Multi-turn Dialogue Speech Evaluation on UniSRM-BENCH T4
88.89AccuracyUniSRM
Evaluation Results
| Method | Links | |
|---|---|---|
| UniSRMModel Category=Proposed Method2026.05 | 88.89 | |
| WERType=Objective Metric2026.05 | 84.1 | |
| Gemini-2.5-ProModel Category=Proprietary2026.05 | 82.4 | |
| GPT-4o-AudioModel Category=Proprietary2026.05 | 71.96 | |
| Gemini-2.5-FlashModel Category=Proprietary2026.05 | 71.43 | |
| Kimi-Audio-7BModel Category=Open-Source2026.05 | 64.29 | |
| MiMo-Audio-7BModel Category=Open-Source2026.05 | 59.52 | |
| Qwen2.5-Omni-7BModel Category=Open-Source2026.05 | 56.35 | |
| DNSMOSType=Objective Metric2026.05 | 50.79 | |
| UTMOSType=Objective Metric2026.05 | 40.48 |