List your company

Language Modeling on OpenWebText GPT-2 (train)

3.41LossStoSignSGD
3.404843.439673.47453.50933Apr 16, 2026

Evaluation results

MethodLinks
StoSignSGDModel=GPT-2, Precision=FP82026.043.4130.26
LionModel=GPT-2, Precision=FP82026.043.45231.55
SignSGDModel=GPT-2, Precision=FP82026.043.53934.43