Predicting Workers’ Compensation Dispute Outcomes with Large Language Models

We evaluate large language models on workers' compensation disputes and find that, in zero-shot settings with context, they match or outperform specialized NLP, offering actuaries decision support for text-heavy processes.