Multimodal Reasoning Efficiency on MM-Vet
3.73Time (s/sample)MIRROR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MIRROR2026.02 | 3.73 | 112.58 | |
| Adaptive-CoFParadigm=Thinking with Images2026.02 | 4.02 | 143.51 | |
| DeepEyesParadigm=Thinking with Images2026.02 | 5.06 | 197.51 | |
| VL-RethinkerParadigm=Text Reflection2026.02 | 5.56 | 229.33 | |
| PixelReasonerParadigm=Thinking with Images2026.02 | 64.12 | 273.95 |