Video Question Answering on MLVU (p@1, p@4, m@4)
67.9p@1VSeek-VETL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VSeek-VETLreward_type=VETL2026.07 | 67.9 | 77.6 | 68.8 | |
| VSeek-EMreward_type=exact-match2026.07 | 66.7 | 75.9 | 67.3 | |
| Qwen3-VL-4B-Instructparameters=4B2026.07 | 61.9 | 66.2 | 62.4 | |
| Qwen3-VL-8B-Instructparameters=8B2026.07 | 61.9 | 66.2 | 62.4 | |
| GPT-5.2frames=642026.07 | 61.3 | 67.1 | 61.5 | |
| Qwen3-VL-4B-Thinkingparameters=4B2026.07 | 58 | 64.3 | 58.8 | |
| VSeek-GPTbase_model=GPT-5.22026.07 | 56.3 | 75.4 | 59.9 | |
| Qwen3-VL-8B-Thinkingparameters=8B2026.07 | 56.3 | 71.5 | 59.4 | |
| Naive RAG2026.07 | 55.6 | 64 | 57.5 | |
| VideoTree2026.07 | 46.1 | 48.2 | 46.8 | |
| InternVL-3.5-4Bparameters=4B2026.07 | 41.8 | 69.1 | 43.4 |