---
type: GlossaryTerm
genre: reference
id: https://noelmcloughlin.github.io/ai-linkmo/knowledge/glossary/evaluation
title: Evaluation
definition: A benchmark or test that measures a risk or a capability, linked to the datasets and tasks it uses.
description: A benchmark or test measuring a risk or a capability, linked to its datasets and tasks; served by the /evaluation endpoint.
resource: docs/key-concepts.md
tags:
- glossary
- ontology
generated:
  by: process:ktl-librarian
  at: "2026-10-17T00:24:57Z"
status: draft
definedBy:
- https://noelmcloughlin.github.io/ai-linkmo/knowledge/references/ai-risk-ontology
relatedTo:
- https://noelmcloughlin.github.io/ai-linkmo/knowledge/glossary/risk
verified:
- by: process:ktl-librarian
  at: "2026-10-16T23:49:47Z"
---

# Evaluation

An **evaluation** is a benchmark or test that measures a [risk](risk.md) or a capability, linked to the datasets and tasks it uses.

The `/evaluation` endpoint and `./ai evaluation` serve them from the packaged ai-atlas-nexus evaluation data, for example `ai_eval_PopQA`. An evaluation is how a risk named in a [taxonomy](taxonomy.md) becomes something measurable on a particular model.