COAT Framework Uses LLM Knowledge for Hidden Causal Discovery

The COAT framework leverages LLM world knowledge to propose latent measured variables for causal discovery in unstructured data. This automates the identification of hidden causal mechanisms that human experts might overlook in complex systems.