List your company

Zero-shot Evaluation on Arc-e, PIQA, Hellaswag, OpenBookQA, Winogrande, MMLU, and BoolQ Suite

53.83Arc-eBHyT
40.51843.97447.4350.886Dec 26, 2025

Evaluation results

MethodLinks
BHyTTraining Stage=Pretrained & SFT, Model=Llama-3B2025.1253.8372.3651.6230.653.0726.8249.8948.31
Peri-LNTraining Stage=Pretrained & SFT, Model=Llama-3B2025.1250.1970.9348.393151.9826.3642.4345.9
BHyTTraining Stage=Pretrained-only, Model=Llama-3B2025.1245.9370.0950.9731.3252.1227.6255.6947.68
Peri-LNTraining Stage=Pretrained-only, Model=Llama-3B2025.1241.0368.9347.4730.6451.727.0247.1944.85