List your company

Speech Quality Assessment on S2S-ARENA

73.3Content ScoreHuman–human agreement
23.06836.10949.1562.191Jan 20, 2026

Evaluation results

MethodLinks
Human–human agreementJudge=Human2026.0173.348.37575
Audio JudgeJudge=Audio-only2026.0158.952.537.7—
TRACEBackbone=Gemini 2.5 Flash, Judge=Tree-based fusion2026.015851.648.1—
LLM JudgeJudge=Transcript-only2026.015755.135.4—
Random GuessJudge=Random2026.0125252525