Model family by Unbabel Inc. · used by 1 application
Unbabel's open-source multilingual LLM fine-tuned to predict translation quality (translation quality estimation without reference translations). Used by Unbabel to filter and curate high-quality translation data for training models like TowerLLM, and for AI quality estimation. Released earlier in 2024 as open-source.
Six dimensions, evidence-linked
Work at Unbabel Inc.? Claim this listing to correct or complete the data.