List your company

Grokking Onset and Speedup on 3-digit Addition (NanoGPT Implementation)

9,800Grokking OnsetNorm Penalty
9,29212,72116,15019,579Jun 30, 2026

Evaluation results

MethodLinks
Norm PenaltyArchitecture=NanoGPT, Regularization=Activation Norm Penalty2026.069,8002.3
No-penalty BaselineArchitecture=NanoGPT, Regularization=None2026.0622,500—