# AI Evals

Canonical URL: https://aievals.net/

AI Evals is a research-backed field guide to evaluating AI agents: systems that
plan, use tools, interact with stateful environments, and act over multiple
steps. It explains why final-answer scoring is insufficient, how to evaluate
outcomes and trajectories, how repeated trials reveal reliability, and how
safety, grader validity, benchmark integrity, and production monitoring fit
into one evaluation program.

The guide synthesizes 25 primary sources from standards bodies, AI labs,
academic benchmarks, and evaluation-framework maintainers. Source links are
provided on the page for verification and further reading.