Compositional Reasoning on WhatsUp
51.3AccuracyReasonCLIP
Evaluation Results
| Method | Links | |
|---|---|---|
| ReasonCLIPViT=B/32, Resolution=224, Data=+58M2026.06 | 51.3 | |
| ReasonSigLIPViT=So/14, Resolution=384, Data=+58M2026.06 | 49.7 | |
| MetaCLIPViT=L/14, Resolution=224, Data=2.5B2026.06 | 48.6 | |
| SigLIPViT=So/14, Resolution=384, Data=10.0B2026.06 | 47.6 | |
| SigLIP2ViT=G/16, Resolution=384, Data=10.0B2026.06 | 46.9 | |
| MetaCLIPViT=B/32, Resolution=224, Data=0.4B2026.06 | 46.5 | |
| Eva-CLIP-02ViT=L/14, Resolution=224, Data=2.0B2026.06 | 46.3 | |
| ReasonCLIPViT=L/14, Resolution=336, Data=+58M2026.06 | 46.3 | |
| ReasonCLIPViT=L/14, Resolution=224, Data=+58M2026.06 | 46 | |
| Long-CLIPViT=L/14, Resolution=224, Data=0.4B2026.06 | 45.6 | |
| SigLIP2ViT=L/16, Resolution=384, Data=10.0B2026.06 | 45.5 | |
| ReasonSigLIP2ViT=G/16, Resolution=384, Data=+58M2026.06 | 45.3 | |
| ReasonCLIP + READViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 45.2 | |
| READ-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 43.9 | |
| NegCLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 42.4 | |
| CLIPViT=L/14, Resolution=336, Data=0.4B2026.06 | 41.8 | |
| Triplet-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 41.6 | |
| GNM-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 41.6 | |
| CLIPViT=B/32, Resolution=224, Data=0.4B2026.06 | 41.2 | |
| CE-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 40.7 | |
| CLIPViT=L/14, Resolution=224, Data=0.4B2026.06 | 40.7 |