ENHANCING GENERATIVE AI RELIABILITY THROUGH SIMILARITY ANALYSIS

A computer-implemented method for improved schema matching of two databases is disclosed. The method can transform one or more text benchmarks into one or more first vector embeddings, retrieve a text output from the generative AI model in response to a user prompt, transform the text output into one or more second vector embeddings, measure one or more matching scores based on comparison of the one or more first vector embeddings with the one or more second vector embeddings, and determine reliability of the text output based on evaluation of the one or more matching scores. Related systems and software for implementing the method are also disclosed.