Multi-discipline Long Video Understanding on MLVU (Context Length)
61,140Context Length (tokens)Video-o3
Evaluation Results
| Method | Links | |
|---|---|---|
| Video-o3Max retrieval / injection=82026.05 | 61,140 | |
| VideoChat-R1.5-7BMax retrieval / injection=22026.05 | 49,872.9 | |
| LongVT-RFTMax retrieval / injection=52026.05 | 49,606.9 | |
| VideoZoomerMax retrieval / injection=42026.05 | 17,518.3 | |
| LOVE-R1-stage3Max retrieval / injection=22026.05 | 16,919.5 | |
| DynFrame-8BMax retrieval / injection=12026.05 | 7,964.2 | |
| Video-R1-7BMax retrieval / injection=02026.05 | 3,968.4 | |
| Qwen2.5-VL-7BMax retrieval / injection=02026.05 | 1,991.5 |