Automatic Speech Recognition on CGN
-38.6Absolute WER ChangeFull fine-tuning
Evaluation results
| Method | Links | |
|---|---|---|
| Full fine-tuningBase Model=OWSM-0.1B2026.01 | -38.6 | |
| SSVD-OOp=100%, l=256, Base Model=OWSM-0.1B2026.01 | -33.4 | |
| Fine-tuning FF layersBase Model=OWSM-0.1B2026.01 | -32 | |
| SSVDp=100%, Base Model=OWSM-0.1B2026.01 | -30.7 | |
| SSVD-OOp=50%, l=256, Base Model=OWSM-0.1B2026.01 | -30.6 | |
| SSVD-OOp=40%, l=256, Base Model=OWSM-0.1B2026.01 | -29.7 | |
| DoRAr=256, Base Model=OWSM-0.1B2026.01 | -27.7 | |
| LoRAr=256, Base Model=OWSM-0.1B2026.01 | -24.6 |