List your company

Scaling Law Fitting on Enoki (train)

0.405Scaling Parameter bAdamW
0.403960.404230.40450.40477Feb 5, 2026

Evaluation results

MethodLinks
AdamWOptimizer=AdamW2026.020.4050.2112.680.031
Dana-Star-MK4Optimizer=Dana-Star-MK4, kappa=0.852026.020.4040.2172.640.030.9999
ADanaOptimizer=ADana, kappa=0.852026.020.4040.2172.640.031
MuonOptimizer=Muon2026.020.4040.2172.650.0290.9999
AdemamixOptimizer=Ademamix, Weight Decay=dec. WD2026.020.4040.2182.640.0291