Evidence Awareness on RCLIP-Bench Evidence Awareness V2
93Score SHuman
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HumanScale=Unbounded Scale2026.06 | 93 | 89.7 | 93.9 | 88.7 | 90 | 91.1 | |
| GPT-4VScale=Unbounded Scale2026.06 | 92.7 | 83.6 | 89.1 | 87.3 | 80 | 86.5 | |
| ReasonCLIPScale=ViT Base (86M), Resolution=224/32, Data=+42M2026.06 | 18.2 | 19.3 | 25 | 31.5 | 21.3 | 23.1 | |
| ReasonSigLIPScale=ViT So400m (428M), Resolution=384/14, Data=+42M2026.06 | 16.1 | 29.1 | 28.8 | 24.9 | 30.1 | 25.8 | |
| ReasonCLIPScale=ViT Large (307M), Resolution=224/14, Data=+42M2026.06 | 11.7 | 19.2 | 15.7 | 34.3 | 15.2 | 19.2 |