Machine Translation on Multi30K eng → deu 2016 (test)
47BLEUSMT-9B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SMT-9BInput Modality=Text, Speech, Speech Source=Synthetic2026.02 | 47 | 87.8 | |
| IMAGEInput Modality=Text, Image, Image Source=Synthetic2026.02 | 45.3 | 83.1 | |
| DeepSeek-V3.1Input Modality=Text2026.02 | 44.2 | 87.3 | |
| Soul-MixInput Modality=Text, Image, Image Source=Authentic2026.02 | 44.2 | — | |
| ConsQA-MMTInput Modality=Text, Image, Image Source=Authentic2026.02 | 44.2 | — | |
| Baseline + Lora (Text only)Input Modality=Text, Adaptation=LoRA2026.02 | 44 | 87 | |
| Gemma3-27B-itInput Modality=Text2026.02 | 43.7 | 87.1 | |
| D2P-MMT (R)Method Category=Image-dependent Methods, Inference Image Source=Reconstructed, Ensemble=false2025.07 | 43.12 | — | |
| Baseline (Text only)Input Modality=Text2026.02 | 42.9 | 87 | |
| D2P-MMT (A)Method Category=Image-dependent Methods, Inference Image Source=Authentic, Ensemble=false2025.07 | 42.72 | — | |
| VALHALLAInput Modality=Text, Image, Image Source=Synthetic2026.02 | 42.7 | — | |
| VALHALLA*Method Category=Image-free Methods, Ensemble=true2025.07 | 42.7 | — | |
| VALHALLAMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 42.6 | — | |
| Noise-robustMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 42.56 | — | |
| MMT-VQAMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 42.55 | — | |
| BridgeInput Modality=Text, Image, Image Source=Synthetic2026.02 | 42.5 | — | |
| Doubly-ATTMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 42.45 | — | |
| RG-MMT-EDCInput Modality=Text, Image, Image Source=Authentic2026.02 | 42 | — | |
| Gated Fusion*Method Category=Image-dependent Methods, Ensemble=true2025.07 | 41.96 | — | |
| Selective AttentionMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 41.84 | — | |
| Qwen3-Next-80B-A3BInput Modality=Text2026.02 | 41.6 | 86.3 | |
| NLLB-moe-54BInput Modality=Text2026.02 | 41.4 | 86.2 | |
| RMMT*Method Category=Image-free Methods, Ensemble=true2025.07 | 41.4 | — | |
| IKD-MMTMethod Category=Image-free Methods, Ensemble=false2025.07 | 41.28 | — | |
| Transformer-TinyMethod Category=Text-Only Transformer, Ensemble=false2025.07 | 40.69 | — | |
| DCCNMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 39.7 | — | |
| WRA-guidedInput Modality=Text, Image, Image Source=Authentic2026.02 | 39.3 | — | |
| Gumbel-AttentionMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 39.2 | — | |
| ImagiTMethod Category=Image-free Methods, Ensemble=false2025.07 | 38.5 | — | |
| UVR-MMTMethod Category=Image-free Methods, Ensemble=false2025.07 | 36.9 | — | |
| DreamLLMInput Modality=Text, Image, Image Source=Synthetic2026.02 | 27.2 | 74.8 |