Compositional Reasoning on CREPE
44.9AccuracyReasonCLIP + READ
Evaluation Results
| Method | Links | |
|---|---|---|
| ReasonCLIP + READViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 44.9 | |
| READ-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 41.5 | |
| Long-CLIPViT=L/14, Resolution=224, Data=0.4B2026.06 | 35.1 | |
| CE-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 34.8 | |
| NegCLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 30.5 | |
| ReasonCLIPViT=L/14, Resolution=224, Data=+58M2026.06 | 24.8 | |
| ReasonCLIPViT=L/14, Resolution=336, Data=+58M2026.06 | 24.8 | |
| ReasonCLIPViT=B/32, Resolution=224, Data=+58M2026.06 | 24 | |
| CLIPViT=B/32, Resolution=224, Data=0.4B2026.06 | 23.9 | |
| Eva-CLIP-02ViT=L/14, Resolution=224, Data=2.0B2026.06 | 21.4 | |
| CLIPViT=L/14, Resolution=336, Data=0.4B2026.06 | 20.9 | |
| CLIPViT=L/14, Resolution=224, Data=0.4B2026.06 | 20.5 | |
| ReasonSigLIP2ViT=G/16, Resolution=384, Data=+58M2026.06 | 20.5 | |
| ReasonSigLIPViT=So/14, Resolution=384, Data=+58M2026.06 | 20.1 | |
| MetaCLIPViT=L/14, Resolution=224, Data=2.5B2026.06 | 19.7 | |
| MetaCLIPViT=B/32, Resolution=224, Data=0.4B2026.06 | 18.2 | |
| SigLIPViT=So/14, Resolution=384, Data=10.0B2026.06 | 18 | |
| GNM-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 17.4 | |
| SigLIP2ViT=G/16, Resolution=384, Data=10.0B2026.06 | 16.5 | |
| SigLIP2ViT=L/16, Resolution=384, Data=10.0B2026.06 | 15.9 | |
| Triplet-CLIPViT=B/32, Resolution=224, Fine-tuned on MS-COCO=true2026.06 | 15 |