Audio Question Answering on ChimeHome (test)
96.7Token-Level AccuracyIndividual
Evaluation results
| Method | Links | |
|---|---|---|
| IndividualTraining Strategy=Full-Data Training, T.E.=~7d2026.05 | 96.7 | |
| GST-T2Training Strategy=Full-Data Training, T.E.=~2d2026.05 | 96.2 | |
| GST-G3Training Strategy=Full-Data Training, T.E.=~2d2026.05 | 96 | |
| GST-G3Training Strategy=Low-Resource Finetuning, T.E.=~1d2026.05 | 95.2 | |
| GST-G2Training Strategy=Full-Data Training, T.E.=~2d2026.05 | 95.1 | |
| Mix AllTraining Strategy=Full-Data Training, T.E.=~4d2026.05 | 94.9 | |
| GST-T2Training Strategy=Low-Resource Finetuning, T.E.=~0.5d2026.05 | 94.7 | |
| GST-G2Training Strategy=Low-Resource Finetuning, T.E.=~0.5d2026.05 | 93.6 | |
| Mix AllTraining Strategy=Low-Resource Finetuning, T.E.=~0.5d2026.05 | 93.1 | |
| SALMONNTraining Strategy=Full-Data Training, T.E.=N/A2026.05 | 39.4 | |
| SequentialTraining Strategy=Full-Data Training, T.E.=~7d2026.05 | 35.2 |