Automatic Speech Recognition on HeartBeats-ASR-Bench
0.1873WER (English, 200)HeartTranscriptor
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HeartTranscriptor2026.01 | 0.1873 | 0.1077 | 0.1042 | 0.1801 | 0.2151 | |
| SongPrep2026.01 | 0.2075 | 0.167 | — | — | — | |
| Whisper-Large-V3size=large, version=v32026.01 | 0.2139 | 0.3316 | 0.1646 | 0.2146 | 0.2798 | |
| Whisper-Mediumsize=medium2026.01 | 0.2206 | 0.3045 | 0.2789 | 0.3405 | 0.352 | |
| Qwen3-Omni-30B-A3B-Captionerparameters=30B, hallucination filtering=error rate > 3 removed2026.01 | 0.261 | 0.2049 | 0.2394 | 0.2804 | 0.232 | |
| FireRedASR-llminput duration constraint=30s2026.01 | 0.3266 | 0.1679 | — | — | — | |
| Whisper-Turbomodel_variant=Turbo2026.01 | 0.3397 | 0.302 | 0.2672 | 0.263 | 0.3328 | |
| Whisper-Smallsize=small2026.01 | 0.4172 | 0.3473 | 0.4612 | 0.5556 | 0.4391 | |
| FireRedASR-aedinput duration constraint=60s2026.01 | 0.5176 | 0.3471 | — | — | — |