High-resolution & Long Video Understanding on HLVid (test)
52.6ScoreNVILA-8B-Video + AutoGaze
Evaluation results
| Method | Links | |
|---|---|---|
| NVILA-8B-Video + AutoGazeMax #F=1024, Max Res.=35842026.03 | 52.6 | |
| Gemini 2.5 Flash-Lite2026.03 | 52.2 | |
| GPT-4o2026.03 | 49.3 | |
| Qwen2.5-VL-7BMax #F=48, Max Res.=8962026.03 | 48.1 | |
| VideoChat-FlashMax #F=10000, Max Res.=4482026.03 | 46.6 | |
| NVILA-8B-VideoMax #F=256, Max Res.=4482026.03 | 42.5 | |
| LongVILA-R1-7BMax #F=8192, Max Res.=4482026.03 | 42.2 | |
| LongVILA-7BMax #F=2048, Max Res.=3842026.03 | 41.4 | |
| LLaVA-OV-8BMax #F=32, Max Res.=3842026.03 | 41.1 | |
| InternVL3.5-8BMax #F=64, Max Res.=4482026.03 | 39.9 | |
| VideoLLaMA3-7BMax #F=180, Max Res.=3842026.03 | 38.8 |