Speech Translation on CoVoST2 (EuroParlST Weighted Average, test)
81.9Score (en-de)Llama-3.1-8B-Instruct
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Llama-3.1-8B-InstructModel Category=Text-only Models, Zero-shot LLM=true2026.07 | 81.9 | 84.1 | 77 | — | |
| SeamlessM4T-v2-large (MT)Model Category=Text-only Models, Zero-shot LLM=true2026.07 | 81.4 | 86.7 | 80.8 | — | |
| Qwen3-4B-Instruct-2507 + LoRA (B)Model Category=Text-only Models, LoRA=B2026.07 | 80.7 | 86.9 | 84.7 | — | |
| SeamlessM4T-v2-large (ASR/ST)Model Category=SFM and Projector-only Model2026.07 | 78.3 | 76.9 | 78 | — | |
| BEST-IWSLT25-IFModel Category=Multimodally Trained Models2026.07 | 77.3 | 84.2 | 80.2 | — | |
| SpeechMapper + LoRA (C) setup 2Model Category=Multimodally Trained Models, LoRA=C, Setup=22026.07 | 76.3 | 84.4 | 81.3 | — | |
| SpeechMapper + LoRA (C) setup 1Model Category=Multimodally Trained Models, LoRA=C, Setup=12026.07 | 75.1 | 83.2 | 80.6 | — | |
| SpeechMapper (A)Model Category=SFM and Projector-only Model, Setup=A2026.07 | 73.5 | 80.1 | 79.7 | — | |
| Qwen3-4B-Instruct-2507Model Category=Text-only Models, Zero-shot LLM=true2026.07 | 71 | 67.7 | 74.3 | — | |
| BEST-IWSLT25-IF2026.07 | — | — | — | 74.3 | |
| SpeechMapperSetup=A2026.07 | — | — | — | 71.1 | |
| SpeechMapper + LoRA (C)Setup=setup 12026.07 | — | — | — | 77.2 | |
| SpeechMapper + LoRA (C)Setup=setup 22026.07 | — | — | — | 78.1 |