List your company
ResearchFormerly WizWand

Generalization gap prediction on QNLI

0.14Generalization Gap Error1-sharpness
0.11840.26420.410.5558May 31, 2023
Updated 8mo ago

Evaluation results

MethodLinks
1-sharpnessTraining scenario=Fine-tuning, Data source=Test loss landscape (dev data, unfair advantage)2023.050.14
HessianTraining scenario=Fine-tuning, Data source=Test loss landscape (dev data, unfair advantage)2023.050.15
InconsistencyTraining scenario=Fine-tuning, Data source=Unlabeled data2023.050.19
1-sharpnessTraining scenario=Fine-tuning, Data source=Training loss landscape2023.050.24
HessianTraining scenario=Fine-tuning, Data source=Training loss landscape2023.050.68