Zero-shot Question Answering on MMLU
1.03Average LossPre-trained
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Pre-trainedModel=LLaMa-3, # Params=8.03B, System=FP322024.11 | 1.03 | — | 1.03 | 1.03 | 36.1 | 16 | |
| LPE-BSRModel=LLaMa-3, # Params=8.03B, System=INT-5, Ensemble size=202024.11 | 1.07 | 0.049 | 1.02 | 0.923 | 36.4 | 8.7 |