Documentation
Everything Evaltude publishes, in four families.
The reusable method, the evaluations it has been applied to, the evidence under them, and the reference material that defines the terms.
Start here
What Evaltude is, and how the claim, the evidence and the method fit together.
Learn the methodA 100-lesson curriculum on evaluating complete LLM systems — defining success, designing evaluation data, diagnosing failures, and understanding what serving choices cost.
Worked evaluationsEvaluations documented end to end: the business question, the systems compared, the method, the results, and what each result does not establish.
ReferenceThe glossary, reproducibility notes, and the sources behind every published number.