Open-ended Visual Question Answering on SLAKE (Accuracy)
64.92AccuracyFrozen Teacher
Evaluation Results
| Method | Links | |
|---|---|---|
| Frozen TeacherScale=4B2026.06 | 64.92 | |
| T2R-R2Scale=4B2026.06 | 64.92 | |
| T2R-R1Scale=4B2026.06 | 61.25 | |
| T2R-R3Scale=4B2026.06 | 60.73 | |
| Base ModelScale=4B2026.06 | 58.11 | |
| RLVR OnlyScale=4B2026.06 | 56.84 | |
| Judge RewardScale=4B2026.06 | 56.54 | |
| Judge RewardScale=2B2026.06 | 56.02 | |
| RLVR OnlyScale=2B2026.06 | 53.92 | |
| T2R-R2Scale=2B2026.06 | 53.92 | |
| Base ModelScale=2B2026.06 | 52.87 | |
| Frozen TeacherScale=2B2026.06 | 52.35 | |
| T2R-R1Scale=2B2026.06 | 52.35 | |
| T2R-R3Scale=2B2026.06 | 50.78 |