AdamS: Momentum Itself Can Be A Normalizer for LLM Pretraining and Post-training EMNLP

Huishuai Zhang | Bohan Wang | Luoxin Chen |

BSFA: Leveraging the Subspace Dichotomy to Accelerate Neural Network Training EMNLP

WenJie Zhou | Bohan Wang | Wei Chen | Xueqi Cheng |