Offline Speech Translation on MCIF (test)
86.59XCOMET-XLOmni + SeedX
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Omni + SeedXCategory=Cascade Systems (ASR → MT), Config=LoRA FT, Image Input=true2025.11 | 86.59 | 1,202.3 | 766.3 | 55.7 | 2,024.3 | |
| OmniFusionCategory=End-to-End (Direct ST), Config=Gated Fusion + Self-Cascade, Image Input=true2025.11 | 86.57 | 1,231 | 719 | 55.3 | 2,005.3 | |
| Omni + SeedXCategory=Cascade Systems (ASR → MT), Config=LoRA FT, Image Input=false2025.11 | 86.42 | 1,239 | 751.3 | 75 | 2,065.3 | |
| OmniFusionCategory=End-to-End (Direct ST), Config=Gated Fusion, Image Input=true2025.11 | 86.24 | 1,255.7 | 736 | 57 | 2,048.7 | |
| OmniFusionCategory=End-to-End (Direct ST), Config=Mid Fusion, Image Input=true2025.11 | 85.98 | 1,256.3 | 749 | 61.7 | 2,067 | |
| Omni + SeedXCategory=Cascade Systems (ASR → MT), Config=No FT, Image Input=true2025.11 | 85.94 | 1,191.3 | 759.3 | 74.7 | 2,025.3 | |
| Omni + SeedXCategory=Cascade Systems (ASR → MT), Config=No FT, Image Input=false2025.11 | 85.88 | 1,212.3 | 773 | 74.3 | 2,059.7 | |
| OmniFusionCategory=End-to-End (Direct ST), Config=Gated Fusion, Image Input=false2025.11 | 83.98 | 1,119.3 | 1,188 | 62.7 | 2,370 | |
| OmniFusionCategory=End-to-End (Direct ST), Config=Mid Fusion, Image Input=false2025.11 | 83.57 | 1,114 | 1,188.3 | 72.3 | 2,374.7 |