cs.LG（2026-10-14）

📊 共 41 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) (20) 支柱九：具身大模型 (Embodied Foundation Models) (16) 支柱一：机器人控制 (Robot Control) (2 🔗1) 支柱八：物理动画 (Physics-based Animation) (1) 支柱七：动作重定向 (Motion Retargeting) (1) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (20 篇)

#题目一句话要点标签🔗⭐
1 Enhancing Latent World Models with Associative Memory 提出记忆增强的潜在世界模型以解决长远规划问题 world model world models JEPA
2 Diagnosing and Correcting Reward Hacking in Verifiable Instruction Following 提出CALIF以解决可验证指令跟随中的奖励黑客问题 reinforcement learning instruction following
3 CARD: Closing the Teacher-Student Loop for Adversarial Robustness Distillation 提出CARD框架以解决对抗鲁棒性蒸馏中的反馈问题 teacher-student distillation
4 TEX-MEX: Similarity-Guided Temporal Contrastive Learning for Time Series 提出TEX-MEX以解决时间序列对比学习中的相似性问题 JEPA contrastive learning foundation model
5 When Does Momentum Help in Reinforcement Learning with Verifiable Rewards? 提出动量方法以提升可验证奖励的强化学习效果 reinforcement learning large language model
6 Adversarial Reinforcement Learning for Robust Planetary Landing 提出对抗性强化学习以增强行星着陆的鲁棒性 reinforcement learning deep reinforcement learning DRL
7 Recovering Reliable Supervision via Critique-Guided On-Policy Distillation 提出CriticOPD以解决教师模型监督可靠性问题 distillation privileged information
8 Harness Policy Distillation: Aligning Agent Policies with Runtime Semantics 提出Harness Policy Distillation以解决运行时语义对齐问题 distillation privileged information
9 RM-OPD: On-Policy Distillation for Generative Reward Models 提出RM-OPD以解决混合领域奖励模型的稀疏性与干扰问题 reinforcement learning distillation
10 Occupancy Modeling via Drifting: One-Step Generation Enables Direct Policy Extraction 提出一种新方法通过漂移建模占用率以直接提取策略 reinforcement learning distillation foundation model
11 Function-Space Temporal Self-Distillation for Policy Optimization 提出功能空间时间自蒸馏以优化策略学习 reinforcement learning distillation
12 What Does Disagreement Mean? A Semantic Identifiability Frontier for Pairwise Preference Learning 提出非参数潜在语义模型以解决配对偏好学习中的不一致性问题 preference learning
13 FlowSep-RF: Rectified Flow Matching for Single-Channel Radio-Frequency Signal Separation 提出FlowSep-RF以解决单通道射频信号分离问题 flow matching
14 In-Context Batch Black-box Optimization via Flow Matching 提出Flow Matching以解决批量黑箱优化问题 flow matching
15 Weighted Adjoint Matching: Finetuning Flow Matching Models with Time-Dependent Control Cost 提出加权伴随匹配以优化流匹配模型的微调 flow matching
16 Provable Transformer Context Compression for In-Context Reinforcement Learning 提出可证明的Transformer上下文压缩以解决强化学习中的内存问题 reinforcement learning
17 Reinforcement Learning via Advantage Guidance 提出优势引导方法以解决强化学习不稳定性问题 reinforcement learning
18 Bridging the Topological Divide: Dual-Branch Graph Distillation with Semantic Priors for Robust Cold-Start Recommendation 提出DualGCN以解决冷启动推荐中的拓扑隔离问题 distillation
19 Test-Time Forecasting: Recurrent Depth Enables In-Context Dynamics Prediction 提出Test-Time Forecasting框架以解决动态系统预测问题 linear attention foundation model
20 MAEM: Mitigating Recency Bias via a Replay-Free, Capacity-Controlled Dynamic Mixture of Experts for Class-Incremental Graph Learning 提出MAEM框架以解决图学习中的近期偏见问题 masked autoencoder MAE

🔬 支柱九：具身大模型 (Embodied Foundation Models) (16 篇)

#题目一句话要点标签🔗⭐
21 On the Limits of Representation-Based Test-Time Adaptation for Multimodal LLMs 探讨多模态大语言模型的测试时适应性局限性 large language model multimodal
22 What Do Foundation Models for Market Microstructure Learn? A Controlled Study with Synthetic Limit Order Books 提出基于合成限价单簿的模型以优化市场微观结构预测 foundation model zero-shot transfer
23 Learning by Examples: Power Grid Steady State Analysis with a Tabular Foundation Model 提出基于预训练表格基础模型的电网稳态分析方法 foundation model
24 OrionMoE-v1.0: A Mixture-of-Experts Foundation Model for Tabular In-Context Learning 提出OrionMoE-Small以解决表格数据学习中的稀疏计算问题 foundation model
25 ChemArgus: Automated Scientific Diagnostics for Lab-Grade Multimodal Chemistry Reasoning: A Frontier Benchmark 提出ChemArgus以解决化学推理评估中的盲点问题 multimodal
26 When Does Depth Help? Understanding Layerwise Inference in Tabular Foundation Models 提出深度分析以优化表格基础模型的推理效率 foundation model
27 TIMERSI: Towards Self-Improved Agentic Deployment for Time-Series Foundation Models 提出TIMERSI以解决时间序列模型部署中的任务特定设计问题 foundation model
28 Text-Informed Training Data Filtering for Enhanced Virtual Flow Meter Performance 提出文本信息驱动的数据过滤方法以提升虚拟流量计性能 large language model multimodal
29 Bayesian Learning to Select Fair Classifiers 提出贝叶斯学习方法以选择公平分类器 large language model
30 SpectrumQuant: Spectrum Alignment for LLM Quantization 提出SpectrumQuant以解决大语言模型量化问题 large language model
31 Safe-FCL: Continual Risk Awareness and Fine-Grained Safety Aggregation for Federated Continual Learning of LLMs 提出Safe-FCL以解决联邦持续学习中的风险检测问题 large language model
32 When Non-Adversarial Noise Hurts Learning: Benchmarking Skill Self-Evolution in Agents 提出RSE-Bench以评估非对抗噪声下的技能自我演化 large language model
33 Energy Consumption Analysis of Discrete Diffusion and Autoregressive Language Models 比较扩散语言模型与自回归模型的能耗效率 large language model
34 Sparse Autoencoders Are a Shared Coordinate System for Symbolic Matching 提出SAE-IDF以解决符号匹配的可解释性问题 large language model
35 A Graph-Based Analysis of Comparability, Scale, and Attribution in Private Transformer Inference 构建图模型以系统化私有Transformer推理的评估与比较 large language model
36 PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference 提出PN-Mix以解决长输出LLM推理中的KV缓存量化问题 large language model

🔬 支柱一：机器人控制 (Robot Control) (2 篇)

#题目一句话要点标签🔗⭐
37 Hyperspherical Contrastive Representations Unlock Planning and Exploration 提出C-Slerp方法以解决自监督强化学习中的规划与探索问题 manipulation reinforcement learning
38 Outperforming 1000 Euclidean Layers With Hyperbolic Goal-Conditioned RL 通过超曲率目标条件强化学习超越1000层欧几里得网络 humanoid reinforcement learning ✅

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

#题目一句话要点标签🔗⭐
39 NeuroDETR: Multiscale Discrete Spatiotemporal Tokenization for Asynchronous EEG Event Detection 提出NeuroDETR以解决异步脑电图事件检测问题 spatiotemporal

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

#题目一句话要点标签🔗⭐
40 When Does Task-Aware Imputation Matter? An Analysis of Robustness Under Train-Test Mismatch 提出任务感知插补方法以应对缺失数据问题 latent optimization

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
41 Rethinking Prediction Horizons: Decoupling Model Queries from State Updates in Few-Step MeanFlow 提出MeanFlow模型以解耦预测视野与状态更新问题 classifier-free guidance

⬅️ 返回 cs.LG 首页 · 🏠 返回主页