Automatic Speech Recognition on OGI-S (Age 11-15)
8.34WERSingle Dataset Fine-tune
Evaluation Results
| Method | Links | |
|---|---|---|
| Single Dataset Fine-tuneFine-tuning Protocol=Encoder+Projector+LoRA2026.06 | 8.34 | |
| C-DR MoERouting Strategy=Hard Routing, Classifier Type=Ground-Truth Routing2026.06 | 8.62 | |
| C-DR MoERouting Strategy=Soft Routing, Classifier Type=Weighted-layer Classifier (Coarse-to-Fine), Entropy-Aware Routing (EAR)=true2026.06 | 8.62 | |
| C-DR MoERouting Strategy=Hard Routing, Classifier Type=Weighted-layer Classifier (Coarse-to-Fine), Entropy-Aware Routing (EAR)=true2026.06 | 8.63 | |
| C-DR MoERouting Strategy=Hard Routing, Classifier Type=Weighted-layer Classifier (Coarse-to-Fine)2026.06 | 8.66 | |
| C-DR MoERouting Strategy=Hard Routing, Classifier Type=Weighted-layer Classifier (Single-Stage)2026.06 | 8.7 | |
| C-DR MoERouting Strategy=Hard Routing, Classifier Type=Top-layer Classifier (Single-Stage)2026.06 | 8.72 | |
| C-DR MoERouting Strategy=Soft Routing, Classifier Type=Weighted-layer Classifier (Coarse-to-Fine)2026.06 | 8.72 | |
| Vanilla-Routing MoEFine-tuning Protocol=Pretrain + Gate-only2026.06 | 9.26 | |
| Single Dataset Fine-tuneFine-tuning Protocol=Projector+LoRA2026.06 | 9.29 | |
| Vanilla-Routing MoEFine-tuning Protocol=Joint2026.06 | 10.37 | |
| Vanilla-Routing MoEFine-tuning Protocol=Pretrain + Joint2026.06 | 10.53 | |
| Single-ExpertArchitecture=One projector, one LoRA2026.06 | 11.07 | |
| Canary-QwenFine-tuning Protocol=Zero-shot2026.06 | 13.34 |