Confidence Level Classification on Confidence Level Dataset (5-fold cross-validation)
74.4F1 Score (Low Bin)Whisper-Base (Proposed)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Whisper-Base (Proposed)Architecture=Whisper-Base2026.05 | 74.4 | 67.2 | 83.6 | 75.1 | |
| WavLMArchitecture=WavLM2026.05 | 72.6 | 67.2 | 81.4 | 73.7 | |
| Whisper-TinyArchitecture=Whisper-Tiny2026.05 | 71.7 | 64.8 | 81.8 | 72.8 | |
| HuBERTArchitecture=HuBERT2026.05 | 70.4 | 63.6 | 80.6 | 71.5 | |
| Wynn et al, 2025Architecture=Co-attention based hybrid2026.05 | 62.8 | 58.2 | 73.2 | 64.7 | |
| Wav2Vec 2.0Architecture=Wav2Vec 2.02026.05 | 60.4 | 59.2 | 78.8 | 66.1 |