to Fool LLM-as-a-Judge 提出数据增强策略以解决大型语言模型的奖励黑客问题 reinforcement learning large language model
32 Metriplectic Conditional Flow Matching for Dissipative Dynamics 提出Metriplectic条件流匹配以解决耗散动力学问题 flow matching
33 Decentralized Autoregressive Generation 提出去中心化自回归生成方法以解决扩展瓶颈问题 flow matching multimodal
34 When Smaller Wins: Dual-Stage Distillation and Pareto-Guided Compression of Liquid Neural Networks for Edge Battery Prognostics 提出DLNet框架以优化边缘设备电池健康预测 distillation
35 GEMSS: A Variational Bayesian Method for Discovering Multiple Sparse Solutions in Classification and Regression Problems 提出GEMSS方法以发现多重稀疏解的分类与回归问题 predictive model multimodal
36 Disentangling Dynamical Systems: Causal Representation Learning Meets Local Sparse Attention 提出一种新方法以识别动态系统的因果结构 representation learning
37 Universal Time Series Generation with Neural Controlled Differential Equations 提出G-SLiCEs以解决时间序列生成问题 flow matching SSM state space model
38 Reliability of Probabilistic Emulation of Physical Systems 提出框架评估物理系统的概率仿真可靠性 flow matching spatiotemporal
39 Decoding Insect Song: A Multitask Semisupervised Orthoptera Bioacoustic Classifier 提出PULSE框架以解决生态声学分类的局限性问题 distillation PULSE
40 Understanding helpfulness and harmless tension in reward models 研究奖励模型中的有益性与无害性张力 reinforcement learning RLHF
41 One Transit Is All You Need: Detecting Exoplanets Through Learned Stellar Behaviour with EXOVEIL 提出EXOVEIL以解决单次过境行星检测问题 world model world models
42 Provably Safe, Yet Scalable Reinforcement Learning 提出PS2-RL框架以解决安全强化学习的可扩展性问题 reinforcement learning
43 Be My Tutor: On-Policy Co-Distillation for Mutual LLM Improvement via Peer Feedback 提出On-Policy Co-Distillation以实现多领域LLM的互助提升 distillation
44 Contract-Based Compositional Shielding for Safe Multi-Agent Reinforcement Learning 提出基于合同的组合屏障以解决安全多智能体强化学习问题 reinforcement learning

🔬 支柱九：具身大模型 (Embodied Foundation Models) (36 篇)

#题目一句话要点标签🔗⭐
45 Emerging Flexible Designs for Geospatial Multimodal Foundation Models 提出灵活设计以提升地理空间多模态基础模型性能 foundation model multimodal
46 CausalMoE: A Billion-Scale Multimodal Foundation Model for Granger Causal Discovery with Pattern-Routed Heterogeneous Experts 提出CausalMoE以解决复杂系统中的格兰杰因果发现问题 foundation model multimodal
47 Earth Science Foundation Models: From Perception to Reasoning and Discovery 提出地球科学基础模型以解决多模态数据整合问题 foundation model multimodal
48 Multimodal Graph Negative Learning 提出GraphMNL以解决多模态图节点表示不平衡问题 multimodal
49 SMGFM: Spectral Multimodal Graph Pretraining for Multimodal-Attributed Graphs 提出SMGFM以解决多模态属性图的语义融合问题 multimodal
50 DeepJEB++: Foundation Model-Driven Large-Scale 3D Engineering Dataset via 2D Latent Space Augmentation 提出DeepJEB++以解决大规模3D工程数据集构建问题 foundation model
51 Authority, Truth, and Citation Bias: A Large-Scale Multi-Domain Benchmark for Studying Epistemic Susceptibility in Large Language Models 提出AuthorityBench基准以研究语言模型中的引用偏见问题 large language model
52 Hölder++: Improving the Quality-Coherence Trade-off in Multimodal VAEs 提出Hölder++以改善多模态变分自编码器的生成质量与一致性问题 multimodal
53 Beyond the Commitment Boundary: Probing Epiphenomenal Chain-of-Thought in Large Reasoning Models 提出早期退出机制以优化大规模推理模型的链式思维 chain-of-thought
54 Decoding the Multimodal Maze: A Systematic Review on the Adoption of Explainability in Multimodal Attention-based Models 系统评估多模态注意力模型的可解释性研究 multimodal
55 Retrieval-Augmented Foundation Models for Water Level Prediction in the Everglades 提出检索增强基础模型以解决埃弗格雷德水位预测问题 foundation model
56 Hellinger Multimodal Variational Autoencoders 提出Hellinger多模态变分自编码器以提升生成学习效果 multimodal
57 Representing Time Series as Structured Programs for LLM Reasoning 提出T2SP方法以解决时间序列分析中的表示问题 large language model instruction following
58 DynamicPTQ: Mitigating Activation Quantization Collapse via Residual-Stream Dynamics 提出DynamicPTQ以解决激活量化崩溃问题 large language model
59 Net-Ev$^2$: A Generative Simulator for Network Event Evolution 提出Net-Ev$^2$以解决网络事件演化模拟问题 multimodal
60 LLM-Powered Personalized Glycemic Assessment in Type 2 Diabetes with Wearable Sensor Data 提出GlyLLM框架以提升2型糖尿病个性化血糖评估 large language model
61 A Stabilized Path-Space Approach to Diffusion-Based Posterior Sampling 提出稳定路径空间框架以改进扩散模型后验采样 multimodal
62 Detecting Functional Memorization in Code Language Models 提出功能性记忆检测方法以解决代码生成模型的审计问题 large language model
63 Multi-Bitwidth Quantization for LLMs Using Additive Codebooks 提出Drop-by-Drop框架以实现LLM的多位宽量化 large language model
64 Where Computation Lives Inside TabPFN: Causal Localisation of Attention Head Function 提出TabPFN的因果机制分析以优化特征注意力头的计算分配 foundation model
65 Existence Precedes Value: Joint Modeling of Observational Existence and Evolving States in Time Series Forecasting 提出Timeflies框架以解决时间序列预测中的观测性问题 TAMP
66 Feature-preserving Latent-EnKF for Data Assimilation of Flows with Shocks 提出特征保留的潜在-EnKF以解决流动中的冲击数据同化问题 multimodal
67 GenAutoML: An Agentic Framework for Dynamic Architecture Generation and Optimization in Time-Series Analysis 提出GenAutoML以解决时间序列分析中的动态架构生成问题 large language model
68 The Range Shrinks, the Threat Remains: Re-evaluating LLM Package Hallucinations on the 2026 Frontier-Model Cohort 评估前沿模型的包名幻觉问题以应对安全威胁 large language model
69 Learning the Context of Errors: Black-Box Online Adaptation of Time Series Foundation Models 提出ORCA以解决黑箱在线适应时间序列基础模型的问题 large language model foundation model ✅
70 Context-aware Modality-Topology Co-Alignment for Multimodal Attributed Graphs 提出CoMAG以解决多模态属性图的任务适应性问题 multimodal
71 Machine-learned particle flow as a foundation model for collider physics 提出机器学习粒子流模型以解决粒子碰撞分析问题 foundation model
72 A Comparative Study of Deep Learning Architectures for Multi-Horizon Behavioural Forecasting for Mobile Health 比较深度学习架构以提升移动健康行为预测精度 foundation model
73 Free Heavy-Tailed Lunch for Muon: A Theoretical Justification of Empirical Success 提出非欧几里得优化方法以解决重尾噪声问题 large language model
74 Code Correctness Signals in LLM Hidden States: Pre-Generation Probing and Repair Geometry 通过隐藏状态探测代码正确性并修复失败尝试 large language model
75 PepALD: Macrocyclic Peptide Generation via Autoregressive Latent Diffusion 提出PepALD以解决宏环肽生成中的多重设计挑战 foundation model
76 SemPiper: Interactive Code Synthesis for Semantic Operators in Machine Learning Pipelines 提出SemPiper以解决机器学习管道开发中的复杂性问题 large language model
77 Can Deep Neural Networks Improve Compression of Very Large Scientific Data? 提出深度学习模型以提升科学数据压缩效果 foundation model
78 When Language Representations Interact: Separability and Cross-Lingual Effects in LLMs 提出因果几何分析以理解多语言大模型的表示交互 large language model
79 LapidaryEngine: Fully Conversational Crystal Generation 提出LapidaryEngine以解决晶体生成的对话交互问题 large language model
80 Adaptive Nucleus Truncation for Long-Form Reasoning 提出自适应核截断方法以提升长文本推理性能 instruction following

🔬 支柱四：生成式动作 (Generative Motion) (4 篇)

#题目一句话要点标签🔗⭐
81 VideoMDM: Towards 3D Human Motion Generation From 2D Supervision 提出VideoMDM以解决3D人类动作生成问题 MDM motion generation human motion
82 Plan, Don't Pose: Long Composite Motion Generation with Text-Aligned BFM 提出Text2BFM以解决长文本到运动生成问题 text-to-motion motion generation motion representation
83 Using Seismic Statistical Features and VQ-VAE to Improve Spatiotemporal Seismicity Predictability 结合地震统计特征与VQ-VAE提升地震性预测能力 VQ-VAE spatiotemporal
84 Self-Guidance: Enhancing Neural Codecs via Decoder Manifold Alignment 提出自引导方法以提升神经音频编解码器性能 VQ-VAE

🔬 支柱一：机器人控制 (Robot Control) (4 篇)

#题目一句话要点标签🔗⭐
85 $μ$VLA: On Recurrent Memory for Partially Observable Manipulation in VLA Models 提出μVLA以解决部分可观测操作中的记忆问题 manipulation vision-language-action VLA
86 Distribution-Agnostic Robust Trajectory Optimization via Chance-Constrained Reinforcement Learning 提出基于机会约束强化学习的分布无关鲁棒轨迹优化框架 trajectory optimization reinforcement learning PULSE
87 DiffCoord: Differentiable Coordination for Distributed Multi-Agent Trajectory Optimization 提出DiffCoord以解决分布式多智能体轨迹优化问题 manipulation trajectory optimization
88 Strategic PAC Learnability via Geometric Definability 通过几何可定义性实现战略PAC可学习性 manipulation

🔬 支柱八：物理动画 (Physics-based Animation) (3 篇)

#题目一句话要点标签🔗⭐
89 TWLA: Achieving Ternary Weights and Low-Bit Activations for LLMs via Post-Training Quantization 提出TWLA框架以实现LLMs的三元权重和低比特激活 AMP large language model
90 Analog Quantum Asynchronous Event-Based Graph Neural Network 提出量子模拟异步事件图神经网络以处理事件相机数据 PULSE
91 Graphical Causal Reasoning for Root Cause Analysis in Cloud Networks 提出图形因果推理方法以解决云网络根本原因分析问题 spatiotemporal

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

#题目一句话要点标签🔗⭐
92 MaxProof: Scaling Mathematical Proof with Generative-Verifier RL and Population-Level Test-Time Scaling 提出MaxProof框架以提升数学证明的竞争能力 IMoS

🔬 支柱七：动作重定向 (Motion Retargeting) (1 篇)

#题目一句话要点标签🔗⭐
93 Decoupled Latent Optimization of Diffusion Models for Full Waveform Inversion 提出解耦潜变量优化方法以解决全波形反演问题 latent optimization

⬅️ 返回 cs.LG 首页 · 🏠 返回主页