Glossary · Core Concepts

Grounding

The practice of anchoring a model's output in verifiable external information (retrieved documents, tool results) rather than relying solely on its internal training knowledge, directly reducing hallucination risk.