This preprint presents AILQA, an AI legal question-answering system designed for the Indian legal context. It evaluates multiple embedding and generative models, including recent LLMs, using lexical and semantic metrics alongside expert legal feedback. The study focuses on retrieval-augmented generation (RAG) for difficult legal questions and includes evaluation on the All India Bar Examination (AIBE). It reports that, under its dataset and rating criteria, some generated answers scored higher than reference answers because they included accurate supporting details. The authors also discuss context sensitivity, hallucination risks, and future improvements. These findings do not establish that AI generally outperforms qualified lawyers.
No heat snapshots are available in the last 24 hours.