Grounding

Grounding is the practice of tying a language model’s output to verifiable external evidence — retrieved documents or data — so its responses reflect real sources rather than its parametric guesses.

A grounded response is generated from supplied evidence, which reduces hallucination and lets the answer be traced back to its source.

Retrieval-augmented generation is the most common grounding technique, and grounding with citations makes the evidence explicit to the reader.