Contract Negotiation OpenEnv Environment

An OpenEnv reinforcement learning environment where an AI acts as a legal reviewer. The agent analyzes contracts, identifies hidden unfair clauses ("traps"), negotiates changes with an adversarial counterparty, and sequentially decides whether to accept, counter, or walk away.

Why Contract Negotiation?

• Novelty: Legal contract review is far less generic than traditional cybersecurity or navigation environments.
• Deep Reasoning: Requires the agent to first investigate text before making blind proposals.
• Rich Assessment: Evaluates on trap detection, actual amendment quality, negotiation efficiency, fairness of outcome, and strategic walk-away logic.

Quickstart

python -m venv .venv
# Activate the environment
# Windows: .venv\Scripts\activate
# Unix: source .venv/bin/activate
pip install -r requirements.txt

# Run main interactive demo
python demo_run.py

# Showcase targeted risk-aware strategy
python showcase_run.py

# Run unit tests and generate the benchmark artifact
python run_all.py

Structure

• contract_negotiation_env/: Core OpenEnv RL implementation, models, grader, and the procedural contract generation tools.
• tests/: Deterministic testing suites for the grader, generators, and environment.
• SUBMISSION.md: Full background, problem framing, and evaluation details for the hackathon.
• RESULTS.md: Benchmark results on various difficulty setups.
• DEPLOYMENT.md: Instructions for Docker images and Hugging Face deployment.

Happy hacking!