cs.LG（2026-09-30）

📊 共 40 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) (23 🔗2) 支柱九：具身大模型 (Embodied Foundation Models) (13 🔗3) 支柱一：机器人控制 (Robot Control) (3) 支柱五：交互与反应 (Interaction & Reaction) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (23 篇)

#题目一句话要点标签🔗⭐
1 Mutual Equilibrium: Multimodal Representation Learning through Reciprocal Feedback 提出互反馈架构MEQ以提升多模态表示学习效果 representation learning multimodal visual grounding
2 DashVMC: Real-Time Discrete World Model Control in Geometry Dash 提出DashVMC以解决实时游戏控制中的动态建模问题 PPO world model world models
3 ReSAIL: Mitigating Collapse in Iterative Agent Self-Distillation 提出ReSAIL以解决迭代代理自蒸馏中的性能崩溃问题 distillation privileged information multimodal
4 Revisiting On-policy Adversarial Black-Box Distillation: Calibrating Groupwise Reward Geometry for Effective Advantage Construction 提出GRGC以解决现有黑箱蒸馏中的奖励几何问题 distillation large language model ✅
5 WinoTS: Wavelet-based Self-Distillation for Time Series Models 提出WinoTS以解决时间序列模型自监督预训练问题 distillation zero-shot transfer
6 GFD-OPD: Guidance-Folded On-Policy Distillation of Diffusion Models Across Scales 提出GFD-OPD以解决大规模扩散模型蒸馏中的分布差异问题 distillation classifier-free guidance
7 Reinforcement Learning-Guided Graph Transformations for SpTRSV Optimization 提出基于强化学习的图变换框架以优化稀疏三角求解 reinforcement learning curriculum learning
8 Robust Transfer Learning for Paper ECG Recognition 提出RobECG-CL以解决纸质心电图识别的鲁棒性问题 representation learning contrastive learning foundation model
9 No Task Vector Is an Island: A Comprehensive Study on the Composability of Task Vectors from On-Policy Distillation 提出任务向量组合方法以提升强化学习效果 reinforcement learning distillation
10 Semifactual Credit-Augmented Policy Optimization 提出半事实信用增强策略优化以改善推理能力 reinforcement learning large language model ✅
11 Distribution Matching Distillation for Continuous Diffusion Language Models 提出分布匹配蒸馏以降低连续扩散语言模型的生成成本 distillation
12 Gromov-Wasserstein Distillation for Inductive Multi-View Embedding 提出Gromov-Wasserstein蒸馏以解决多视图嵌入的归纳问题 distillation
13 Do Better Goal Representations Improve Goal-Conditioned Reinforcement Learning? 提出改进状态表示以提升目标条件强化学习性能 reinforcement learning
14 Free Everywhere, Exact on Trees: PPO's Dropped Correction Buys Sample Efficiency Under Aggressive Reuse 提出一种新方法提升PPO在树形结构上的样本效率 PPO
15 Correcting CondOT: Exact Finite-Step Sampling in Gaussian Flow Matching 提出CondOT调度以解决有限步采样中的误差问题 flow matching
16 From Search to Signal: Online Post-Training in Automatic Heuristic Design 提出在线后训练方法以优化自动启发式设计 reinforcement learning large language model
17 Wavelet Flow Matching for Time Series 提出小波流匹配方法以生成多变量时间序列 flow matching
18 Learning Beyond Full Imitation: Task-Preserving Knowledge Distillation 提出任务保留知识蒸馏以解决全模仿的局限性 distillation
19 T-Router: Learning Thalamic Routing for Reasoning with Parameter-Efficient Reinforcement Learning 提出T-Router以实现高效的推理强化学习 reinforcement learning
20 Switching Linear Attention 提出切换线性注意力以解决序列建模效率问题 linear attention
21 K2P: Label-Free Knowledge to Prompt Distillation 提出K2P以解决无标签知识蒸馏问题 distillation
22 From Imitation to Reward Discovery: On-Policy Warmup for Agentic RL 提出On-Policy Warmup以解决稀疏奖励下的强化学习问题 reinforcement learning distillation
23 Decoupled and Distilled: Task-Adaptive LoRA-Teachers with Ensemble Knowledge Transfer for Few-Shot Class-Incremental Learning 提出TALON框架以解决少样本类增量学习问题 representation learning distillation

🔬 支柱九：具身大模型 (Embodied Foundation Models) (13 篇)

#题目一句话要点标签🔗⭐
24 Ranking-Aware Prompt Optimization for Multimodal Clinical Diagnosis 提出Ranking-PE以优化多模态临床诊断中的提示策略 large language model multimodal
25 NodeGround: A Node Classification Benchmark in the Graph Foundation Model Era 提出NodeGround基准以评估图基础模型在节点分类中的表现 foundation model ✅
26 CIDER-FM: Foundation Models for Causal Inference from Diverse Experimental Regimes 提出CIDER-FM以提高因果推断的准确性 foundation model
27 Raw-Routed Mixture of Adapters: A Causal Intervention for Routing Collapse in Time Series Foundation Models 提出Raw-Routed Mixture of Adapters以解决时间序列基础模型中的路由崩溃问题 foundation model
28 When, Not How Much: Evaluating Time-Series Foundation Models on Sparse Events 提出时间序列基础模型评估方法以解决稀疏事件预测问题 foundation model
29 RATIO: Reasoning Analysis and Token-level Inference Optimization for Quantized Reasoning Models 提出RATIO框架以优化量化推理模型的推理性能 large language model chain-of-thought ✅
30 Is Weight Tying Still Beneficial for Decoder-Only LLMs in Private Settings Under DP-SGD? 探讨解码器仅模型在差分隐私下的权重绑定效益 large language model
31 OpenTSLM TeeMoE: A Unified Time-Series Language Model for Forecasting, Contextual Prediction, and Reasoning 提出OpenTSLM TeeMoE以解决时间序列预测与推理问题 foundation model
32 Predicting Multi-View Rashomon Representation: Can We Learn Where Models Disagree? 提出单视图预测模型不一致性的方法以提升输入可靠性 foundation model
33 A Comprehensive Benchmark of Source-Free Universal Domain Adaptation on Time Series Representations 提出源无关的通用领域适应方法以解决时间序列数据的挑战 foundation model
34 Network-based Spatial Context Retrieval for Open-weight LLMs: A Faithfulness Benchmark for Grounded Geographic Reasoning 提出基于网络的空间上下文检索以提升开放权重LLM的地理推理能力 large language model
35 PatchKV: Weight-Space Compensation of KV Cache 提出PatchKV以解决长上下文推理中的KV缓存瓶颈问题 large language model
36 Blackboard Intelligence Can Surpass Autoregressive on Globally Constrained Problems 提出黑板智能以解决复杂全局约束问题 large language model ✅

🔬 支柱一：机器人控制 (Robot Control) (3 篇)

#题目一句话要点标签🔗⭐
37 World-as-Graph: Relational World Modeling Through Latent Space Graphs 提出World-As-Graph以解决对象关系建模不足问题 manipulation world model world models
38 Role-Adaptive Policy Optimization for Offline Reinforcement Learning 提出角色自适应策略优化以解决离线强化学习中的价值估计不确定性问题 locomotion reinforcement learning TD3
39 Game-Guided Skill Discovery through Self-Play for Playable Agent Control 提出游戏引导的技能发现框架以解决可玩代理控制问题 Unitree policy learning

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

#题目一句话要点标签🔗⭐
40 Trust the Critic More 提出AC2方法以优化长序列的信用分配问题 IMoS

⬅️ 返回 cs.LG 首页 · 🏠 返回主页