cs.LG（2026-02-19）

📊 共 27 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) (13) 支柱九：具身大模型 (Embodied Foundation Models) (11 🔗1) 支柱三：空间感知与语义 (Perception & Semantics) (1) 支柱一：机器人控制 (Robot Control) (1) 支柱八：物理动画 (Physics-based Animation) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (13 篇)

#题目一句话要点标签🔗⭐
1 Canonicalizing Multimodal Contrastive Representation Learning 提出正交映射以实现多模态对比表示学习的统一性 representation learning multimodal
2 Spatio-temporal dual-stage hypergraph MARL for human-centric multimodal corridor traffic signal control 提出STDSH-MARL以解决多模态交通信号控制问题 reinforcement learning deep reinforcement learning multimodal
3 SMAC: Score-Matched Actor-Critics for Robust Offline-to-Online Transfer SMAC：通过分数匹配的Actor-Critic方法实现鲁棒的离线到在线迁移 reinforcement learning TD3 offline RL
4 2Mamba2Furious: Linear in Complexity, Competitive in Accuracy 提出2Mamba，通过简化和改进Mamba-2，在长文本建模中实现精度与效率的平衡。 Mamba linear attention
5 LexiSafe: Offline Safe Reinforcement Learning with Lexicographic Safety-Reward Hierarchy 提出LexiSafe框架以解决离线安全强化学习中的安全问题 reinforcement learning offline RL
6 Optimal Unconstrained Self-Distillation in Ridge Regression: Strict Improvements, Precise Asymptotics, and One-Shot Tuning 提出最优无约束自蒸馏方法以提升岭回归性能 distillation
7 A Theoretical Framework for Modular Learning of Robust Generative Models 提出模块化生成模型训练框架，提升LLM在混合数据上的鲁棒性与效率 distillation large language model
8 MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning MASPO：统一梯度利用、概率质量和信号可靠性的LLM鲁棒推理与高效采样 reinforcement learning large language model
9 RLGT: A reinforcement learning framework for extremal graph theory 提出RLGT框架，系统化图论极值问题，提升强化学习求解效率。 reinforcement learning
10 TIFO: Time-Invariant Frequency Operator for Stationarity-Aware Representation Learning in Time Series 提出时不变频率算子TIFO，解决非平稳时间序列预测中的分布偏移问题。 representation learning
11 Action-Graph Policies: Learning Action Co-dependencies in Multi-Agent Reinforcement Learning 提出行动图策略以解决多智能体强化学习中的协调问题 reinforcement learning
12 VP-VAE: Rethinking Vector Quantization via Adaptive Vector Perturbation VP-VAE：通过自适应向量扰动改进向量量化变分自编码器 representation learning VQ-VAE
13 MARS: Margin-Aware Reward-Modeling with Self-Refinement 提出MARS以解决奖励模型训练中的不确定性问题 PPO RLHF

🔬 支柱九：具身大模型 (Embodied Foundation Models) (11 篇)

#题目一句话要点标签🔗⭐
14 Reverso: Efficient Time Series Foundation Models for Zero-shot Forecasting 提出Reverso，一种高效时间序列基础模型，用于零样本预测。 foundation model
15 Retrospective In-Context Learning for Temporal Credit Assignment with Large Language Models 提出基于大语言模型的回顾性上下文学习，解决强化学习中的时序信用分配问题 large language model
16 Structured Prototype-Guided Adaptation for EEG Foundation Models 提出SCOPE框架，通过结构化原型引导自适应脑电基础模型，提升少样本跨被试泛化能力。 foundation model
17 TimeOmni-VL: Unified Models for Time Series Understanding and Generation 提出TimeOmni-VL以解决时间序列理解与生成的分裂问题 multimodal chain-of-thought
18 Pushing the Frontier of Black-Box LVLM Attacks via Fine-Grained Detail Targeting M-Attack-V2：通过细粒度细节攻击提升黑盒LVLM对抗攻击性能 multimodal ✅
19 Towards Anytime-Valid Statistical Watermarking 提出基于E-value的统计水印框架，实现LLM生成内容的可信溯源与高效检测。 large language model
20 Privacy-Preserving Mechanisms Enable Cheap Verifiable Inference of LLMs 利用隐私保护机制实现低成本、可验证的大语言模型推理 large language model
21 Powering Up Zeroth-Order Training via Subspace Gradient Orthogonalization ZO-Muon：基于子空间梯度正交化的零阶优化方法，提升大模型微调效率 large language model
22 FLoRG: Federated Fine-tuning with Low-rank Gram Matrices and Procrustes Alignment FLoRG：基于低秩Gram矩阵和Procrustes对齐的联邦微调方法，解决分解漂移问题。 large language model
23 MeGU: Machine-Guided Unlearning with Target Feature Disentanglement 提出MeGU，通过目标特征解耦实现机器引导的有效率的机器学习遗忘 large language model
24 Adam Improves Muon: Adaptive Moment Estimation with Orthogonalized Momentum 提出NAMO与NAMO-D优化器，将正交动量与自适应矩估计相结合，提升大语言模型训练效果。 large language model

🔬 支柱三：空间感知与语义 (Perception & Semantics) (1 篇)

#题目一句话要点标签🔗⭐
25 i-PhysGaussian: Implicit Physical Simulation for 3D Gaussian Splatting 提出i-PhysGaussian，将3D高斯溅射与隐式MPM积分器结合，实现更稳定的物理仿真。 3D gaussian splatting 3DGS gaussian splatting

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
26 Continual uncertainty learning 提出基于课程学习的持续不确定性学习框架，用于解决多重不确定性下的机械系统鲁棒控制问题。 sim-to-real domain randomization reinforcement learning

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

#题目一句话要点标签🔗⭐
27 Learning a Latent Pulse Shape Interface for Photoinjector Laser Systems 提出基于Wasserstein自编码器的光阴极激光系统脉冲整形潜空间学习方法 PULSE

⬅️ 返回 cs.LG 首页 · 🏠 返回主页