MEET-MR

Advancing English–Thai Machine Translation Evaluation

MEET-MR provides high-quality human-annotated benchmarks and models to bridge the evaluation gap for the English–Thai language pair. Our work focuses on fine-grained error analysis (MQM) and human preference alignment.

Resources

• Dataset: 20k segments with MQM and human rankings.
• Models: Optimized COMET for English-Thai quality estimation.

Publication

EACL 2026: Machine Translation Evaluation English-Thai MQM Ranking Dataset. OpenReview Paper

For inquiries or collaborations, please reach out via our primary repository issues or the contacts listed in our paper.