cs.LG（2025-08-30）

📊 共 10 篇论文

🎯 兴趣领域导航

#	题目	一句话要点	标签
1	RoFt-Mol: Benchmarking Robust Fine-Tuning with Molecular Graph Foundation Models	提出ROFT-MOL以解决分子图基础模型的鲁棒微调问题	foundation model
2	Universal Properties of Activation Sparsity in Modern Large Language Models	提出通用框架以研究现代大语言模型的激活稀疏性	large language model
3	Privacy-Preserving Offloading for Large Language Models in 6G Vehicular Networks	提出隐私保护的卸载框架以解决6G车联网中的数据安全问题	large language model
4	TimeCopilot	提出TimeCopilot框架以实现高效的时间序列预测	large language model foundation model
5	SQL-of-Thought: Multi-agentic Text-to-SQL with Guided Error Correction	提出SQL-of-Thought框架以解决文本到SQL转换问题	chain-of-thought
6	Memory Limitations of Prompt Tuning in Transformers	提出对Transformer记忆限制的理论分析以解决提示调优问题	large language model
7	Metis: Training LLMs with FP4 Quantization	提出Metis框架以解决大语言模型低位训练中的量化偏差问题	large language model

#	题目	一句话要点	标签
8	LLM-Driven Policy Diffusion: Enhancing Generalization in Offline Reinforcement Learning	提出LLM驱动的策略扩散以解决离线强化学习的泛化问题	reinforcement learning policy learning offline RL
9	An Efficient GNNs-to-KANs Distillation via Self-Attention Dynamic Sampling with Potential for Consumer Electronics Edge Deployment	提出SA-DSD框架以提升GNN知识蒸馏至KAN的效率	teacher-student distillation
10	Are We Really Learning the Score Function? Reinterpreting Diffusion Models Through Wasserstein Gradient Flow Matching	提出Wasserstein梯度流匹配以重新理解扩散模型	flow matching