Machine Translation on Multi30K eng → deu 2017 (test)
41.8BLEUSMT-9B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SMT-9BInput Modality=Text, Speech, Speech Source=Synthetic2026.02 | 41.8 | 87.3 | |
| DeepSeek-V3.1Input Modality=Text2026.02 | 41.1 | 86.8 | |
| Gemma3-27B-itInput Modality=Text2026.02 | 40.3 | 86.3 | |
| NLLB-moe-54BInput Modality=Text2026.02 | 39.7 | 85.8 | |
| Baseline + Lora (Text only)Input Modality=Text, Adaptation=LoRA2026.02 | 39.4 | 86.4 | |
| Baseline (Text only)Input Modality=Text2026.02 | 38.8 | 86.4 | |
| IMAGEInput Modality=Text, Image, Image Source=Synthetic2026.02 | 38.6 | 81.9 | |
| Qwen3-Next-80B-A3BInput Modality=Text2026.02 | 37.6 | 85.9 | |
| ConsQA-MMTInput Modality=Text, Image, Image Source=Authentic2026.02 | 37.6 | — | |
| Soul-MixInput Modality=Text, Image, Image Source=Authentic2026.02 | 37.1 | — | |
| BridgeInput Modality=Text, Image, Image Source=Synthetic2026.02 | 36 | — | |
| D2P-MMT (R)Method Category=Image-dependent Methods, Inference Image Source=Reconstructed, Ensemble=false2025.07 | 35.54 | — | |
| D2P-MMT (A)Method Category=Image-dependent Methods, Inference Image Source=Authentic, Ensemble=false2025.07 | 35.24 | — | |
| VALHALLAInput Modality=Text, Image, Image Source=Synthetic2026.02 | 35.1 | — | |
| VALHALLA*Method Category=Image-free Methods, Ensemble=true2025.07 | 35.1 | — | |
| VALHALLAMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 35.1 | — | |
| Noise-robustMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 35.09 | — | |
| MMT-VQAMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 34.58 | — | |
| Selective AttentionMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 34.32 | — | |
| Transformer-TinyMethod Category=Text-Only Transformer, Ensemble=false2025.07 | 34.26 | — | |
| Doubly-ATTMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 33.95 | — | |
| IKD-MMTMethod Category=Image-free Methods, Ensemble=false2025.07 | 33.83 | — | |
| Gated Fusion*Method Category=Image-dependent Methods, Ensemble=true2025.07 | 33.59 | — | |
| RG-MMT-EDCInput Modality=Text, Image, Image Source=Authentic2026.02 | 33.4 | — | |
| RMMT*Method Category=Image-free Methods, Ensemble=true2025.07 | 32.9 | — | |
| WRA-guidedInput Modality=Text, Image, Image Source=Authentic2026.02 | 32.3 | — | |
| ImagiTMethod Category=Image-free Methods, Ensemble=false2025.07 | 32.1 | — | |
| Gumbel-AttentionMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 31.4 | — | |
| DCCNMethod Category=Image-dependent Methods, Ensemble=false2025.07 | 31 | — | |
| UVR-MMTMethod Category=Image-free Methods, Ensemble=false2025.07 | 28.6 | — | |
| DreamLLMInput Modality=Text, Image, Image Source=Synthetic2026.02 | 19.5 | 73.5 |