Human Preference Evaluation on TPI (test)
66.05Preferred RateTPI-Full
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TPI-FullEvaluation Protocol=Independent Quality Preference2026.04 | 66.05 | 7.63 | 26.32 | |
| TPI-Corpus (GT)Evaluation Protocol=Independent Quality Preference2026.04 | 64.75 | 5.5 | 29.75 | |
| TPI-FullEvaluation Protocol=Direct Model Comparison (A/B Preference Test), Baseline=Qwen2.5-Omni-7B2026.04 | 60.5 | 11.5 | 28 |