Speech-to-Speech Translation on CVSS-C Average
82.86ASR-BLEUGround Truth
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Ground TruthEvaluation Setting=Reference2026.05 | 82.86 | — | |
| S2ST-Omni 2Evaluation Setting=Single many-to-one model2026.05 | 35 | 37.73 | |
| S2ST-OmniEvaluation Setting=Single many-to-one model2026.05 | 33.45 | 35.67 | |
| RosettaSpeechEvaluation Setting=Single many-to-one model2026.05 | 27.68 | 29.08 | |
| StreamSpeechEvaluation Setting=Originally reported2026.05 | 25.54 | 28.77 | |
| SimulS2S-LLMEvaluation Setting=Originally reported2026.05 | 24.92 | — | |
| UnitYEvaluation Setting=Originally reported2026.05 | 23.82 | — | |
| ComSpeechEvaluation Setting=Originally reported2026.05 | 23.7 | 25.55 | |
| Translatotron 2Evaluation Setting=Originally reported2026.05 | 21.97 | 24.43 | |
| DASpeechEvaluation Setting=Originally reported2026.05 | 20.85 | — | |
| Whisper–Qwen S2TTEvaluation Setting=Text-level reference2026.05 | — | 36.54 |