Audio-Visual Question Answering on StreamingBench Audio-Visual QA 1.0 (test)
40.4Error Rate (ER)ROMA-7B
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ROMA-7BAudio=true2026.05 | 40.4 | 34.8 | 50.4 | 58.8 | 46.1 | |
| Qwen2.5-VL-7BAudio=false2026.05 | 43.6 | 23.2 | 45.2 | 54 | 41.5 | |
| Qwen3-VL-30B-A3BAudio=false2026.05 | 44.8 | 30.4 | 46.4 | 62.4 | 46 | |
| Qwen2.5-Omni-7BAudio=true2026.05 | 48 | 22.4 | 50 | 70 | 49.6 | |
| Qwen3-Omni-30B-A3BAudio=true2026.05 | 53.6 | 40.8 | 70.4 | 79.2 | 61 | |
| StreamOVAudio=true2026.05 | 63.6 | 59.2 | 79.2 | 90.4 | 68.6 |