cs.LG（2025-04-14）

📊 共 25 篇论文 | 🔗 1 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (11 🔗1) 支柱二：RL算法与架构 (RL & Architecture) (10) 支柱四：生成式动作 (Generative Motion) (1) 支柱七：动作重定向 (Motion Retargeting) (1) 支柱八：物理动画 (Physics-based Animation) (1) 支柱一：机器人控制 (Robot Control) (1)

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

#题目一句话要点标签🔗⭐
1 Enhancing Ultra-Low-Bit Quantization of Large Language Models Through Saliency-Aware Partial Retraining 提出基于显著性感知的局部重训练方法，提升大语言模型超低比特量化性能。 large language model ✅
2 Foundation models for electronic health records: representation dynamics and transferability 研究电子病历基础模型的表征动态与迁移能力，提升临床预测任务性能。 foundation model
3 Satellite Federated Fine-Tuning for Foundation Models in Space Computing Power Networks 提出卫星联邦微调框架，解决星载计算资源受限和空间通信挑战。 foundation model
4 How Instruction and Reasoning Data shape Post-Training: Data Quality through the Lens of Layer-wise Gradients 通过层级梯度谱分析，揭示指令和推理数据质量对LLM后训练的影响 large language model instruction following
5 LLM-based AI Agent for Sizing of Analog and Mixed Signal Circuit 提出基于LLM的AI Agent，用于模拟和混合信号电路的晶体管尺寸设计 large language model
6 Can LLMs Handle WebShell Detection? Overcoming Detection Challenges with Behavioral Function-Aware Framework 提出BFAD框架，提升LLM在WebShell检测中的性能，超越传统方法。 large language model
7 Efficient Process Reward Model Training via Active Learning 提出ActPRM主动学习方法，高效训练过程奖励模型，降低标注成本。 large language model
8 Graph Neural Networks Based Analog Circuit Link Prediction 提出GNN-ACLP方法，利用图神经网络进行模拟电路链路预测，提升电路设计自动化水平。 large language model
9 FedRecon: Missing Modality Reconstruction in Heterogeneous Distributed Environments FedRecon：异构分布式环境下缺失模态重建的联邦学习方法 multimodal
10 KeepKV: Achieving Periodic Lossless KV Cache Compression for Efficient LLM Inference KeepKV：实现LLM高效推理的周期性无损KV缓存压缩 large language model
11 Ember: A Compiler for Efficient Embedding Operations on Decoupled Access-Execute Architectures Ember编译器：为解耦访问-执行架构优化嵌入操作 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (10 篇)

#题目一句话要点标签🔗⭐
12 On the Value of Cross-Modal Misalignment in Multimodal Representation Learning 通过建模跨模态不对齐，提升多模态表征学习的性能与可解释性 representation learning contrastive learning multimodal
13 M1: Towards Scalable Test-Time Compute with Mamba Reasoning Models 提出基于Mamba的混合线性RNN推理模型M1，提升测试时计算效率。 Mamba distillation large language model
14 Achieving Optimal Tissue Repair Through MARL with Reward Shaping and Curriculum Learning 提出基于MARL的组织修复框架，通过奖励塑造和课程学习优化修复过程 reinforcement learning curriculum learning reward shaping
15 Adaptive Sensor Steering Strategy Using Deep Reinforcement Learning for Dynamic Data Acquisition in Digital Twins 提出基于深度强化学习的自适应传感器控制策略，用于数字孪生中的动态数据采集。 reinforcement learning deep reinforcement learning
16 STaRFormer: Semi-Supervised Task-Informed Representation Learning via Dynamic Attention-Based Regional Masking for Sequential Data STaRFormer：基于动态注意力区域掩码的半监督任务感知序列数据表征学习 representation learning contrastive learning spatiotemporal
17 Reasoning without Regret 提出BARS框架以解决稀疏奖励信号的有效性问题 reward shaping large language model chain-of-thought
18 Using Reinforcement Learning to Integrate Subjective Wellbeing into Climate Adaptation Decision Making 提出强化学习框架以整合主观幸福感于气候适应决策中 reinforcement learning
19 AimTS: Augmented Series and Image Contrastive Learning for Time Series Classification AimTS：通过增强序列和图像对比学习提升时间序列分类性能 contrastive learning
20 Improving Controller Generalization with Dimensionless Markov Decision Processes 提出基于无量纲MDP的强化学习方法，提升控制器在不同环境下的泛化能力 reinforcement learning world model
21 Moderate Actor-Critic Methods: Controlling Overestimation Bias via Expectile Loss 提出基于期望分位损失的适度Actor-Critic方法，抑制Q函数过估计偏差 reinforcement learning SAC

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
22 RadarLLM: Empowering Large Language Models to Understand Human Motion from Millimeter-Wave Point Cloud Sequence RadarLLM：利用大语言模型理解毫米波雷达点云序列中的人体运动 VQ-VAE human motion large language model

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

#题目一句话要点标签🔗⭐
23 A Structure-Preserving Framework for Solving Parabolic Partial Differential Equations with Neural Networks 提出Sidecar框架，增强神经网络求解抛物型偏微分方程的物理一致性 structure preservation

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

#题目一句话要点标签🔗⭐
24 Air Quality Prediction with A Meteorology-Guided Modality-Decoupled Spatio-Temporal Network 提出MDSTNet，利用气象引导的解耦时空网络进行空气质量预测 spatiotemporal

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
25 Undermining Federated Learning Accuracy in EdgeIoT via Variational Graph Auto-Encoders 提出数据独立模型操控攻击以解决EdgeIoT中的联邦学习准确性问题 manipulation

⬅️ 返回 cs.LG 首页 · 🏠 返回主页