Video Understanding on VideoRefer-BenchQ
78.3Overall AccuracySWIM
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SWIMPrompt types=Text2026.05 | 78.3 | 83.8 | 75 | 66.7 | 93.7 | 80.7 | |
| GAR-8BMethod Category=Region MLLMs, Zero-shot=true, Sampled frames=162025.10 | 72 | 77.2 | 71 | 61.7 | 86.6 | 68.1 | |
| VideoRefer-7BMethod Category=Region MLLMs, Zero-shot=true, in-domain training=true2025.10 | 71.9 | 75.4 | 68.6 | 59.3 | 89.4 | 78.1 | |
| VideoRefer-7BPrompt types=Mask2026.05 | 71.9 | 75.4 | 68.6 | 59.3 | 89.4 | 78.1 | |
| Qwen2.5-VL-7BPrompt types=Text2026.05 | 71.8 | 78 | 69.7 | 58.2 | 79.9 | 73.2 | |
| GPT-4oMethod Category=General MLLMs, Zero-shot=true2025.10 | 71.3 | 62.3 | 74.5 | 66 | 88 | 73.7 | |
| GPT-4oPrompt types=Text2026.05 | 71.3 | 62.3 | 74.5 | 66 | 88 | 73.7 | |
| GAR-1BMethod Category=Region MLLMs, Zero-shot=true, Sampled frames=162025.10 | 69.9 | 75 | 69.9 | 59.7 | 83.2 | 63.7 | |
| Qwen2.5-Omni-7BPrompt types=Text2026.05 | 68.2 | 65 | 67.6 | 54 | 84.3 | 70.3 | |
| LLaVA-OneVision-7BMethod Category=General MLLMs, Zero-shot=true2025.10 | 67.4 | 58.7 | 62.9 | 64.7 | 87.4 | 76.3 | |
| LLaVA-OV-7BPrompt types=Text2026.05 | 67.4 | 58.7 | 62.9 | 64.7 | 87.4 | 76.3 | |
| Qwen2-VL-7BMethod Category=General MLLMs, Zero-shot=true2025.10 | 66 | 62 | 69.6 | 54.9 | 87.3 | 74.6 | |
| Qwen2-VL-7BPrompt types=Text2026.05 | 66 | 62 | 69.6 | 54.9 | 87.3 | 74.6 | |
| GPT-4o-miniPrompt types=Text2026.05 | 65.8 | 57.6 | 67.1 | 56.5 | 85.9 | 75.4 | |
| InternVL2-26BMethod Category=General MLLMs, Zero-shot=true2025.10 | 65 | 58.5 | 63.5 | 53.4 | 88 | 78.9 | |
| InternVL2-26BPrompt types=Text2026.05 | 65 | 58.5 | 63.5 | 53.4 | 88 | 78.9 | |
| LongVA-7BPrompt types=Text2026.05 | 61.8 | 56.2 | 62.5 | 52 | 83.9 | 65.8 | |
| LongVU-7BPrompt types=Text2026.05 | 61 | 47.2 | 61.3 | 57.5 | 85.3 | 65.8 | |
| Ferret-7BMethod Category=Region MLLMs, Zero-shot=true2025.10 | 48.8 | 35.2 | 44.7 | 41.9 | 70.4 | 74.6 | |
| Ferret-7BPrompt types=Point, Box, Mask2026.05 | 48.8 | 35.2 | 44.7 | 41.9 | 70.4 | 74.6 | |
| Osprey-7BMethod Category=Region MLLMs, Zero-shot=true2025.10 | 39.9 | 45.9 | 47.1 | 30 | 48.6 | 23.7 | |
| Osprey-7BPrompt types=Point, Mask2026.05 | 39.9 | 45.9 | 47.1 | 30 | 48.6 | 23.7 |