Video Question Answering on StreamGaze
70NFI ScoreHuman
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Human2026.03 | 70 | 82.7 | 88.9 | 90.3 | 70.7 | 96 | 92 | 80 | 84 | 76.5 | 78 | |
| GazeQwenParams=7B, Frames=Adapt.2026.03 | 65.7 | 63.9 | 64.4 | 54.8 | 60.9 | 59.4 | 76.6 | 80.9 | 40.8 | 60.3 | 74.8 | |
| GPT-4oFrames=162026.03 | 60.1 | 53.4 | 44.9 | 53.5 | 58 | 72.9 | 73 | 59.6 | 37 | 59.7 | 14.9 | |
| Qwen2.5-VLParams=7B, Frames=Adapt.2026.03 | 51.8 | 47.8 | 35 | 45 | 48.3 | 59 | 55.8 | 54.8 | 39.1 | 48.6 | 40.7 | |
| Claude Sonnet4Frames=162026.03 | 50 | 47.4 | 55.4 | 42.5 | 32.5 | 52.1 | 53.3 | 56.1 | 43.9 | 53.5 | 35 | |
| InternVL3.5Params=8B, Frames=Adapt.2026.03 | 49 | 44.4 | 31.1 | 57.3 | 54.8 | 62.7 | 62.8 | 46.6 | 37.2 | 37.3 | 5.1 | |
| VITA 1.5Params=7B, Frames=162026.03 | 47.4 | 38.4 | 36.5 | 34.6 | 37.8 | 45.5 | 39.6 | 43.7 | 37 | 35.1 | 26.7 | |
| ViSpeakParams=7B, Frames=1 fps2026.03 | 46.3 | 46.7 | 35.8 | 41.7 | 47.3 | 57.2 | 58.1 | 40.6 | 30.9 | 63.5 | 45.8 | |
| MiniCPM-VParams=8B, Frames=322026.03 | 43 | 36.5 | 37.4 | 35.4 | 29.6 | 33.4 | 37.9 | 43.8 | 34.5 | 48 | 21.6 | |
| Claude Opus4Frames=162026.03 | 37.2 | 39.9 | 39.2 | 46 | 16.6 | 43.1 | 43.6 | 43 | 35.1 | 46.6 | 49 | |
| DispiderParams=7B, Frames=1 fps2026.03 | 36.6 | 32.4 | 36.5 | 38.1 | 26.3 | 33.6 | 33.8 | 35.3 | 32.1 | 25.2 | 26.1 | |
| KangarooParams=7B, Frames=642026.03 | 36.3 | 35.1 | 36.5 | 31.9 | 27.5 | 45.4 | 48.4 | 40.2 | 41.2 | 24.2 | 19.8 | |
| AssistGazeParams=26M, Frames=322026.03 | 29.4 | 24 | 13.1 | 31 | 29.4 | 27.8 | 25 | 10.9 | 25.4 | — | — | |
| Flash-VStreamParams=7B, Frames=1 fps2026.03 | 24.9 | 24.3 | 20.2 | 33.6 | 22 | 28.9 | 14.7 | 4.4 | 28 | 44.3 | 21.7 | |
| VideoLLM-onlineParams=8B, Frames=2 fps2026.03 | 0 | 8.1 | 0 | 0 | 0 | 0.6 | 0.6 | 0.2 | 0 | 45.8 | 33.3 |