Automated Short Answer Scoring on Private-Biology (ID)
50.7AccuracyEDIT-RL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EDIT-RLStrategy=EDIT, Training=RL2026.06 | 50.7 | 83.1 | 0.717 | 0.754 | |
| GRPOStrategy=RL-only2026.06 | 49.1 | 80.9 | 0.753 | 0.727 | |
| EDIT-SFTStrategy=EDIT, Training=SFT2026.06 | 48.5 | 82.1 | 0.756 | 0.738 | |
| DGPOStrategy=RL-only2026.06 | 47 | 80.1 | 0.79 | 0.704 | |
| BaseStrategy=Base2026.06 | 45.6 | 79.3 | 0.821 | 0.701 | |
| InTStrategy=InT, Training=RL2026.06 | 44.8 | 82.8 | 0.803 | 0.648 | |
| InTStrategy=InT, Training=SFT2026.06 | 43.7 | 81.2 | 0.838 | 0.63 |