Visual Question Answering on RAD-VQA (COMET)
86.4COMET ScoreR_form + R_exact + R_DTW
Evaluation results
| Method | Links | |
|---|---|---|
| R_form + R_exact + R_DTWDescription=Trajectory-Aware Process Supervision2026.04 | 86.4 | |
| R_form + R_exact2026.04 | 86.1 | |
| SFT2026.04 | 82.6 | |
| R_form2026.04 | 82.4 |