Audit glossary

Evaluating generative AI (evals)

Testing generative AI systems for quality, accuracy and safety.

AAIA

Evals use test sets of prompts with expected behaviours to measure factual accuracy, groundedness, relevance, harmful output, bias and resistance to jailbreaks. Because outputs vary, tests are repeated and sampled, and human review is combined with automated scoring.

On the exam

Generative systems need ongoing evaluation, not a single test, especially after any model, prompt or data change.

Related terms

  • Large language model (LLM)A generative model trained on huge amounts of text to understand and produce language.
  • AI red teamingDeliberately attacking an AI system to find weaknesses before others do.
  • HallucinationConfident but false output from a generative AI model.

AAIA Test yourself on this with timed AAIA practice questions in the AAIA Exam Simulation Hub, or read the free AAIA exam guide.

Test yourself with today’s question of the day.