Replication Data for ACL 2026 paper: \"JudgeMeNot: Personalizing Large Language Models to Emulate Judicial Reasoning in Hebrew\"

This dataset contains the replication data for the ACL 2026 paper: \"JudgeMeNot: Personalizing Large Language Models to Emulate Judicial Reasoning in Hebrew\". In contain Hebrew text from Israeli court decisions in which a single judge decided the case (sole-judge rulings, as opposed to multi-judge panels). The materials are organized by judge. Two complementary tables are provided for each judge: Question–answer (QA) pairs: Legal questions paired with reference answers taken as excerpts from the underlying rulings, plus a train / evaluation / test split label. These rows support instruction-style modeling and evaluation. Raw ruling text: The same rulings are also provided as contiguous text segments (shorter passages cut from the full opinion text), with the same split labels. These segments support continuation-style or causal language modeling of each judge’s writing. The text is in Modern Hebrew