Audio-language Question Answering on ST-AudioQA Type B
14.3mAPST-AudioLM
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ST-AudioLMTraining=QA-trained2026.06 | 14.3 | 83.8 | 54.2 | 32.8 | 71.6 | 70 | 66.2 | 49.7 | 47.4 | 59.3 | |
| SpatialAST-FOA + OLMo2Training=QA-trained2026.06 | 12.6 | 83.2 | 51 | 30.9 | 65.2 | 68.8 | 73.9 | 48.8 | 40.2 | 70 | |
| PSELDNets-mACCDOA + OLMo2Training=QA-trained2026.06 | 5.4 | 76.1 | 51.2 | 28.2 | 50.5 | 58.3 | 56.1 | 46.1 | 32.8 | 30.1 | |
| Qwen2-AudioTraining=Zero-shot2026.06 | 4.7 | 69.1 | 20.5 | 0 | 48.8 | 36.3 | 36.9 | 3.6 | 19.4 | 4.7 | |
| BATTraining=QA-trained2026.06 | 1.8 | 60.5 | 35.6 | 21 | 49.8 | 50.4 | 64.7 | 39.4 | 37.4 | 57 | |
| Random chance2026.06 | 1.5 | 50 | 22.9 | 10 | 50 | 50 | 50 | 25 | 25 | 25 |