一作NeurIPS 2026Poster· Optimization / Large Scale, Parallel and Distributed
BReD: Block Replay Dithering for Stable Low-Bit EMA Optimizer States
Heshen Zhan · Youhan Huang · Yunke Peng · Yao Wang · Linghui Kong · Ziwei Zhu · Qingyu Han · Yaoyuan Wang · Congliang Chen · Ruoyu Sun