Automated Short Answer Scoring on Private-Physics (ID)
51.2AccuracyInT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| InTStrategy=InT, Training=RL2026.06 | 51.2 | 82.2 | 0.782 | 0.551 | |
| EDIT-RLStrategy=EDIT, Training=RL2026.06 | 51.2 | 81.5 | 0.795 | 0.555 | |
| EDIT-SFTStrategy=EDIT, Training=SFT2026.06 | 50.6 | 78.2 | 0.812 | 0.527 | |
| DGPOStrategy=RL-only2026.06 | 49.5 | 79.5 | 0.801 | 0.535 | |
| GRPOStrategy=RL-only2026.06 | 48.9 | 78 | 0.823 | 0.48 | |
| InTStrategy=InT, Training=SFT2026.06 | 45.4 | 80.4 | 0.813 | 0.454 | |
| BaseStrategy=Base2026.06 | 43.8 | 76.2 | 0.906 | 0.477 |