Utterance-level Pairwise Preference Judgement on UniSRM-BENCH T1
65.06AccuracyUniSRM
Evaluation Results
| Method | Links | |
|---|---|---|
| UniSRMModel Category=Proposed Method2026.05 | 65.06 | |
| GPT-4o-AudioModel Category=Proprietary2026.05 | 61.04 | |
| Gemini-2.5-ProModel Category=Proprietary2026.05 | 60.67 | |
| Gemini-2.5-FlashModel Category=Proprietary2026.05 | 60.44 | |
| WERType=Objective Metric2026.05 | 59.24 | |
| SpeechJudgeModel Category=Open-Source2026.05 | 57.2 | |
| Kimi-Audio-7BModel Category=Open-Source2026.05 | 52.81 | |
| Qwen2.5-Omni-7BModel Category=Open-Source2026.05 | 51.2 | |
| MiMo-Audio-7BModel Category=Open-Source2026.05 | 50.4 | |
| UTMOSType=Objective Metric2026.05 | 50.2 | |
| DNSMOSType=Objective Metric2026.05 | 49.8 | |
| SIMType=Objective Metric2026.05 | 47.99 |