MELA introduces a multilingual benchmark for linguistic acceptability judgment, showcasing the importance of in-language training data and its impact on syntax-related tasks.
