Speech Enhancement on StillSonicSet Reverb+Noise
1.464PESQStoRM
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| StoRMModel Type=Diffusion2026.03 | 1.464 | 0.469 | 8.758 | 3.174 | 3.823 | 2.792 | 0.869 | |
| DiT-FlowModel Type=Flow Matching2026.03 | 1.378 | 0.484 | 7.576 | 3.189 | 3.856 | 2.818 | 0.858 | |
| SGMSEModel Type=Diffusion2026.03 | 1.248 | 0.373 | 7.223 | 3.318 | 3.234 | 2.719 | 0.865 | |
| Noisy (lower bound)Model Type=–2026.03 | 1.124 | 0.45 | 4.604 | 1.788 | 1.543 | 1.376 | 0.831 |