cs.LG（2025-02-14）

📊 共 22 篇论文 | 🔗 2 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (11 🔗2) 支柱九：具身大模型 (Embodied Foundation Models) (8) 支柱一：机器人控制 (Robot Control) (2) 支柱八：物理动画 (Physics-based Animation) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (11 篇)

#题目一句话要点标签🔗⭐
1 AdaPTS: Adapting Univariate Foundation Models to Probabilistic Multivariate Time Series Forecasting AdaPTS：通过适配器将单变量预训练模型应用于概率多变量时间序列预测 representation learning foundation model ✅
2 AttenGluco: Multimodal Transformer-Based Blood Glucose Forecasting on AI-READI Dataset AttenGluco：基于多模态Transformer的血糖预测框架，有效融合CGM和活动数据。 MAE multimodal
3 Ten Challenging Problems in Federated Foundation Models 综述联邦大模型中十大挑战性问题，为理论研究和实际应用提供指导 teacher-student foundation model
4 Preference learning made easy: Everything should be understood through win rate 基于胜率的偏好学习框架，简化生成模型对齐偏好数据的过程 preference learning RLHF DPO
5 From Markov to Laplace: How Mamba In-Context Learns Markov Chains Mamba在上下文学习中高效学习马尔可夫链的拉普拉斯平滑估计器 Mamba SSM
6 Representation Learning on Out of Distribution in Tabular Data 提出TCL，一种轻量级表格数据OOD表示学习方法，提升分类性能。 representation learning contrastive learning
7 Direct Preference Optimization-Enhanced Multi-Guided Diffusion Model for Traffic Scenario Generation 提出DPO增强的多引导扩散模型，用于生成逼真可控的交通场景 DPO direct preference optimization
8 Do We Need to Verify Step by Step? Rethinking Process Supervision from a Theoretical Perspective 从理论视角重新审视过程监督：结果监督与过程监督统计难度相当 reinforcement learning large language model
9 Dynamic Reinforcement Learning for Actors Dynamic RL：通过直接控制系统动力学实现强化学习的动态化 reinforcement learning
10 Filtration-Based Representation Learning for Temporal Graphs 提出基于δ-时间模体的时序图过滤表示学习方法，用于时序图分类。 representation learning
11 Process Reward Models for LLM Agents: Practical Framework and Directions 提出AgentPRM框架，通过过程奖励模型提升LLM Agent交互能力 RLHF reward shaping ✅

🔬 支柱九：具身大模型 (Embodied Foundation Models) (8 篇)

#题目一句话要点标签🔗⭐
12 SPIRIT: Short-term Prediction of solar IRradIance for zero-shot Transfer learning using Foundation Models SPIRIT：利用基础模型进行太阳辐照度短期预测，实现零样本迁移学习 foundation model zero-shot transfer
13 Accelerating Unbiased LLM Evaluation via Synthetic Feedback 提出一种结合人类与合成反馈的无偏LLM评估加速方法。 large language model
14 MixMin: Finding Data Mixtures via Convex Minimization MixMin：通过凸优化寻找最优数据混合比例，提升模型性能。 large language model
15 Exploring the Boundary of Diffusion-based Methods for Solving Constrained Optimization 提出DiOpt：一种基于扩散模型的约束优化求解框架 multimodal
16 Accelerated Parallel Tempering via Neural Transports 利用神经传输加速并行退火算法，提升复杂分布采样效率 multimodal
17 Efficient Zero-Order Federated Finetuning of Language Models for Resource-Constrained Devices 提出高效零阶联邦微调方法，加速资源受限设备上LLM的微调收敛。 large language model
18 A novel approach to data generation in generative model 提出收敛融合范式（CFP）理论，解决生成模型中数据生成结构的捕获问题。 large language model
19 Identifiable Steering via Sparse Autoencoding of Multi-Concept Shifts 提出稀疏移位自编码器，实现多概念变化下的可辨识语言模型操控 large language model

🔬 支柱一：机器人控制 (Robot Control) (2 篇)

#题目一句话要点标签🔗⭐
20 Memory, Benchmark & Robots: A Benchmark for Solving Complex Tasks with Reinforcement Learning 提出MIKASA：用于评估强化学习智能体记忆能力的机器人操作基准 manipulation reinforcement learning
21 Shaping Inductive Bias in Diffusion Models through Frequency-Based Noise Control 通过频率控制噪声塑造扩散模型的归纳偏置 manipulation

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

#题目一句话要点标签🔗⭐
22 Robust Anomaly Detection via Tensor Pseudoskeleton Decomposition 提出基于张量伪骨架分解的鲁棒异常检测方法，适用于高维张量数据。 spatiotemporal

⬅️ 返回 cs.LG 首页 · 🏠 返回主页