Video Question Answering on VideoMME (Efficiency and Latency Metrics)
73ScoreKeye-VL-1.5
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Keye-VL-1.5Model Class=8B2026.05 | 73 | — | — | — | |
| LLaVA-OV-2Model Class=8B2026.05 | 71.9 | — | — | — | |
| Qwen3-VLModel Class=8B2026.05 | 71.4 | — | — | — | |
| InternVL-3.5Model Class=8B2026.05 | 65.9 | — | — | — | |
| LLaVA-OV-1.5Model Class=8B2026.05 | 61.1 | — | — | — | |
| PLMModel Class=8B2026.05 | 60.5 | — | — | — | |
| EchoPruneSettings=F160/20%, Base Model=LLaVA-OV-7B2026.05 | 60.4 | 42.8 | 491.3 | 2,122.1 | |
| UniCompSettings=F160/20%, Base Model=LLaVA-OV-7B2026.05 | 60.2 | 511.2 | 818.8 | 2,449.6 | |
| HoliTomSettings=F160/20%, Base Model=LLaVA-OV-7B2026.05 | 60.1 | 830.4 | 1,423 | 3,053.8 | |
| FlashVIDSettings=F160/20%, Base Model=LLaVA-OV-7B2026.05 | 60 | 180.3 | 624.4 | 2,255.2 | |
| LLaVA-OV-7BSettings=F64, Base Model=LLaVA-OV-7B2026.05 | 58.8 | — | 917.2 | 1,569.4 | |
| EchoPruneSettings=F64/20%, Base Model=LLaVA-OV-7B2026.05 | 58.7 | 18.2 | 192.9 | 845.1 | |
| LLaVA-OV-7BSettings=F160, Base Model=LLaVA-OV-7B2026.05 | 58.6 | — | 2,760.8 | 4,391.6 | |
| UniCompSettings=F64/20%, Base Model=LLaVA-OV-7B2026.05 | 58.6 | 210.1 | 353.5 | 1,005.7 | |
| FlashVIDSettings=F64/20%, Base Model=LLaVA-OV-7B2026.05 | 58.5 | 92.1 | 258 | 910.2 | |
| HoliTomSettings=F64/20%, Base Model=LLaVA-OV-7B2026.05 | 58.4 | 280.7 | 523.7 | 1,175.9 |