| Benchmark | Model | Rank | Results |
|---|---|---|---|
| video-question-answering-on-mlvu | S3-prune (LLaVA-VID-7B, 25% vision-token retention) | – | Accuracy: 68.89 |
| video-question-answering-on-video-mme-with-subtitles | S3-prune (LLaVA-VID-7B, 25% vision-token retention) | – | Accuracy: 71.96 |
| zero-shot-video-question-answer-on-egoschema-fullset | S3-prune (LLaVA-OV-7B, 25% vision-token retention) | – | Accuracy: 61.11 |