Calibrated large language models for multi-label classifications

Aspects disclosed provide system and methods for providing reliability measures to outputs of large language models (LLMs). The system and methods do this by integrating Large Language Models (LLMs) in a multi-label classification setting, utilizing the Conformal Prediction (CP) framework. This approach ensures that the predictions made by the LLM are accompanied by mathematically guaranteed error bounds, enhancing the LLMs reliability and trustworthiness.