Video Question Answering on NExT-QA Main Dataset
0.796AccuracyVideoMultiAgents
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VideoMultiAgentsZero-shot=true2025.04 | 0.796 | 0.847 | 0.757 | — | 80.5 | |
| TarsierZero-shot=true, Additional training data=true2025.04 | 0.792 | — | — | — | — | |
| LLaMA-VQALanguage Model=LLaMA 33B, Total Parameters=33B, Trainable Parameters=9.2M2023.10 | 0.755 | 0.788 | 0.726 | 0.762 | — | |
| ENTERZero-shot=true2025.04 | 0.751 | 0.792 | 0.682 | — | 77.9 | |
| LLaMA-VQALanguage Model=LLaMA 13B, Total Parameters=13B, Trainable Parameters=6M2023.10 | 0.742 | 0.759 | 0.717 | 0.753 | — | |
| SeViLALanguage Model=FlanT5, Total Parameters=12.1B, Trainable Parameters=188M2023.10 | 0.738 | 0.813 | 0.694 | 0.742 | — | |
| SEVILANumber of Frames=32 -> 4, Keyframe Selection=refined LOCALIZER2023.05 | 0.738 | 0.813 | 0.694 | — | 74.2 | |
| LLoViZero-shot=true2025.04 | 0.738 | 0.819 | 0.702 | — | 73.7 | |
| TS-LLAVAZero-shot=true2025.04 | 0.736 | 0.815 | 0.682 | — | 74.6 | |
| VideoTreeZero-shot=true2025.04 | 0.735 | 0.813 | 0.67 | — | 75.2 | |
| SEVILA†Number of Frames=32 -> 4, Keyframe Selection=zero-shot LOCALIZER2023.05 | 0.734 | 0.835 | 0.688 | — | 73.4 | |
| LVNetZero-shot=true2025.04 | 0.729 | 0.815 | 0.655 | — | 75 | |
| BLIP-2 concat (ANSWERER)Number of Frames=4, Protocol=feature concatenation2023.05 | 0.726 | 0.812 | 0.681 | — | 72.9 | |
| LLaMA-VQALanguage Model=LLaMA 7B, Total Parameters=7B, Trainable Parameters=4.5M2023.10 | 0.72 | 0.758 | 0.692 | 0.727 | — | |
| LLaMA-VQALanguage Model=LLaMA, Number of Trainable Parameters=4.5M2023.10 | 0.72 | 0.758 | 0.692 | — | 0.727 | |
| VideoAgentZero-shot=true2025.04 | 0.713 | 0.811 | 0.645 | — | 72.7 | |
| IG-VLMZero-shot=true2025.04 | 0.709 | 0.773 | 0.657 | — | 72.2 | |
| BLIP-2Language Model=FlanT5, Total Parameters=12.1B, Trainable Parameters=188M2023.10 | 0.701 | 0.801 | 0.652 | 0.701 | — | |
| BLIP-2 votingNumber of Frames=4, Protocol=majority voting2023.05 | 0.701 | 0.801 | 0.652 | — | 70.1 | |
| InternVideoLanguage Model=CLIP text encoder, Number of Trainable Parameters=1.3B2023.10 | 0.632 | 0.758 | 0.585 | — | 0.625 | |
| InternVideoNumber of Frames=8, Tested_by_authors=true2023.05 | 0.632 | 0.758 | 0.585 | — | 62.5 | |
| HiTeASupervision=Yes2023.03 | 0.631 | — | — | — | — | |
| HiTeA2023.10 | 0.631 | 0.756 | 0.583 | — | 0.624 | |
| HiTeANumber of Frames=162023.05 | 0.631 | 0.756 | 0.583 | — | 62.4 | |
| SeVIT-FIDNumber of Frames=102023.05 | 0.606 | — | — | — | — | |
| ViperGPTSupervision=No2023.03 | 0.6 | — | — | — | — | |
| COVGTNumber of Frames=322023.05 | 0.6 | 0.693 | 0.574 | — | 58.8 | |
| VFCNumber of Frames=322023.05 | 0.586 | 0.728 | 0.533 | — | 57.6 | |
| MISTLanguage Model=BERT2023.10 | 0.572 | 0.669 | 0.566 | — | 0.546 | |
| MISTNumber of Frames=322023.05 | 0.571 | 0.669 | 0.566 | — | 54.6 | |
| VGTSupervision=Yes2023.03 | 0.569 | — | — | — | — | |
| VGTLanguage Model=BERT2023.10 | 0.569 | 0.695 | 0.564 | — | 0.534 | |
| VGTNumber of Frames=322023.05 | 0.55 | 0.64 | 0.55 | — | 52.2 | |
| Temp[ATP] + ATP2022.06 | 0.543 | 0.668 | 0.502 | 0.531 | — | |
| ATPSupervision=Yes2023.03 | 0.543 | — | — | — | — | |
| ATPLanguage Model=CLIP text encoder2023.10 | 0.543 | 0.668 | 0.502 | — | 0.531 | |
| JustAskNumber of Frames=20, Speech Input=true2023.05 | 0.523 | 0.631 | 0.514 | — | 49.6 | |
| Temp[ATP]2022.06 | 0.515 | 0.65 | 0.493 | 0.486 | — | |
| Temp[ATP]Number of Frames=322023.05 | 0.515 | 0.65 | 0.493 | — | 48.6 | |
| AIONumber of Trainable Parameters=110M2023.10 | 0.506 | 0.632 | 0.486 | — | 0.48 | |
| All-in-OneNumber of Frames=322023.05 | 0.506 | 0.632 | 0.486 | — | 48 | |
| HGA + CLIP2022.06 | 0.504 | 0.593 | 0.521 | 0.468 | — | |
| HGALanguage Model=GRU2023.10 | 0.504 | 0.593 | 0.521 | — | 0.468 | |
| HGA2022.06 | 0.497 | 0.593 | 0.507 | 0.463 | — | |
| ATP2022.06 | 0.492 | 0.589 | 0.467 | 0.483 | — | |
| HCRNLanguage Model=LSTM, Number of Trainable Parameters=44M2023.10 | 0.482 | 0.537 | 0.493 | — | 0.459 | |
| CLIPMode=single-frame2022.06 | 0.437 | 0.531 | 0.39 | 0.438 | — | |
| Random2022.06 | 0.2 | 0.2 | 0.2 | 0.2 | — |