LM-Polygraph: Uncertainty Estimation for LLMs

LM-Polygraph is an open-source toolkit that helps detect when AI language models are uncertain about their outputs — a key signal for identifying hallucinations. It offers dozens of built-in uncertainty estimation methods, works with both open and closed AI models, and serves as a widely adopted benchmark used by hundreds of researchers and companies to evaluate hallucination detection techniques, making AI-generated text safer and more reliable.