Visio-linguistics Compositionality on Hard Positive Compositional Benchmark (test)
68.3Original Test AccuracyMACCO-CLIP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MACCO-CLIP2026.06 | 68.3 | 48.5 | |
| CLIPtraining=pre-trained2026.06 | 63.2 | 47.2 | |
| CLIP-FTtraining=fine-tuned2026.06 | 62.3 | 48.3 |