Correlation with Human Judgments on FinePhyEval
0.478Pearson's rPQSG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PQSGBackbone=GPT-5.52026.06 | 0.478 | 0.336 | 0.456 | |
| PQSGBackbone=Gemini-2.5-Pro2026.06 | 0.467 | 0.306 | 0.406 | |
| Direct VQA2026.06 | 0.382 | 0.29 | 0.36 | |
| VideoPhy-2-Autoeval2026.06 | 0.346 | 0.277 | 0.349 | |
| DSG2026.06 | 0.302 | 0.195 | 0.265 | |
| VideoScore2026.06 | 0.289 | 0.262 | 0.378 | |
| PhyGenEval2026.06 | 0.272 | 0.22 | 0.264 |