L-MARS: Legal Multi-Agent Workflow with Orchestrated Reasoning and Agentic Search
This addresses the need for precise legal retrieval and deliberation in high-stakes domains, offering a scalable blueprint for deploying LLMs in such applications.
The paper tackles the problem of hallucination and uncertainty in legal question answering by introducing L-MARS, a multi-agent system that decomposes queries, performs targeted searches, and verifies evidence before synthesis. Results on a new benchmark of 200 legal questions show substantial improvements in factual accuracy, reduced uncertainty, and higher preference scores from experts and LLM judges.
We present L-MARS (Legal Multi-Agent Workflow with Orchestrated Reasoning and Agentic Search), a system that reduces hallucination and uncertainty in legal question answering through coordinated multi-agent reasoning and retrieval. Unlike single-pass retrieval-augmented generation (RAG), L-MARS decomposes queries into subproblems, issues targeted searches across heterogeneous sources (Serper web, local RAG, CourtListener case law), and employs a Judge Agent to verify sufficiency, jurisdiction, and temporal validity before answer synthesis. This iterative reasoning-search-verification loop maintains coherence, filters noisy evidence, and grounds answers in authoritative law. We evaluated L-MARS on LegalSearchQA, a new benchmark of 200 up-to-date multiple choice legal questions in 2025. Results show that L-MARS substantially improves factual accuracy, reduces uncertainty, and achieves higher preference scores from both human experts and LLM-based judges. Our work demonstrates that multi-agent reasoning with agentic search offers a scalable and reproducible blueprint for deploying LLMs in high-stakes domains requiring precise legal retrieval and deliberation.