Audio-language Question Answering on ST-AudioQA Type A
27.6mAPST-AudioLM
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| ST-AudioLMTraining=QA-trained2026.06 | 27.6 | 93.7 | 81.6 | 51.5 | 91.1 | 87.8 | 99.8 | |
| SpatialAST-FOA + OLMo2Training=QA-trained2026.06 | 25.3 | 93.3 | 71.4 | 44.9 | 78.4 | 82.4 | 99.6 | |
| PSELDNets-mACCDOA + OLMo2Training=QA-trained2026.06 | 9.1 | 89.9 | 72 | 39.7 | 51.3 | 62.6 | 73.4 | |
| Qwen2-AudioTraining=Zero-shot2026.06 | 8.8 | 76.3 | 15.9 | 0 | 51.2 | 34.3 | 37.2 | |
| BATTraining=QA-trained2026.06 | 1 | 64.6 | 48.5 | 29.8 | 49.7 | 52.8 | 99.4 | |
| Random chance2026.06 | 0.8 | 50 | 22.9 | 10 | 50 | 50 | 50 |