Video Question Answering on NExT-QA Hard Split (val)
51.8Causal AccuracyCoVGT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoVGT2023.02 | 51.8 | 50.5 | |
| HiTeA# PT Data=5M2022.12 | 47.8 | 48.6 | |
| HGA2022.12 | 43.3 | 45.3 | |
| HGA2023.01 | 43.3 | 45.3 | |
| SeViT_FiDFrame Sampling=Frame Retrieval2023.01 | 43.3 | 46.5 | |
| HGA2023.02 | 43.3 | 45.3 | |
| SeViT_MARFrame Sampling=Frame Retrieval2023.01 | 43.2 | 46.3 | |
| SeViT_FiDFrame Sampling=Uniform Sampling2023.01 | 42.7 | 46.4 | |
| SeViT_MARFrame Sampling=Uniform Sampling2023.01 | 41.9 | 44.7 | |
| ATP# PT Data=400M2022.12 | 38.4 | 36.5 | |
| Temp[APT]2023.01 | 38.4 | 36.5 | |
| Temporal [ATP]2023.02 | 38.4 | 36.5 | |
| APT2023.01 | 19.6 | 22.6 | |
| ATP2023.02 | 19.6 | 22.6 |