Automated Short Answer Scoring on Private-Biology (OOD)
0.489AccuracyEDIT-RL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EDIT-RLStrategy=EDIT, Training=RL2026.06 | 0.489 | 0.84 | 0.687 | 0.575 | |
| GRPOStrategy=RL-only2026.06 | 0.478 | 0.825 | 0.718 | 0.535 | |
| DGPOStrategy=RL-only2026.06 | 0.478 | 0.817 | 0.725 | 0.533 | |
| EDIT-SFTStrategy=EDIT, Training=SFT2026.06 | 0.475 | 0.816 | 0.735 | 0.548 | |
| BaseStrategy=Base2026.06 | 0.46 | 0.785 | 0.786 | 0.524 | |
| InTStrategy=InT, Training=RL2026.06 | 0.425 | 0.836 | 0.772 | 0.508 | |
| InTStrategy=InT, Training=SFT2026.06 | 0.422 | 0.837 | 0.777 | 0.494 |