liveBook · Manning

Fidelity

As explained in AI Model Evaluation

Groundedness is a metric measuring whether an output is supported by the information it is expected to rely on.

Fidelity, also called groundedness here, measures whether generated answers are supported by the retrieved context. In a RAG chatbot, it helps assess whether responses stay anchored to information they were expected to use. section 1.3.4 section 2.3.6

Compare model versions using a fixed evaluation setup and a defined metric. An LLM judge can score groundedness, but its output is an evaluation signal, not absolute truth; validate and interpret it carefully. section 1.3.4 section 1.4

Make results useful to product decisions by translating a score change into plain language—for example, more generated answers were supported by retrieved context. A benchmark improvement alone may not capture live product behavior. section 2.3.6 section 1.4

How AI Model Evaluation covers it

How deep it goes
Defines groundedness and gives a product-language interpretation, without detailing a calculation method.
Examples
Uses a RAG chatbot and explains improvement as more answers supported by retrieved context.
What you can do after
Evaluate whether retrieved context supports generated answers and interpret score changes in product-relevant terms.
How it connects
Pairs groundedness with relevance and instruction following, and treats LLM-as-a-judge as a scoring method rather than a metric.

Also explained in

Compare how 11 books explain fidelity →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Model Evaluation. The section links point to the parts of the book it draws on.