| image-sentence-alignment-on-valse | ViLBERT 12-in-1 | #1 | average pairwise accuracy: 75.1Average Accuracy: 63.2 |
| image-sentence-alignment-on-valse | CLIP | #2 | average pairwise accuracy: 64.0 |
| image-sentence-alignment-on-valse | ViLBERT | #3 | average pairwise accuracy: 63.7Average Accuracy: 51.3 |
| image-sentence-alignment-on-valse | GPT1 | #4 | average pairwise accuracy: 60.7 |
| image-sentence-alignment-on-valse | GPT2 | #5 | average pairwise accuracy: 60.1 |
| image-sentence-alignment-on-valse | LXMERT | #6 | average pairwise accuracy: 59.6Average Accuracy: 53.5 |
| image-sentence-alignment-on-valse | VisualBERT | #7 | average pairwise accuracy: 46.4Average Accuracy: 48.8 |
| image-sentence-alignment-on-valse-actant-swap | GPT2 | #1 | pairwise accuracy: 76.9 |
| image-sentence-alignment-on-valse-actant-swap | GPT1 | #2 | pairwise accuracy: 72.2 |
| image-sentence-alignment-on-valse-actant-swap | CLIP | #3 | pairwise accuracy: 68.6 |
| image-sentence-alignment-on-valse-actant-swap | ViLBERT | #4 | pairwise accuracy: 68.3Accuracy (%): 50.4 |
| image-sentence-alignment-on-valse-actant-swap | ViLBERT 12-in-1 | #5 | pairwise accuracy: 58.9Accuracy (%): 52.2 |
| image-sentence-alignment-on-valse-actant-swap | LXMERT | #6 | pairwise accuracy: 45.8Accuracy (%): 48.5 |
| image-sentence-alignment-on-valse-actant-swap | VisualBERT | #7 | pairwise accuracy: 44.4Accuracy (%): 49.7 |
| image-sentence-alignment-on-valse-action | CLIP | #1 | pairwise accuracy: 75.6 |
| image-sentence-alignment-on-valse-action | ViLBERT | #2 | pairwise accuracy: 70.7Accuracy (%): 52.6 |
| image-sentence-alignment-on-valse-action | GPT2 | #3 | pairwise accuracy: 66.8 |
| image-sentence-alignment-on-valse-action | ViLBERT 12-in-1 | #4 | pairwise accuracy: 65.9Accuracy (%): 57.3 |
| image-sentence-alignment-on-valse-action | GPT1 | #5 | pairwise accuracy: 65.4 |
| image-sentence-alignment-on-valse-action | LXMERT | #6 | pairwise accuracy: 54.8Accuracy (%): 51.1 |
| image-sentence-alignment-on-valse-action | VisualBERT | #7 | pairwise accuracy: 49.2Accuracy (%): 48.8 |
| image-sentence-alignment-on-valse-coreference | ViLBERT 12-in-1 | #1 | pairwise accuracy: 75.7Accuracy (%): 54.4 |
| image-sentence-alignment-on-valse-coreference | GPT2 | #2 | pairwise accuracy: 54.5 |
| image-sentence-alignment-on-valse-coreference | CLIP | #3 | pairwise accuracy: 52.1 |
| image-sentence-alignment-on-valse-coreference | VisualBERT | #4 | pairwise accuracy: 49.5Accuracy (%): 50.0 |
| image-sentence-alignment-on-valse-coreference | ViLBERT | #5 | pairwise accuracy: 47.2Accuracy (%): 50.0 |
| image-sentence-alignment-on-valse-coreference | LXMERT | #6 | pairwise accuracy: 46.8Accuracy (%): 49.8 |
| image-sentence-alignment-on-valse-coreference | GPT1 | #7 | pairwise accuracy: 45.6 |
| image-sentence-alignment-on-valse-coreference-1 | ViLBERT 12-in-1 | #1 | pairwise accuracy: 69.2Accuracy (%): 54.3 |
| image-sentence-alignment-on-valse-coreference-1 | GPT2 | #2 | pairwise accuracy: 50.0 |
| image-sentence-alignment-on-valse-coreference-1 | CLIP | #3 | pairwise accuracy: 49.7 |
| image-sentence-alignment-on-valse-coreference-1 | ViLBERT | #4 | pairwise accuracy: 48.1Accuracy (%): 50.0 |
| image-sentence-alignment-on-valse-coreference-1 | VisualBERT | #5 | pairwise accuracy: 47.6Accuracy (%): 50.0 |
| image-sentence-alignment-on-valse-coreference-1 | GPT1 | #6 | pairwise accuracy: 45.2 |
| image-sentence-alignment-on-valse-coreference-1 | LXMERT | #7 | pairwise accuracy: 44.2Accuracy (%): 49.0 |
| image-sentence-alignment-on-valse-counting | ViLBERT 12-in-1 | #1 | pairwise accuracy: 76.7Accuracy (%): 64.9 |
| image-sentence-alignment-on-valse-counting | LXMERT | #2 | pairwise accuracy: 62.2Accuracy (%): 52.0 |
| image-sentence-alignment-on-valse-counting | CLIP | #3 | pairwise accuracy: 62.1 |
| image-sentence-alignment-on-valse-counting | ViLBERT | #4 | pairwise accuracy: 58.6Accuracy (%): 50.7 |
| image-sentence-alignment-on-valse-counting | GPT2 | #5 | pairwise accuracy: 51.6 |
| image-sentence-alignment-on-valse-counting | GPT1 | #6 | pairwise accuracy: 51.2 |
| image-sentence-alignment-on-valse-counting | VisualBERT | #7 | pairwise accuracy: 48.2Accuracy (%): 48.3 |
| image-sentence-alignment-on-valse-counting-1 | ViLBERT 12-in-1 | #1 | pairwise accuracy: 80.2Accuracy (%): 69.2 |
| image-sentence-alignment-on-valse-counting-1 | LXMERT | #2 | pairwise accuracy: 69.2Accuracy (%): 55.4 |
| image-sentence-alignment-on-valse-counting-1 | ViLBERT | #3 | pairwise accuracy: 62.9Accuracy (%): 50.6 |
| image-sentence-alignment-on-valse-counting-1 | CLIP | #4 | pairwise accuracy: 62.5 |
| image-sentence-alignment-on-valse-counting-1 | GPT2 | #5 | pairwise accuracy: 49.8 |
| image-sentence-alignment-on-valse-counting-1 | GPT1 | #6 | pairwise accuracy: 48.7 |
| image-sentence-alignment-on-valse-counting-1 | VisualBERT | #7 | pairwise accuracy: 48.2Accuracy (%): 47.8 |
| image-sentence-alignment-on-valse-counting-2 | ViLBERT 12-in-1 | #1 | pairwise accuracy: 77.3Accuracy (%): 66.7 |
| image-sentence-alignment-on-valse-counting-2 | ViLBERT | #2 | pairwise accuracy: 73.7Accuracy (%): 51.8 |
| image-sentence-alignment-on-valse-counting-2 | GPT1 | #3 | pairwise accuracy: 69.5 |
| image-sentence-alignment-on-valse-counting-2 | CLIP | #4 | pairwise accuracy: 57.5 |
| image-sentence-alignment-on-valse-counting-2 | VisualBERT | #5 | pairwise accuracy: 50.0Accuracy (%): 50.0 |
| image-sentence-alignment-on-valse-counting-2 | GPT2 | #6 | pairwise accuracy: 45.3 |
| image-sentence-alignment-on-valse-counting-2 | LXMERT | #7 | pairwise accuracy: 42.6Accuracy (%): 49.9 |
| image-sentence-alignment-on-valse-existence | ViLBERT 12-in-1 | #1 | pairwise accuracy: 95.6Accuracy (%): 89.0 |
| image-sentence-alignment-on-valse-existence | LXMERT | #2 | pairwise accuracy: 78.6Accuracy (%): 55.8 |
| image-sentence-alignment-on-valse-existence | CLIP | #3 | pairwise accuracy: 66.9 |
| image-sentence-alignment-on-valse-existence | ViLBERT | #4 | pairwise accuracy: 66.5Accuracy (%): 2.4 |
| image-sentence-alignment-on-valse-existence | GPT1 | #5 | pairwise accuracy: 61.8 |
| image-sentence-alignment-on-valse-existence | GPT2 | #6 | pairwise accuracy: 58.0 |
| image-sentence-alignment-on-valse-existence | VisualBERT | #7 | pairwise accuracy: 39.7Accuracy (%): 49.3 |
| image-sentence-alignment-on-valse-foil-it | CLIP | #1 | pairwise accuracy: 88.8 |
| image-sentence-alignment-on-valse-foil-it | LXMERT | #2 | pairwise accuracy: 87.1Accuracy (%): 70.8 |
| image-sentence-alignment-on-valse-foil-it | ViLBERT 12-in-1 | #3 | pairwise accuracy: 86.9Accuracy (%): 71.5 |
| image-sentence-alignment-on-valse-foil-it | ViLBERT | #4 | pairwise accuracy: 86.9Accuracy (%): 55.9 |
| image-sentence-alignment-on-valse-foil-it | GPT2 | #5 | pairwise accuracy: 80.7 |
| image-sentence-alignment-on-valse-foil-it | GPT1 | #6 | pairwise accuracy: 77.5 |
| image-sentence-alignment-on-valse-foil-it | VisualBERT | #7 | pairwise accuracy: 48.5Accuracy (%): 46.6 |
| image-sentence-alignment-on-valse-plurality | ViLBERT 12-in-1 | #1 | pairwise accuracy: 72.4Accuracy (%): 62.0 |
| image-sentence-alignment-on-valse-plurality | LXMERT | #2 | pairwise accuracy: 64.4Accuracy (%): 55.1 |
| image-sentence-alignment-on-valse-plurality | ViLBERT | #3 | pairwise accuracy: 61.2Accuracy (%): 50.3 |
| image-sentence-alignment-on-valse-plurality | CLIP | #4 | pairwise accuracy: 56.2 |
| image-sentence-alignment-on-valse-plurality | GPT1 | #5 | pairwise accuracy: 53.1 |
| image-sentence-alignment-on-valse-plurality | GPT2 | #6 | pairwise accuracy: 51.9 |
| image-sentence-alignment-on-valse-plurality | VisualBERT | #7 | pairwise accuracy: 45.7Accuracy (%): 46.5 |
| image-sentence-alignment-on-valse-spatial | GPT1 | #1 | pairwise accuracy: 77.2 |
| image-sentence-alignment-on-valse-spatial | GPT2 | #2 | pairwise accuracy: 75.0 |
| image-sentence-alignment-on-valse-spatial | ViLBERT 12-in-1 | #3 | pairwise accuracy: 67.7Accuracy (%): 53.4 |
| image-sentence-alignment-on-valse-spatial | CLIP | #4 | pairwise accuracy: 64.3 |
| image-sentence-alignment-on-valse-spatial | LXMERT | #5 | pairwise accuracy: 60.2Accuracy (%): 50.8 |
| image-sentence-alignment-on-valse-spatial | ViLBERT | #6 | pairwise accuracy: 57.2Accuracy (%): 49.9 |
| image-sentence-alignment-on-valse-spatial | VisualBERT | #7 | pairwise accuracy: 39.7Accuracy (%): 49.3 |