Automatic Speech Recognition on CORAAL (test)
9.25WERGumbel-BEARD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gumbel-BEARDBackbone=Whisper-medium2026.06 | 9.25 | — | |
| LoRAWhisper model size=large-v32026.06 | 9.7 | — | |
| GC-LoRAWhisper model size=large-v32026.06 | 9.7 | — | |
| Full FTParams=764M, Backbone=Whisper-Medium2026.06 | 9.8 | — | |
| Baseline (SFT)Backbone=Whisper-medium2026.06 | 9.81 | — | |
| GC-LoRAWhisper model size=medium2026.06 | 9.9 | — | |
| GC-LoRAParams=447k, Backbone=Whisper-Medium2026.06 | 9.9 | — | |
| LoRAWhisper model size=medium2026.06 | 10.1 | — | |
| LoRAParams=829k, Backbone=Whisper-Medium2026.06 | 10.1 | — | |
| Gumbel-BEARDBackbone=Whisper-small2026.06 | 11.01 | — | |
| Baseline (SFT)Backbone=Whisper-small2026.06 | 11.7 | — | |
| GC-LoRAWhisper model size=small2026.06 | 12.3 | — | |
| LoRAWhisper model size=small2026.06 | 12.7 | — | |
| Zero-shotParams=0, Backbone=Whisper-Medium2026.06 | 17 | — | |
| Zero-shotBackbone=Whisper-medium2026.06 | 17.09 | — | |
| Zero-shotBackbone=Whisper-small2026.06 | 18.69 | — | |
| GC-LoRAWhisper model size=base2026.06 | 20.1 | — | |
| LoRAWhisper model size=base2026.06 | 20.3 | — | |
| ClozeGERBackbone=SpeechGPT, Source Speech=true, LoRA=true, Logits Calibration=true, Post-processing=true2024.05 | 26.7 | 8.6 | |
| ClozeGERBackbone=LLaMA-2-7b, Source Speech=false, LoRA=true, Logits Calibration=true, Post-processing=true2024.05 | 26.8 | 8.2 | |
| GC-LoRAWhisper model size=tiny2026.06 | 29 | — | |
| Whisper Baseline2024.05 | 29.2 | — | |
| LoRAWhisper model size=tiny2026.06 | 29.4 | — |