MMODELYST
People/Shiwei Liu
PPL

Shiwei Liu

Papers (2)
One LR Doesn't Fit All: Heavy-Tail Guided Layerwise Learning Rates for LLMsLayer Collapse in Diffusion Language Models