Pro MQM-Annotated MT Dataset Released
A professional linguist-annotated dataset for machine translation quality evaluation is now open-sourced on Hugging Face. It features 362 segments across 16 language pairs, annotated by 48 experts using MQM framework following WMT guidelines. Inter-annotator agreement reaches Kendall's τ = 0.317, 2.6x higher than typical WMT campaigns.






