Glossary term
Glossary term
Evaluation and Benchmarks
Strangely, an abbreviation for Conference on Machine Translation. (The abbreviation is WMT because the original name was Workshop on Machine Translation.) The conference focuses on developments in machine translation systems.
Created for this library
An LLM evaluation team uses WMT benchmarks to measure translation quality across language pairs.
A translation vendor reports WMT BLEU scores when announcing model improvements on key language pairs.
A research lab uses WMT datasets as the standard evaluation suite for machine translation experiments.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License