| zero-shot-action-recognition-on-kinetics-400-zero-shot-action-classification | MoE-ViE-H/14 | #2 | Accuracy: 76.9 |
| zero-shot-action-recognition-on-ucf101 | MoE-ViE-H/14 | #2 | Top-1 Accuracy: 92.8 |
| zero-shot-video-retrieval-on-msr-vtt-video-to-text | MoE-ViE-L/16 | #2 | R@1: 49.8 |
| image-classification-on-objectnet | MoE-ViE-H/14 | #3 | Top-1 Accuracy: 87.0 |
| zero-shot-action-recognition-on-hmdb51 | MoE-ViE-H/14 | #3 | Top-1 Accuracy: 61.3 |
| image-classification-on-imagenet-v2 | MoE-ViE-H/14 | #16 | Top 1 Accuracy: 80.0 |
| video-question-answering-on-mvbench | MoE-ViE-H/14 + Llama 3.1 8B (42M alignment) | #25 | Avg.: 71.5 |
| zero-shot-cross-modal-retrieval-on-coco-2014 | MoE-ViE-L/16 | #26 | Text-to-image R@1: 57.2 |
| image-classification-on-imagenet | MoE-ViE-H/14 (linear probe) | #32 | Top 1 Accuracy: 88.85% |
| visual-question-answering-on-docvqa-test | MoE-ViE-H/14 + Llama 3.1 8B (42M alignment) | #40 | ANLS: 0.904 |
| zero-shot-video-question-answer-on-video-mme-w-o-subs | MoE-ViE-H/14 + Llama 3.1 8B (42M alignment) | #51 | Accuracy (%): 53.0 |
| chart-question-answering-on-chartqa | MoE-ViE-H/14 + Llama 3.1 8B (42M alignment) | #66 | 1:1 Accuracy: 80.8 |