Image Classification on MNIST (Top-1 Accuracy)
63.6Top-1 AccuracyCLIP
Evaluation results
| Method | Links | |
|---|---|---|
| CLIPZero-shot=true, Backbone=ViT-B/16, Pre-training Dataset=Laion100M, Epochs=302026.03 | 63.6 | |
| ITOZero-shot=true, Backbone=ViT-B/16, Pre-training Dataset=Laion100M, Epochs=302026.03 | 58.1 | |
| SLIPZero-shot=true, Backbone=ViT-B/16, Pre-training Dataset=Laion100M, Epochs=302026.03 | 57 | |
| SigLIPPre-training Dataset=CC12M, Backbone=ViT-B/16, Epochs=30, Evaluation protocol=Zero-shot2026.03 | 12.9 | |
| ITO sub2Pre-training Dataset=CC12M, Backbone=ViT-B/16, Epochs=30, Evaluation protocol=Zero-shot2026.03 | 11.9 | |
| ITOPre-training Dataset=CC12M, Backbone=ViT-B/16, Epochs=30, Evaluation protocol=Zero-shot2026.03 | 10.3 | |
| CLIPPre-training Dataset=CC12M, Backbone=ViT-B/16, Epochs=30, Evaluation protocol=Zero-shot2026.03 | 9.6 | |
| SLIPPre-training Dataset=CC12M, Backbone=ViT-B/16, Epochs=30, Evaluation protocol=Zero-shot2026.03 | 9 | |
| FLAIRPre-training Dataset=CC12M, Backbone=ViT-B/16, Epochs=30, Evaluation protocol=Zero-shot2026.03 | 4.6 |