Question Answering on PubMedQA (Accuracy/F1)
98.1Accuracy/F1ReGrad + ICL
Evaluation results
| Method | Links | |
|---|---|---|
| ReGrad + ICLModel=LLaMA-3.1 8B2026.06 | 98.1 | |
| ReGrad + ICLModel=LLaMA-3.2 1B2026.06 | 97.9 | |
| ReGrad + ICLModel=LLaMA-3.2 3B2026.06 | 97.9 | |
| ReGradModel=LLaMA-3.2 1B2026.06 | 93.3 | |
| ReGradModel=LLaMA-3.1 8B2026.06 | 93.2 | |
| ReGradModel=LLaMA-3.2 3B2026.06 | 92.8 |