Vision-Language Compositional Reasoning on ARO (test)
71.4VG-Relsyn-CLIP
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| syn-CLIPtraining=finetuned on SyViC2023.03 | 71.4 | 66.94 | 59.06 | 70.96 | 67.09 | |
| syn-CyCLIPtraining=finetuned on SyViC2023.03 | 69.02 | 63.65 | 49.17 | 59.36 | 60.3 | |
| CyCLIPtraining=pre-trained on large-scale real data2023.03 | 59.12 | 65.41 | 20.82 | 29.54 | 43.72 | |
| CLIPtraining=pre-trained on large-scale real data2023.03 | 58.84 | 63.19 | 47.2 | 59.46 | 57.17 |