L-MARS - Enhanced Multi-Agent Workflow for Legal QA

📄 Paper: L-MARS: Legal Multi-Agent Workflow with Orchestrated Reasoning and Agentic Search

L-MARS (Legal Multi-Agent Framework for Orchestrated Reasoning and Agentic Search) combines multiple search strategies, agent-based reasoning, and evaluation systems for legal QA.

Key Features

Operating Mode

• Simple Mode: Fast single-turn pipeline with multiple search results
• Multi-Turn Mode: Iterative refinement with judge agent evaluation and deep content extraction

Retrieval Source

1. Online Search (Default): Web search via Serper API - Always enabled
   • Simple mode: Quick search returning 5 results with snippets
   • Multi-turn mode: Deep search returning 3 results with full content extraction
2. Local RAG (Optional): BM25-based local document retrieval from inputs/ folder - Enable with --offline-rag
3. Case Law (Optional): Legal case retrieval via CourtListener API - Enable with --courtlistener

Quick Start

Installation

# Clone the repository
git clone https://github.com/yourusername/l-mars.git
cd l-mars

# Install the package with dependencies
pip install -e .

# Set up environment variables
cp .env.example .env
# Edit .env and add your API keys:
# - OPENAI_API_KEY (required)
# - SERPER_API_KEY (required, for web search)
# - COURTLISTENER_API_TOKEN (optional, for case law)

Setting Up Local Documents for Offline RAG

Place your legal documents in markdown format in the inputs/ folder:

# Create inputs directory
mkdir -p inputs

# Add your markdown documents
cp your_legal_docs/*.md inputs/

The system will automatically index and search these documents.

Command Line Usage

Simple Mode

# Quick legal research with online search only (default)
python main.py "Your legal question"

# Enable offline RAG for local documents
python main.py --offline-rag "Your legal question"

# Enable all sources (offline RAG + CourtListener + web search)
python main.py --all-sources "Your legal question"

# With verbose output
python main.py -v "Your legal question"

Multi-Turn Mode

# Thorough research with refinement
python main.py --multi "Complex contract dispute..."

# With custom iterations
python main.py --multi --max-iterations 5 "Your question"

Citation

If you use L-MARS in your research, please cite:

@misc{wang2025lmarslegalmultiagentworkflow,
      title={L-MARS: Legal Multi-Agent Workflow with Orchestrated Reasoning and Agentic Search}, 
      author={Ziqi Wang and Boqin Yuan},
      year={2025},
      eprint={2509.00761},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2509.00761}, 
}