Automated Short Answer Scoring on SAS Physics subject (OOD)
37.8AccuracyEDIT-SFT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EDIT-SFTCategory=EDIT (Ours)2026.06 | 37.8 | 55.6 | 1.6 | 70.1 | |
| EDIT-RLCategory=EDIT (Ours)2026.06 | 37.8 | 53.3 | 1.644 | 72.9 | |
| RLCategory=InT2026.06 | 35.6 | 44.4 | 2.022 | 63 | |
| BaseCategory=Base2026.06 | 33.3 | 51.1 | 1.756 | 69.3 | |
| GRPOCategory=RL-only2026.06 | 31.1 | 48.9 | 2.133 | 51.8 | |
| SFTCategory=InT2026.06 | 31.1 | 44.4 | 2.444 | 42.7 | |
| DGPOCategory=RL-only2026.06 | 28.9 | 48.9 | 1.956 | 60.2 |