Machine Translation on Multi30K eng → fra 2016 (test)
67.5BLEUIMAGE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| IMAGEInput Modality=Text, Image, Image Source=Synthetic2026.02 | 67.5 | 88.3 | |
| SMT-9BInput Modality=Text, Speech, Speech Source=Synthetic2026.02 | 67 | 90 | |
| ConsQA-MMTInput Modality=Text, Image, Image Source=Authentic2026.02 | 64.8 | — | |
| Soul-MixInput Modality=Text, Image, Image Source=Authentic2026.02 | 64.7 | — | |
| BridgeInput Modality=Text, Image, Image Source=Synthetic2026.02 | 63.7 | — | |
| VALHALLAInput Modality=Text, Image, Image Source=Synthetic2026.02 | 63.1 | — | |
| RG-MMT-EDCInput Modality=Text, Image, Image Source=Authentic2026.02 | 62.9 | — | |
| WRA-guidedInput Modality=Text, Image, Image Source=Authentic2026.02 | 61.8 | — | |
| Baseline + Lora (Text only)Input Modality=Text, Adaptation=LoRA2026.02 | 55.5 | 88.1 | |
| Gemma3-27B-itInput Modality=Text2026.02 | 55.4 | 87.9 | |
| DeepSeek-V3.1Input Modality=Text2026.02 | 55.3 | 88.2 | |
| NLLB-moe-54BInput Modality=Text2026.02 | 55.1 | 87.4 | |
| Qwen3-Next-80B-A3BInput Modality=Text2026.02 | 53.2 | 87.8 | |
| Baseline (Text only)Input Modality=Text2026.02 | 52.4 | 87.7 | |
| DreamLLMInput Modality=Text, Image, Image Source=Synthetic2026.02 | 36.9 | 81.1 |