For language-model responses, readability is an evaluation metric for clarity and accessibility: it asks whether content is easy for its intended users to understand. It complements measures such as accuracy and contextual relevance. section 5.4.2
Assessing readability
Evaluate responses using readability scores or evaluator ratings. These checks can sit within a hybrid evaluation process that combines automated assessment with human review. section 5.4.2
