Open-ended Visual Question Answering on Med-CXR
31.73AccuracyT2R-R3
Evaluation Results
| Method | Links | |
|---|---|---|
| T2R-R3Scale=2B2026.06 | 31.73 | |
| Judge RewardScale=2B2026.06 | 28 | |
| T2R-R2Scale=2B2026.06 | 27.51 | |
| Frozen TeacherScale=4B2026.06 | 24.41 | |
| T2R-R3Scale=4B2026.06 | 23.23 | |
| T2R-R2Scale=4B2026.06 | 22.87 | |
| T2R-R1Scale=2B2026.06 | 22.5 | |
| Judge RewardScale=4B2026.06 | 22.38 | |
| Frozen TeacherScale=2B2026.06 | 21.74 | |
| Base ModelScale=4B2026.06 | 21.74 | |
| RLVR OnlyScale=4B2026.06 | 21.34 | |
| T2R-R1Scale=4B2026.06 | 21.27 | |
| RLVR OnlyScale=2B2026.06 | 16.87 | |
| Base ModelScale=2B2026.06 | 15.94 |