cs.LG（2025-03-25）

📊 共 14 篇论文 | 🔗 3 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (8 🔗3) 支柱二：RL算法与架构 (RL & Architecture) (5) 支柱八：物理动画 (Physics-based Animation) (1)

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

#题目一句话要点标签🔗⭐
1 VectorFit : Adaptive Singular & Bias Vector Fine-Tuning of Pre-trained Foundation Models VectorFit：通过自适应奇异向量与偏置向量微调预训练模型，提升参数效率。 foundation model
2 Data-centric Federated Graph Learning with Large Language Models 提出LLM4FGL框架，利用大语言模型解决联邦图学习中的数据异构性问题。 large language model
3 LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration LayerCraft：利用CoT推理和分层对象集成增强文本到图像生成 large language model chain-of-thought ✅
4 Why Representation Engineering Works: A Theoretical and Empirical Study in Vision-Language Models 针对视觉-语言模型，提出基于主特征向量的表征工程理论框架，提升模型透明性。 large language model multimodal
5 LogQuant: Log-Distributed 2-Bit Quantization of KV Cache with Superior Accuracy Preservation LogQuant：基于对数分布的2比特KV缓存量化，显著提升大模型推理精度 large language model ✅
6 Automated Video-EEG Analysis in Epilepsy Studies: Advances and Challenges 综述癫痫视频脑电分析进展与挑战，提出基于概念学习的治疗效果评估新流程 multimodal
7 Optimization through In-Context Learning and Iterative LLM Prompting for Nuclear Engineering Design Problems 提出基于上下文学习和迭代LLM提示的优化方法，解决核工程设计问题 large language model
8 QUAD: Quantization and Parameter-Efficient Tuning of LLM with Activation Decomposition 提出QUAD框架以解决大语言模型量化中的激活异常问题 large language model ✅

🔬 支柱二：RL算法与架构 (RL & Architecture) (5 篇)

#题目一句话要点标签🔗⭐
9 ExCoT: Optimizing Reasoning for Text-to-SQL with Execution Feedback ExCoT：利用执行反馈优化Text-to-SQL的推理能力 DPO direct preference optimization large language model
10 Beyond Verifiable Rewards: Scaling Reinforcement Learning for Language Models to Unverifiable Data 提出JEPO算法，扩展强化学习至不可验证数据的语言模型训练 reinforcement learning chain-of-thought
11 LERO: LLM-driven Evolutionary framework with Hybrid Rewards and Enhanced Observation for Multi-Agent Reinforcement Learning LERO：基于LLM驱动的演化框架，通过混合奖励和增强观测提升多智能体强化学习性能 reinforcement learning large language model
12 Abstracting Geo-specific Terrains to Scale Up Reinforcement Learning 提出基于抽象地形的多智能体强化学习方法，加速军事仿真训练。 reinforcement learning
13 Flow to Learn: Flow Matching on Neural Network Parameters 提出FLoWN，通过流匹配学习生成神经网络参数，提升图像任务的元学习能力。 flow matching

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

#题目一句话要点标签🔗⭐
14 Tensor Generalized Approximate Message Passing 提出张量广义近似消息传递算法（TeG-AMP）用于低秩张量推断，解决张量补全和分解问题。 AMP

⬅️ 返回 cs.LG 首页 · 🏠 返回主页