Video Question Answering on KPM-QA (test)
94.05Accuracy (All)KPM
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| KPMUtilized-LLM=Qwen2.5, Size=7B2026.02 | 94.05 | 93.23 | 89.78 | 80.7 | 98.27 | 97.09 | |
| KPM (w/o MoPE)Utilized-LLM=Qwen2.5, Size=7B2026.02 | 94.03 | 93.2 | 89.78 | 80.7 | 98.27 | 97.05 | |
| InternVideo-2.5Utilized-LLM=InternLM2.5, Size=8B2026.02 | 85.95 | 83.7 | 75.74 | 77.42 | 92.21 | 94.22 | |
| GPT-4.12026.02 | 84.41 | 75.34 | 77.62 | 72.88 | 88.13 | 95.79 | |
| Gemini-2.5Pro2026.02 | 81.55 | 77.31 | 76.22 | 71.08 | 87.93 | 95.19 | |
| LLaVA-OneVisionUtilized-LLM=QWen2.5, Size=7B2026.02 | 79.09 | 73.41 | 75.75 | 68.42 | 86.84 | 93.91 | |
| Qwen2-VLUtilized-LLM=Qwen2, Size=7B2026.02 | 76.39 | 75.92 | 58.19 | 61.4 | 85.58 | 81.98 | |
| Qwen2.5-VL (w MoPE)Utilized-LLM=Qwen2.5, Size=7B2026.02 | 75.04 | 73.63 | 59.23 | 31.58 | 82.34 | 87.69 | |
| Qwen2.5-VLUtilized-LLM=Qwen2.5, Size=7B2026.02 | 69.34 | 65.43 | 56.56 | 24.56 | 86.15 | 81.61 | |
| VideoLLaMA3Utilized-LLM=QWen2.5, Size=7B2026.02 | 58.69 | 57.25 | 47.85 | 54.39 | 83.85 | 61.26 |