This research explores methods for making large language models more trustworthy in critical applications by accurately measuring their confidence levels.

• LLMs often produce plausible but incorrect responses, creating significant risks in healthcare, legal, and other high-stakes domains
• Uncertainty quantification (UQ) helps estimate confidence in LLM outputs, enabling better risk management
• Traditional UQ methods face challenges with modern LLMs due to their scale and complexity
• Effective confidence calibration is especially critical in medical applications where incorrect AI recommendations could impact patient safety and treatment outcomes

Uncertainty Quantification and Confidence Calibration in Large Language Models: A Survey