Multiple-choice Video Question Answering on How2QA
86.7AccuracyFrozenBiLM
Evaluation results
| Method | Links | |
|---|---|---|
| FrozenBiLMSupervision=100% (fully-supervised)2022.06 | 86.7 | |
| FrozenBiLMSupervision=10% (few-shot)2022.06 | 75.8 | |
| FrozenBiLMSupervision=1% (few-shot)2022.06 | 71.7 | |
| FrozenBiLMTraining Data=WebVid10M, Speech=true, Zero-shot=true2022.06 | 58.4 | |
| FrozenBiLMSupervision=0% (zero-shot)2022.06 | 58.4 | |
| FrozenBiLMTraining Data=WebVid10M, Zero-shot evaluation=true2022.06 | 57.9 | |
| Just AskTraining Data=HowToVQA69M + WebVidVQA3M, Speech=false, Zero-shot=true2022.06 | 53.1 | |
| Just AskTraining Data=HowToVQA69M + WebVidVQA3M, Zero-shot evaluation=true2022.06 | 53.1 | |
| CLIP ViT-L/14Training Data=400M image-texts, Speech=false, Zero-shot=true2022.06 | 47.7 | |
| CLIP ViT-L/14Training Data=400M image-texts, Zero-shot evaluation=true2022.06 | 47.7 | |
| FrozenBiLMTraining Data=WebVid10M, Speech=false, Zero-shot=true2022.06 | 41.9 | |
| RandomZero-shot=true2022.06 | 25 | |
| RandomTraining Data=—, Zero-shot evaluation=true2022.06 | 25 |