cs.LG（2025-10-19）

📊 共 7 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) (4 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (3 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (4 篇)

#题目一句话要点标签🔗⭐
1 Hierarchical Federated Unlearning for Large Language Models 提出层级联邦卸载学习方法，解决LLM异构数据下的持续卸载难题。 large language model
2 Peering Inside the Black Box: Uncovering LLM Errors in Optimization Modelling through Component-Level Evaluation 提出组件级评估框架，诊断LLM在优化建模中的错误，提升模型性能。 large language model chain-of-thought
3 Forgetting to Forget: Attention Sink as A Gateway for Backdooring LLM Unlearning 提出基于注意力汇聚的后门LLM卸载方法，实现可控的知识遗忘与恢复。 large language model ✅
4 Utility-Diversity Aware Online Batch Selection for LLM Supervised Fine-tuning 提出UDS框架，通过效用-多样性感知在线批量选择优化LLM监督微调。 large language model ✅

🔬 支柱二：RL算法与架构 (RL & Architecture) (3 篇)

#题目一句话要点标签🔗⭐
5 Leave It to the Experts: Detecting Knowledge Distillation via MoE Expert Signatures 利用MoE专家签名检测知识蒸馏，有效应对提示工程攻击。 distillation large language model
6 NeuCo-Bench: A Novel Benchmark Framework for Neural Embeddings in Earth Observation NeuCo-Bench：面向地球观测的神经嵌入评估基准框架，解决表征学习的标准化评估问题 representation learning foundation model
7 Fly-CL: A Fly-Inspired Framework for Enhancing Efficient Decorrelation and Reduced Training Time in Pre-trained Model-based Continual Representation Learning Fly-CL：受果蝇启发的持续表征学习框架，提升去相关性并加速训练。 representation learning ✅

⬅️ 返回 cs.LG 首页 · 🏠 返回主页