Human-likeness classification on S2S Turing (test)
96.48Accuracy (H-H)GPT-4o-Audio-Preview
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-4o-Audio-Preview2026.02 | 96.48 | 27.08 | 0.69 | 41.16 | |
| Baichuan-Audio2026.02 | 81.69 | 15.28 | 12.5 | 36.28 | |
| Qwen2.5-Omni2026.02 | 78.17 | 23.61 | 23.61 | 41.63 | |
| Phi-4-Multimodal2026.02 | 77.46 | 14.58 | 22.22 | 37.91 | |
| Human Judgement2026.02 | 70.28 | 83.57 | 63.84 | 72.84 | |
| MiniCPM-o 2.62026.02 | 67.61 | 43.06 | 29.86 | 46.74 | |
| Gemini 2.5 pro2026.02 | 57.75 | 72.92 | 57.64 | 62.79 | |
| Voxtral Mini2026.02 | 51.41 | 50.69 | 38.89 | 46.98 | |
| Gemma 3n2026.02 | 46.48 | 44.44 | 40.28 | 43.72 | |
| Seallms-Audio2026.02 | 11.27 | 84.72 | 72.92 | 56.51 |