Tag
This article introduces Magis-Bench, a benchmark for evaluating large language models on magistrate-level legal tasks such as judicial reasoning and sentence drafting, using data from Brazilian judicial exams.