Machine Translation on Multi30K MSCOCO (eng → fra)
55.3BLEUSMT-9B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SMT-9BInput Modality=Text, Speech, Speech Source=Synthetic2026.02 | 55.3 | 86.7 | |
| DeepSeek-V3.1Input Modality=Text2026.02 | 53.5 | 85.8 | |
| Baseline + Lora (Text only)Input Modality=Text, Adaptation=LoRA2026.02 | 53.4 | 85.9 | |
| NLLB-moe-54BInput Modality=Text2026.02 | 53.3 | 85.3 | |
| Baseline (Text only)Input Modality=Text2026.02 | 52.6 | 86.1 | |
| Qwen3-Next-80B-A3BInput Modality=Text2026.02 | 50.4 | 85.1 | |
| Gemma3-27B-itInput Modality=Text2026.02 | 49.6 | 85 | |
| IMAGEInput Modality=Text, Image, Image Source=Synthetic2026.02 | 49.3 | 82.5 | |
| Soul-MixInput Modality=Text, Image, Image Source=Authentic2026.02 | 49.2 | — | |
| ConsQA-MMTInput Modality=Text, Image, Image Source=Authentic2026.02 | 48.5 | — | |
| VALHALLAInput Modality=Text, Image, Image Source=Synthetic2026.02 | 46.5 | — | |
| BridgeInput Modality=Text, Image, Image Source=Synthetic2026.02 | 46.3 | — | |
| RG-MMT-EDCInput Modality=Text, Image, Image Source=Authentic2026.02 | 45.1 | — | |
| WRA-guidedInput Modality=Text, Image, Image Source=Authentic2026.02 | 43.4 | — | |
| DreamLLMInput Modality=Text, Image, Image Source=Synthetic2026.02 | 36.6 | 79.2 |