Video Question Answering on NExT-QA ATPhard
71.2Overall AccuracyVideoHV-Agent
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VideoHV-AgentEvaluation Protocol=Zero-shot2026.03 | 71.2 | — | — | — | |
| LinVT-Qwen2-VLEvaluation Protocol=Supervised2026.03 | 69.1 | — | — | — | |
| VideoChat2Evaluation Protocol=Supervised2026.03 | 68.2 | — | — | — | |
| VideoAgent2Evaluation Protocol=Zero-shot2026.03 | 68.2 | — | — | — | |
| VideoAgentEvaluation Protocol=Zero-shot2024.03 | 58.4 | — | 58.8 | 57.8 | |
| VideoAgentEvaluation Protocol=Zero-shot2026.03 | 58.4 | — | — | — | |
| SeViLAEvaluation Protocol=Zero-shot2026.03 | 50.8 | — | — | — | |
| GFEvaluation Protocol=Supervised2024.03 | 49.3 | — | 50.3 | 48.7 | |
| HGA2022.06 | 44.1 | 51.2 | 45.3 | 43.3 | |
| HGAEvaluation Protocol=Fine-Tuned, Additional features=Uses additional motion features2023.04 | 44.1 | — | 45.3 | 43.3 | |
| Owl-ConZero-shot=true2023.11 | 41.1 | — | — | — | |
| VFCEvaluation Protocol=Fine-Tuned2023.04 | 39.3 | — | 38.3 | 39.9 | |
| End-to-End VNLIZero-shot=true2023.11 | 39 | — | — | — | |
| Temp[ATP]2022.06 | 38.8 | 46.8 | 36.5 | 38.4 | |
| ATPEvaluation Protocol=Supervised2024.03 | 38.8 | — | 36.5 | 38.4 | |
| PAXIONPatcher Training Loss=VTC+DVDM2023.05 | 38.5 | — | 38.1 | 38.8 | |
| Temp[ATP]Evaluation Protocol=Fine-Tuned2023.04 | 37.6 | — | 36.5 | 38.4 | |
| Owl-RandZero-shot=true2023.11 | 37.2 | — | — | — | |
| Owl-BaseZero-shot=true2023.11 | 37.1 | — | — | — | |
| Side-TuningPatcher Training Loss=VTC+DVDM2023.05 | 36.8 | — | 36 | 37.4 | |
| KP-Transformer FTPatcher Training Loss=VTC2023.05 | 33 | — | 33.6 | 32.5 | |
| VFCEvaluation Protocol=Zero-Shot2023.04 | 31.4 | — | 30 | 32.2 | |
| VFCZero-shot=true2023.11 | 31.4 | — | — | — | |
| VFCEvaluation Protocol=Zero-shot2024.03 | 31.4 | — | 30 | 32.2 | |
| KP-Perceiver FTPatcher Training Loss=VTC2023.05 | 30.7 | — | 31.6 | 30.1 | |
| TACTZero-shot=true2023.11 | 27.6 | — | — | — | |
| InternVideo Backbone2023.05 | 27.1 | — | 27.3 | 27 | |
| ImageBindZero-shot=true, Modalities=video-language2023.11 | 25.4 | — | — | — | |
| CLIPZero-shot=true2023.11 | 23.8 | — | — | — | |
| VideoCLIPZero-shot=true2023.11 | 23.4 | — | — | — | |
| CLIPEvaluation Protocol=Zero-Shot2023.04 | 23 | — | 21.8 | 23.8 | |
| ATPEvaluation Protocol=Fine-Tuned2023.04 | 20.8 | — | 22.6 | 19.6 | |
| ATP2022.06 | 20.2 | 23.9 | 22.6 | 19.6 | |
| HiTeAEvaluation Protocol=Supervised2024.03 | — | — | 48.6 | 47.8 | |
| SeViTEvaluation Protocol=Supervised2024.03 | — | — | 46.5 | 43.3 |