cs.LG（2026-10-01）

📊 共 82 篇论文 | 🔗 7 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (42 🔗5) 支柱九：具身大模型 (Embodied Foundation Models) (31 🔗1) 支柱八：物理动画 (Physics-based Animation) (3 🔗1) 支柱一：机器人控制 (Robot Control) (3) 支柱五：交互与反应 (Interaction & Reaction) (2) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (42 篇)

#题目一句话要点标签🔗⭐
1 Distillation of Tabular Foundation Models into Efficient Predictors 提出知识蒸馏方法以提高表格基础模型的推理效率 distillation foundation model ✅
2 Beyond Linear Concepts: Discovering and Aligning Non-Linear Concept Manifolds in Large Language Models 提出非线性概念发现与对齐方法以提升大语言模型理解能力 DPO feature matching large language model
3 Reachability-Informed Reinforcement Learning for Multi-Impulse Interplanetary Transfers 提出基于可达性分析的强化学习方法以优化多脉冲星际转移 reinforcement learning reward shaping PULSE
4 In CEM, a World Model Is Also a Proposal Mechanism 提出CEM方法中的世界模型作为提议机制以优化决策过程 world model world models predictive model
5 Towards Fast and Disentangled Counterfactuals for Visual Foundation Models 提出Disentangled Diffusion Autoencoders解决视觉基础模型的反事实生成问题 distillation foundation model
6 The Missing Primitive: Diagnosing and Repairing Mathematical Reasoning in Large Language Models 提出数学原语以诊断和修复大语言模型的数学推理问题 distillation large language model
7 Higher-Order Molecular Grammars for Generative and Foundation Models in Chemistry 提出高阶语法表示以解决分子生成中的拓扑编码问题 representation learning foundation model
8 Same Reward, Different Skills: When Multimodal RL Learns to Look 提出视觉可解析性原则以提升多模态强化学习性能 reinforcement learning multimodal
9 Selection-Based Structured Reasoning: Toward Efficient Multimodal Search Agents 提出选择性结构推理以解决多模态搜索代理的效率问题 reinforcement learning multimodal ✅
10 Supervise What Decides Success: Criterion-Aligned Auxiliary Losses for Latent World-Model Planning 提出辅助损失以优化潜在世界模型的成功判定 world model world models
11 Learning Commute-Time-Preserving World Models for Planning 提出通勤时间保持的世界模型以优化规划问题 world model world models
12 CARM: Cancellation-Aware Response Masking for LLM Reinforcement Learning 提出CARM以解决LLM强化学习中的响应掩蔽问题 reinforcement learning large language model
13 Latent JEPA: Abstract Future Prediction for Latent Reasoning in Chemistry 提出Latent JEPA以解决化学推理中的未来预测问题 JEPA large language model
14 Flowing Faster to Coordinate: One-Step Online Multi-Agent Flow Policies 提出OMAF框架以解决在线多智能体协调学习问题 reinforcement learning policy learning multimodal
15 iADD: Improving Alignment and Diversity in Diffusion Policy Optimization 提出iADD以改善扩散策略优化中的对齐与多样性问题 reinforcement learning diffusion policy
16 The hidden advantage of mask resampling: a theory of masked autoencoders 提出掩码重采样理论以提升掩码自编码器性能 masked autoencoder MAE
17 Learning Rate Transfer for Hybrid Transformer-SSM Architectures 提出学习率转移方法以优化混合Transformer-SSM架构 Mamba SSM
18 From Gradients to Capabilities: Understanding Multi-Teacher On-Policy Distillation 提出多教师在线蒸馏方法以提升强化学习模型性能 distillation
19 Faynt: Scaling and Optimizing Policies for Competitive Melee 提出Faynt以优化竞争性近战游戏中的策略 reinforcement learning transformer policy distillation
20 Finetuning with Sampling: SFT Learns Better Than You Think 提出MCMC采样算法以提升SFT模型的泛化能力 reinforcement learning policy learning privileged information
21 Bellman Meets Lyapunov: Unsupervised Reinforcement Learning via Mastering Chaos 提出F-CIP以解决无监督强化学习中的奖励信号设计问题 reinforcement learning
22 Asynchronous LLM Post-Training: Group-Mass Capping and Convergence Analysis 提出GMC-GRPO以解决异步强化学习中的过时回滚问题 reinforcement learning large language model
23 Function-Structured Reinforcement Learning with Executable Verifiers for Mathematical Reasoning 提出功能结构化强化学习以解决数学推理问题 reinforcement learning ✅
24 Continual Reinforcement Learning with Neuroevolution 提出神经进化方法以解决强化学习中的适应与遗忘问题 reinforcement learning
25 Towards Optimal Policy Improvement 提出最优策略改进方法以解决近似评估下的强化学习问题 reinforcement learning SAC offline RL
26 Reusing Past Samples in Proximal Policy Optimization: When and How Does It Help? 提出样本重用策略以提升PPO的样本效率 reinforcement learning deep reinforcement learning PPO
27 Context-Aware Error Mitigation Orchestration for Hybrid Quantum Reinforcement Learning on NISQ Systems 提出自适应策略引导的错误缓解以提升量子强化学习性能 reinforcement learning
28 Counterfactual Generation via Flow Matching: Coupling-Sensitive End-to-End Rates 提出流匹配方法以解决反事实生成问题 flow matching
29 Does Scaling Reinforcement Learning Really Require More Training? 提出SURGE方法以优化强化学习训练资源利用 reinforcement learning
30 Iterative Policy Refinement through Semantic Rollout Analysis 提出闭环框架通过语义回滚分析迭代优化结构化策略 reinforcement learning imitation learning
31 Drive vs. Decay: On the Training Dynamics of Joint-Embedding Predictive Architectures 提出早期训练稳定性理论以解决JEPAs的表示崩溃问题 JEPA Joint-Embedding Predictive Architecture joint-embedding predictive architecture ✅
32 Slow-Fast Multi-Teacher On-Policy Distillation for Capability Preservation 提出Slow-Fast多教师在线蒸馏以解决能力干扰问题 distillation large language model multimodal
33 Energy Saving in 5G and Beyond Networks: A Quantum Reinforcement Learning Approach 提出量子强化学习方法以解决5G及未来网络的能耗问题 reinforcement learning deep reinforcement learning DRL
34 Test-time Multi-agent Coordination by Decomposed Value Gradient Flow 提出SCOUT框架以解决离线多智能体协调问题 reinforcement learning flow matching foundation model
35 Latent-MOPD: Latent Multi-Teacher On-Policy Distillation 提出Latent-MOPD以解决多教师蒸馏训练的局限性 teacher-student distillation
36 Lexicographic Multi-Objective On-Policy Distillation 提出LMOPD以解决多目标强化学习中的奖励优先级问题 reinforcement learning distillation
37 Sharpen Before You Adapt: Data-Free Entry-State Sharpening for Test-Time Reinforcement Learning 提出数据无关的入口状态锐化以提升测试时强化学习效果 reinforcement learning distillation
38 Reward Inflation: A Healthy Stimulus for Reinforcement Learning 提出奖励膨胀以促进强化学习的适应性 reinforcement learning
39 Student-Guided Teacher Distillation for Efficient LLM Task Routing: Positioning Against Jev-Style System-1 Classifiers 提出学生引导教师蒸馏方法以提高LLM任务路由效率 distillation
40 Flow Matching for Fast Posterior Sampling in Bayesian Inverse Problems 提出条件流匹配以加速贝叶斯逆问题的后验采样 flow matching
41 Continual Reinforcement Learning with Neuroevolution 提出神经进化方法以解决强化学习中的适应与遗忘问题 reinforcement learning
42 Structure-agnostic Causal Representation Learning 提出SaCRL框架以解决因果表示学习中的结构假设问题 representation learning ✅

🔬 支柱九：具身大模型 (Embodied Foundation Models) (31 篇)

#题目一句话要点标签🔗⭐
43 Wasserstein Gradient Flows and Forward-Only Diffusion Are Not Enough for Multimodal Sampling 揭示WGF和FODP在多模态采样中的局限性 multimodal
44 RelICL: Training-free Relational Learning with Tabular Foundation Models 提出RelICL以解决深度特征合成中的特征爆炸与交互盲点问题 foundation model
45 Parameter-Efficient Distributionally Robust Adaptation of Tabular Foundation Models under Subpopulation Shift 提出DR-TFM框架以应对子群体转变下的模型适应问题 foundation model
46 Have an LLM Write Your Anomaly Detector: Autonomous Discovery of Compact, Interpretable Detectors for Time Series 利用大型语言模型自主发现时间序列异常检测器 large language model foundation model
47 SIEVE: Selective attention-value Suppression for Vision-Language Models Unlearning 提出SIEVE以解决视觉语言模型的选择性遗忘问题 multimodal
48 TACO: Ternary Absolute-max Column-wise One-sparse Optimizer for LLM Fine-Tuning 提出TACO优化器以解决大语言模型微调中的内存瓶颈问题 large language model
49 Foundations without Fundamentals: Zero-Shot Blind Spots in Time Series FMs 提出SimpleTimeBench以解决时间序列基础模型的盲点问题 foundation model
50 FastCI: Efficient GPU-Intensive CI for LLM Training Frameworks 提出FastCI以解决LLM训练框架中的CI效率问题 large language model
51 Learning to Predict Distributions over Weight Updates for Test-Time Adaptation 提出分布式超网络以优化测试时适应性问题 large language model
52 QK-Wanda: Coupling Queries and Keys for Unstructured Pruning 提出QK-Wanda以优化无结构剪枝方法 large language model
53 Range-GRPO: Policy Optimization via Pairwise Relations among Reward Intervals 提出Range-GRPO以解决奖励不确定性问题 large language model
54 FedFit: Federated Fine-Tuning of LLMs via Vector-Bank Parameterization and Quantization 提出FedFit以解决联邦学习中的通信开销问题 large language model
55 Mixture-Trained Merging for Unified Multi-Objective Models 提出混合训练合并方法以解决多目标模型的兼容性问题 instruction following
56 STEER: Reducing Inference Cost in Relational Foundation Models through Semantically Informed Sampling 提出STEER以降低关系基础模型的推理成本 large language model foundation model
57 Fisher-Guided Submodular Data Selection for Continual Pre-Training of Large Language Models 提出Fisher引导的子模数据选择以解决持续预训练问题 large language model
58 LiteEMG-FM: An Efficient and Deployable Foundation Model for Robust EMG Sensing 提出LiteEMG-FM以解决EMG信号泛化不足问题 foundation model
59 VisAudit: Evaluating Multimodal Agents for Visual Diagnosis and Repair 提出VisAudit以解决多模态代理在可视化审查中的不足问题 multimodal
60 From Behavior to Provenance: Attributing Tabular Foundation Models to Synthetic Pretraining Data 提出合成预训练数据的可追溯性方法以解决模型行为归因问题 foundation model
61 Joint Movement and Compression Ratio Design for Mobile Embodied AI Networks (MEAN) 提出联合移动与压缩比设计以优化移动智能体网络能效 embodied AI
62 AI-driven Thermal-aware Data Center Capacity Planning 提出AI驱动的热管理数据中心容量规划框架以应对挑战 large language model
63 A Generative Model of Complex Networks Using Graphons and Neural Inverse Operators 提出基于图形和神经逆算子的生成模型以解决复杂网络建模问题 foundation model
64 Are you Synthesizing or Recalling? Evaluating LLMs on Algorithmic Code Retrieval 提出AlgoREval基准以评估LLMs在算法代码检索中的表现 large language model ✅
65 Inherit-MAS: Test-Time Evolution of Multi-Agent Systems through Workflow and Execution Inheritance 提出Inherit-MAS以优化多智能体系统的工作流与执行过程 large language model
66 Hesitation Has a Geometry: Entropy-Trained Hyperbolic Probes for Sparse Activation Steering 提出超曲率熵引导以优化稀疏激活问题 large language model
67 Why Does Adaptive Batching Help LLM Pretraining? A Perspective from Unbounded Variance 提出自适应批处理调度器以优化LLM预训练中的方差控制 large language model
68 Does Every User Need a Private LoRA? Decoupling Personalization from Per-User Adaptation 提出LINEUP以解决个性化大语言模型的用户适应性问题 large language model
69 Intent-Hiding Jailbreaks: An Information-Theoretic Framework for Compositional Attacks 提出信息论框架以解决语言模型的意图隐藏攻击问题 large language model
70 Range-GRPO: Policy Optimization via Pairwise Relations among Reward Intervals 提出Range-GRPO以解决奖励不确定性问题 large language model
71 SLIM: Simplex-Lattice Interpolation Merging 提出SLIM以优化大语言模型的合并系数评估 large language model
72 Generalist Representation, Specialist Detection: TS-Router for Time-Series Anomaly Detection 提出TS-Router以解决时间序列异常检测的泛化问题 foundation model
73 Match the Distribution, Not the Compute: Post-Training Multi-Token Prediction Heads 提出后训练多标记预测头以提升推理效率 chain-of-thought

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

#题目一句话要点标签🔗⭐
74 Physics-Refined Spatiotemporal Forecasting on Open-Boundary Hydrologic Graphs 提出物理精炼的时空预测方法以解决开放边界水文图的不稳定性问题 spatiotemporal
75 Sim+Real: Joint Simulation - Experiment Training Improves Balanced Prediction in Physical Systems 提出联合训练方法以提升物理系统的预测平衡性 spatiotemporal ✅
76 Effects of interpulse-interval variation on deep-learning classification of bat vocalizations 研究间隔脉冲变化对蝙蝠声学分类的影响 PULSE

🔬 支柱一：机器人控制 (Robot Control) (3 篇)

#题目一句话要点标签🔗⭐
77 Variational Streaming Flow: Probabilistic Forecasting in Physical Time 提出变分流动模型以解决复杂动态系统的概率预测问题 manipulation motion planning flow matching
78 Multi-Fidelity Policy Gradients Stabilize Data-Scarce Reinforcement Learning 提出多保真度策略梯度方法以解决数据稀缺的强化学习问题 locomotion reinforcement learning PPO
79 Co-design Gym: A Unified Benchmark for Embodiment-Policy Co-optimization 提出Co-Design Gym以解决共同优化行为策略与体现设计的问题 locomotion manipulation policy learning

🔬 支柱五：交互与反应 (Interaction & Reaction) (2 篇)

#题目一句话要点标签🔗⭐
80 SkillEvoLean: Mutation-enhanced skill evolution for Lean provers 提出Mutation增强的技能自演化框架以提升Lean证明器 IMoS large language model
81 Clifford Sheaf Neural Networks 提出Clifford Sheaf神经网络以解决几何图的回归问题 OMOMO

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
82 SoftServe: A Scalable Quasi-Newton Method for Deep Learning 提出SoftServe以解决深度学习中的优化问题 physics-informed diffusion

⬅️ 返回 cs.LG 首页 · 🏠 返回主页