Multiple-choice Video Question Answering on TVQA
82AccuracyFrozenBiLM
Evaluation Results
| Method | Links | |
|---|---|---|
| FrozenBiLMSupervision=100% (fully-supervised)2022.06 | 82 | |
| FrozenBiLMSupervision=10% (few-shot)2022.06 | 77.6 | |
| FrozenBiLMSupervision=1% (few-shot)2022.06 | 72.5 | |
| FrozenBiLMTraining Data=WebVid10M, Speech=true, Zero-shot=true2022.06 | 59.7 | |
| FrozenBiLMSupervision=0% (zero-shot)2022.06 | 59.7 | |
| MiniGPT4-VideoBackbone=Mistral-7B, Using Subtitles=true, Zero-shot=true2024.04 | 54.21 | |
| MiniGPT4-VideoBackbone=Llama 2-7B, Using Subtitles=true, Zero-shot=true2024.04 | 46.3 | |
| MiniGPT4-VideoBackbone=Llama 2-7B, Using Subtitles=false, Zero-shot=true2024.04 | 36.45 | |
| MiniGPT4-VideoBackbone=Mistral-7B, Using Subtitles=false, Zero-shot=true2024.04 | 33.9 | |
| FrozenBiLMUsing Subtitles=false, Zero-shot=true2024.04 | 29.7 | |
| FrozenBiLMTraining Data=WebVid10M, Speech=false, Zero-shot=true2022.06 | 29.7 | |
| CLIP ViT-L/14Training Data=400M image-texts, Speech=false, Zero-shot=true2022.06 | 26.1 | |
| Video-ChatGPTUsing Subtitles=false, Zero-shot=true2024.04 | 23.35 | |
| RandomZero-shot=true2022.06 | 20 |