ClauseLens: Clause-Grounded, CVaR-Constrained Reinforcement Learning for Trustworthy Reinsurance Pricing

2025/10/09 by Dong, Stella C., Finlay, James R.
#68T05 #91G70 #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #I.2.6 #I.2.7 #J.1 #Machine Learning (cs.LG) #Machine Learning (stat.ML)

paper · doi:10.48550/arxiv.2510.08429

Abstract

Reinsurance treaty pricing must satisfy stringent regulatory standards, yet current quoting practices remain opaque and difficult to audit. We introduce ClauseLens, a clause-grounded reinforcement learning framework that produces transparent, regulation-compliant, and risk-aware treaty quotes. ClauseLens models the quoting task as a Risk-Aware Constrained Markov Decision Process (RA-CMDP). Statutory and policy clauses are retrieved from legal and underwriting corpora, embedded into the agent's observations, and used both to constrain feasible actions and to generate clause-grounded natural language justifications. Evaluated in a multi-agent treaty simulator calibrated to industry data, ClauseLens reduces solvency violations by 51%, improves tail-risk performance by 27.9% (CVaR₀.10), and achieves 88.2% accuracy in clause-grounded explanations with retrieval precision of 87.4% and recall of 91.1%. These findings demonstrate that embedding legal context into both decision and explanation pathways yields interpretable, auditable, and regulation-aligned quoting behavior consistent with Solvency II, NAIC RBC, and the EU AI Act.

Citations

• Dynamic Reinsurance Treaty Bidding via Multi-Agent Reinforcement Learning
• Retrieval-Augmented Reinforcement Learning
• When Does Pretraining Help? Assessing Self-Supervised Learning for Law and the CaseHOLD Dataset
• LEGAL-BERT: The Muppets straight out of Law School
• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• Constrained Reinforcement Learning Has Zero Duality Gap
• Explainable Reinforcement Learning Through a Causal Lens
• Reward Constrained Policy Optimization
• Proximal Policy Optimization Algorithms
• Towards A Rigorous Science of Interpretable Machine Learning
• Risk-Constrained Reinforcement Learning with Percentile Risk Criteria
• Optimizing the CVaR via Sampling

Related

• RKEFino1: A Regulation Knowledge-Enhanced Large Language Model
• Natural Language Processing for Financial Regulation
• Context Reasoner: Incentivizing Reasoning Capability for Contextualized Privacy and Safety Compliance via Reinforcement Learning
• Scaling Policy Compliance Assessment in Language Models with Policy Reasoning Traces
• Towards Secure and Explainable Smart Contract Generation with Security-Aware Group Relative Policy Optimization