Answers are grounded in this paper and cite their sources.
System design study demonstrates a retrieval-augmented generation framework for legal question answering, highlighting reduced hallucination and verifiable citation grounding.
This preprint presents an AI-based legal information retrieval chatbot designed to provide natural-language answers grounded in verified legal sources. The proposed system uses retrieval-augmented generation to retrieve relevant passages from a curated legal knowledge base containing statutes, regulations, and case summaries. A hybrid retrieval mechanism combines dense semantic retrieval with BM25 lexical retrieval, followed by grounded answer generation, atomic claim extraction, citation matching, and entailment verification. A confidence-based abstention mechanism prevents the system from returning unsupported legal claims when sufficient evidence is unavailable. The proposed evaluation considers answer accuracy, retrieval performance, citation correctness, and hallucination rate, with the objective of improving the reliability and traceability of conversational legal information access.
No takes yet. Share an insight, caveat, or question.
Adishree Gupta (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: