Video Understanding on LSDBench (test)
56.2AccuracyGemini-2.0-Flash
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-2.0-FlashFrames=2700, Visual Token=696.6k2026.03 | 56.2 | |
| Qwen2.5-VLFrames=768, Visual Token=499.2k2026.03 | 52.5 | |
| Qwen2.5-VL(RHS)Frames=225, Visual Token=146.2k2026.03 | 52.2 | |
| EVAFrames=76.9, Visual Token=10.3k2026.03 | 51 | |
| Qwen2.5-VLFrames=256, Visual Token=166.4k2026.03 | 50.1 | |
| LongVilaFrames=256, Visual Token=65.5k2026.03 | 49.8 | |
| Qwen2.5-VL*Frames=32, Visual Token=21.0k2026.03 | 49.2 | |
| Qwen2-VLFrames=256, Visual Token=166.4k2026.03 | 48 | |
| Qwen2.5-VL*Frames=16, Visual Token=10.5k2026.03 | 47.7 | |
| LongVAFrames=512, Visual Token=73.7k2026.03 | 33 | |
| LongVAFrames=1024, Visual Token=147.5k2026.03 | 32.5 | |
| LongVAFrames=256, Visual Token=36.9k2026.03 | 31.3 |