Ragas
Definition
Evaluate AI workflows based on defined criteria.
Ragas is a framework for evaluating AI applications, known especially for assessing RAG workflows and their answers.
Simply explained
A test bench considers both the retrieved documents and the formulated answer derived from them.
Why it matters in practice
Different errors in retrieval and generation can be examined separately.
Example from business
A hypothetical team evaluates the source coverage and answer fidelity of its document assistant.
Limits and pitfalls
- Automatic metrics and model-based judgments are not infallible professional acceptance.
Origin & history
Ragas originated in the context of evaluating retrieval-augmented generation.
Use in mid-sized companies
- Select representative cases and align automatic scores with human evaluations.
Frequently asked questions
Is a high overall score sufficient?
Not without checking the relevant types of errors.
Sources
Editorially reviewed on: