Emphasis Generation on EmphAssess (test)
66.95PrecisionGPT-4O-AUDIO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-4O-AUDIOGroup=Topline2026.07 | 66.95 | 63.19 | 65.02 | |
| SPEECHCOMBINEGroup=Group B2026.07 | 25.91 | 39.9 | 31.42 | |
| GLM-4-VOICEGroup=Group B2026.07 | 24.67 | 28.75 | 26.55 | |
| FUN-AUDIO-CHATGroup=Group B2026.07 | 23.01 | 22.82 | 22.91 | |
| STEP-AUDIO2-MINI-THINKGroup=Group B2026.07 | 22.75 | 23.11 | 22.93 | |
| STEP-AUDIO2-MINIGroup=Group B2026.07 | 22.48 | 23.23 | 22.85 | |
| QWEN-2.5-OMNIGroup=Group B2026.07 | 18.82 | 34.03 | 24.24 | |
| KIMI-AUDIOGroup=Group B2026.07 | 18.21 | 26.81 | 21.69 | |
| CONT. PRE-TRAIN + SFTGroup=Group A2026.07 | 16.76 | 26.95 | 20.67 | |
| OSUM-ECHATGroup=Group B2026.07 | 13.87 | 19.61 | 16.25 | |
| ASR + TEXT LLMGroup=Group A, Generation engine=Kokoro text-to-speech2026.07 | 11.35 | 29.71 | 16.42 | |
| CONT. PRE-TRAINGroup=Group A2026.07 | 6.9 | 17.5 | 9.89 |