General Evaluation on Datacomp small (38 tasks)
19.7Average ScoreFinal Self-Filtered 30% Data Mix
Evaluation results
| Method | Links | |
|---|---|---|
| Final Self-Filtered 30% Data MixBackbone=ViT-B/32, Data=Final data mix from last round of Self-Filtering, Evaluation Protocol=Zero-shot2026.06 | 19.7 | |
| CLIP with OAI CLIP Filtered data + AllBackbone=ViT-B/32, Data=OAI Filtered + All data mix, Evaluation Protocol=Zero-shot2026.06 | 19.6 | |
| Self-Filtering 40%Backbone=ViT-B/32, Data=40% subset selected via Self-Filtering, Evaluation Protocol=Zero-shot2026.06 | 19.5 | |
| Self-Filtering 30%Backbone=ViT-B/32, Data=30% subset selected via Self-Filtering, Evaluation Protocol=Zero-shot2026.06 | 19.2 | |
| CLIP with All dataBackbone=ViT-B/32, Data=11.49M samples (Datacomp small), Evaluation Protocol=Zero-shot2026.06 | 19 | |
| CLIP with OAI CLIP Filtered dataBackbone=ViT-B/32, Data=Data selected by OpenAI’s CLIP ViT-L/14, Evaluation Protocol=Zero-shot2026.06 | 17.3 |