Real-Time Trustworthiness Scoring for LLM Structured Outputs and Data Extraction

This paper introduces a novel framework for real-time trustworthiness scoring of structured outputs generated by large language models (LLMs). It addresses the critical need for verifiable and reliable data extraction in practical applications by evaluating the consistency and factual accuracy of LLM responses against external knowledge sources or predefined constraints. The proposed method enables immediate feedback on the quality of extracted data, significantly enhancing the utility and safety of LLMs in enterprise environments.