Description on iEEG clinical dataset Foreground
91.7AVG(G, P, T)Qwen2-Audio
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen2-AudioSelection=Oracle, Fine-tuned=true2025.02 | 91.7 | |
| AAD-LLMAttention Strategy=oracle2025.02 | 89.9 | |
| AAD-LLMAttention Strategy=brain-decoded2025.02 | 89.3 | |
| AAD-LLMAttention Strategy=brain-decoded, Training Data=clinical-15m2025.02 | 89.2 | |
| Qwen2-AudioSelection=Extracted, Fine-tuned=true2025.02 | 88.1 | |
| Qwen2-AudioSelection=Random, Fine-tuned=true2025.02 | 69.3 | |
| Qwen2-AudioSelection=Extracted, Fine-tuned=false2025.02 | 56.2 | |
| SALMONNIntention=None2025.02 | 55.3 | |
| Qwen2-AudioIntention=None2025.02 | 50.9 | |
| Qwen2-AudioSelection=Random, Fine-tuned=false2025.02 | 49.9 | |
| LTU-ASIntention=None2025.02 | 48.9 | |
| GAMAIntention=None2025.02 | 45.9 | |
| WavLLMIntention=None2025.02 | 41.7 | |
| Qwen-AudioIntention=None2025.02 | 36.1 |