Audio-Visual Question Answering on UGC-AVQA
69.1Average ScoreGemini-3.1-pro
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Gemini-3.1-proModel Type=Proprietary Models (API)2026.05 | 69.1 | 68.7 | 71.4 | 68.5 | 68 | |
| Gemini-3.1-flash-liteModel Type=Proprietary Models (API)2026.05 | 67.1 | 68.5 | 68 | 64.3 | 67.5 | |
| Qwen3-Omni-30B-A3BModel Type=Open-source Models2026.05 | 65 | 65.3 | 66.8 | 60.9 | 67 | |
| Gemma-4-E4B-itModel Type=Open-source Models2026.05 | 59.7 | 58.5 | 61.7 | 55.6 | 63.1 | |
| Gemma-3n-E4B-itModel Type=Open-source Models2026.05 | 54.6 | 58 | 57 | 47.6 | 55.6 | |
| Qwen2.5-Omni-7BModel Type=Open-source Models2026.05 | 54.2 | 52.7 | 54.4 | 51.2 | 58.7 | |
| Gemma-3n-E2B-itModel Type=Open-source Models2026.05 | 53 | 55.8 | 55.6 | 46.4 | 54.4 | |
| Qwen2.5-Omni-3BModel Type=Open-source Models2026.05 | 49 | 44.7 | 52.4 | 48.3 | 50.5 | |
| Gemma-4-E2B-itModel Type=Open-source Models2026.05 | 45.6 | 46.1 | 45.2 | 42 | 49 |