Long Video Understanding on LongVideoBench (LongVB)
69.2AccuracyGemini 2.5 Pro
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini 2.5 ProEvaluation frame rate=1 FPS, self-collected via API=true2025.05 | 69.2 | |
| GPT4-O2024.12 | 66.7 | |
| Seed1.5-VLThinking mode=true, Evaluation frame rate=1 FPS2025.05 | 64.6 | |
| Gemini 1.5 Pro2024.12 | 64.4 | |
| Seed1.5-VLThinking mode=false, Evaluation frame rate=1 FPS2025.05 | 64 | |
| GPT4-V2024.12 | 60.7 | |
| LinVT-InternVL2Size=8B2024.12 | 57.4 | |
| LinVT-Qwen2-VLSize=7B2024.12 | 57.2 | |
| LinVT-MolmoSize=7B2024.12 | 56.8 | |
| LinVT-BLIP-3Size=4B2024.12 | 56.6 | |
| OryxSize=7B2024.12 | 55.3 | |
| KangarooSize=8B2024.12 | 54.8 | |
| PLLAVASize=34B2024.12 | 53.5 | |
| LinVT-AquilaSize=2B2024.12 | 53.5 | |
| LinVT-MiphaSize=1.6B2024.12 | 49.7 | |
| ShareGPT4VideoSize=8B2024.12 | 41.8 | |
| Video-LLaVASize=7B2024.12 | 37.6 |