List your company
ResearchFormerly WizWand

Language Modeling on GPT-2

0.44Delta PPLRandom
-8.708453.0433114.795176.5467Apr 3, 2026
Updated 6mo ago

Evaluation results

MethodLinks
RandomModel=Large (36L), k=12026.040.44—
SpectralModel=Large (36L), k=12026.040.51.5
WorstModel=Large (36L), k=12026.040.62—
WorstModel=Medium (24L), k=12026.040.73—
Last-NModel=Large (36L), k=12026.040.75—
SpectralModel=Medium (24L), k=12026.040.842.4
SpectralModel=Large (36L), k=22026.041.041.9
RandomModel=Medium (24L), k=12026.041.06—
RandomModel=Large (36L), k=22026.041.06—
RandomModel=Large (36L), k=32026.041.51—
SpectralModel=Large (36L), k=32026.041.682
Last-NModel=Large (36L), k=22026.041.92—
Last-NModel=Medium (24L), k=12026.041.98—
RandomModel=Large (36L), k=42026.042.18—
RandomModel=Medium (24L), k=22026.042.36—
WorstModel=Medium (24L), k=22026.042.66—
SpectralModel=Medium (24L), k=22026.042.672.3
SpectralModel=Large (36L), k=42026.042.961.9
Last-NModel=Large (36L), k=32026.043.41—
RandomModel=Medium (24L), k=32026.043.79—
SpectralModel=Medium (24L), k=32026.044.892.8
Last-NModel=Large (36L), k=42026.045.55—
RandomModel=Medium (24L), k=42026.045.62—
Last-NModel=Medium (24L), k=22026.046.04—
SpectralModel=Medium (24L), k=42026.049.652.6
Last-NModel=Medium (24L), k=32026.0413.49—
WorstModel=Medium (24L), k=32026.0419.49—
WorstModel=Large (36L), k=22026.0421.2—
Last-NModel=Medium (24L), k=42026.0425.06—
WorstModel=Large (36L), k=32026.0426.8—
WorstModel=Large (36L), k=42026.0433.36—
WorstModel=Medium (24L), k=42026.04229.15—