Video Question Answering on EventHallusion
70.2Accuracy (Entire)Gemini-1.5-Pro
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Gemini-1.5-ProModel Type=Proprietary2025.01 | 70.2 | 77.7 | 96.1 | 80.2 | |
| GPT-4oModel Type=Proprietary2025.01 | 65.8 | 90.7 | 92.2 | 84.1 | |
| LLaVA-Video-7BModel Type=Open-Source, Model Scale=≤10B2025.01 | 61.4 | 48.7 | 96 | 64 | |
| Tarsier2-7BModel Type=Open-Source, Model Scale=≤10B2025.01 | 60.5 | 93.3 | 95.1 | 84.6 | |
| LLaVA-Video-72BModel Type=Open-Source, Model Scale=>10B2025.01 | 57.9 | 11.9 | 96 | 45.6 | |
| InternVL2.5-78BModel Type=Open-Source, Model Scale=>10B2025.01 | 57.9 | 67.9 | 88.2 | 70.2 | |
| Tarsier-34BModel Type=Open-Source, Model Scale=>10B2025.01 | 49.1 | 92.7 | 69.6 | 74.8 | |
| LLAVA-OV-72BModel Type=Open-Source, Model Scale=>10B2025.01 | 47.4 | 26.9 | 90.1 | 48.3 | |
| LLAVA-OV-7BModel Type=Open-Source, Model Scale=≤10B2025.01 | 46.5 | 67.4 | 86.1 | 66.2 | |
| InternVL2.5-8BModel Type=Open-Source, Model Scale=≤10B2025.01 | 46.5 | 69.2 | 90.2 | 68.2 | |
| Tarsier-7BModel Type=Open-Source, Model Scale=≤10B2025.01 | 43.9 | 82.4 | 79.4 | 70.9 | |
| Qwen2-VL-7BModel Type=Open-Source, Model Scale=≤10B2025.01 | 35.1 | 94.3 | 57.4 | 68.6 | |
| Qwen2-VL-72BModel Type=Open-Source, Model Scale=>10B2025.01 | 33.3 | 77.7 | 56.4 | 60 |