Contextual Automatic Speech Recognition on Spoken Darwin-Science (test)
4.5WERWhisper-large-v3
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Whisper-large-v3CoT=false, Anchor=-2026.05 | 4.5 | 8.35 | 30.56 | 16.86 | |
| Qwen3-ASR-1.7BCoT=false, Anchor=-2026.05 | 5.6 | 10.55 | 30.93 | 18.37 | |
| Fun-Audio-Chat-8B + LaSRCoT=true, Anchor=- 0.50s2026.05 | 6.09 | 10.79 | 25 | 16.23 | |
| Fun-Audio-Chat-8B + LaSRCoT=true, Anchor=Random2026.05 | 6.17 | 11.1 | 25.13 | 16.42 | |
| Fun-Audio-Chat-8BCoT=false, Anchor=-2026.05 | 6.25 | 12.68 | 32.74 | 20.37 | |
| Qwen3-ASR-0.6BCoT=false, Anchor=-2026.05 | 6.27 | 18.5 | 42.93 | 27.84 | |
| Fun-Audio-Chat-8B + SFTCoT=false, Anchor=-2026.05 | 6.9 | 13.23 | 26.64 | 18.32 | |
| Fun-Audio-Chat-8B + LaSRCoT=true, Anchor=+ 0.15s2026.05 | 6.97 | 12.05 | 26.64 | 17.64 | |
| Fun-Audio-Chat-8B + SFTCoT=true, Anchor=-2026.05 | 7.93 | 18.83 | 28.08 | 22.39 |