arXivShaokun Zhang, Yifan Zhang, Jian Hu, Yueying Li, Hao Zhang, Binfeng Xu, Jan Kautz, Yi DongMon, Oct 5, 2026, 9:28 AM PDT
score 17.0
Low-Rank Gradient Sketches Cut Memory for LLM Reinforcement Learning
Original: LoGRA: Scaling LLM Reinforcement Learning with Low-Rank Gradient Sketches
Source: arxiv.org ↗
Writing ELI5 summary…