Understanding when a language model is uncertain about its outputs is crucial for improving the safety and reliability of large language models (LLMs). This challenge is particularly pressing given LLMs’ tendency to produce fluent but inaccurate information—a phenomenon commonly referred to as hallucination. In this talk, we will explore the role of uncertainty estimation as a powerful tool for detecting when a model lacks confidence in its own predictions. We will provide a comprehensive overview of various uncertainty quantification methods for LLMs, highlighting their unique characteristics and strengths. Additionally, we will discuss strategies for combining these methods to enhance performance, evaluate their effectiveness, and propose practical approaches for integrating multiple estimators into a cohesive and efficient uncertainty estimation framework.