Transcription on iEEG clinical dataset Background
19.4WERQwen2-Audio
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen2-AudioSelection=Oracle, Fine-tuned=true2025.02 | 19.4 | |
| AAD-LLMAttention Strategy=brain-decoded, Training Data=clinical-15m2025.02 | 22.5 | |
| Qwen2-AudioSelection=Extracted, Fine-tuned=true2025.02 | 24.6 | |
| AAD-LLMAttention Strategy=oracle2025.02 | 33.9 | |
| AAD-LLMAttention Strategy=brain-decoded2025.02 | 35.2 | |
| Qwen2-AudioSelection=Extracted, Fine-tuned=false2025.02 | 73.2 | |
| Qwen2-AudioSelection=Random, Fine-tuned=true2025.02 | 74.6 | |
| Qwen2-AudioSelection=Random, Fine-tuned=false2025.02 | 106.9 | |
| Qwen-AudioIntention=None2025.02 | 112.8 | |
| Qwen2-AudioIntention=None2025.02 | 124.7 | |
| WavLLMIntention=None2025.02 | 128.3 | |
| LTU-ASIntention=None2025.02 | 172.6 | |
| SALMONNIntention=None2025.02 | 225.3 |