Automated Short Answer Scoring on SAS Average across subjects (OOD)
42.9AccuracyEDIT-RL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EDIT-RLCategory=EDIT (Ours)2026.06 | 42.9 | 58.6 | 1.52 | 79.4 | |
| EDIT-SFTCategory=EDIT (Ours)2026.06 | 42.1 | 58.6 | 1.606 | 78.1 | |
| GRPOCategory=RL-only2026.06 | 39.4 | 54 | 1.85 | 68.9 | |
| BaseCategory=Base2026.06 | 38.9 | 52.5 | 1.95 | 72.7 | |
| RLCategory=InT2026.06 | 38.3 | 53.4 | 1.753 | 73.1 | |
| DGPOCategory=RL-only2026.06 | 37.2 | 56.3 | 1.784 | 72.1 | |
| SFTCategory=InT2026.06 | 36.8 | 53.5 | 1.943 | 68.1 |