AI & LLM Evals / Quiz
RAGAS: evaluate() and the Four Columns
user_input, response, retrieved_contexts, reference: the data schema behind ragas.evaluate(), and how its core metrics are actually computed.
- Difficulty
- Easy
- Format
- Quiz
- Points
- 100
- Estimate
- 8 min
// MISSION BRIEF
Your Mission
RAGAS evaluates RAG pipelines from a dataset of samples: the question, the answer, the fetched chunks, and (for some metrics) the ground truth. Get the columns right and evaluate() hands you per-metric, per-row scores; get them wrong and it hands you exceptions.
Six questions on the evaluate() workflow and what each core metric computes. Speed bonus active.
// FIRST CONTACT
Battle teaser
A RAGAS evaluation sample carries which fields?
- AOnly the model's answer; RAGAS infers the rest
- BA screenshot of the chat window
- Cuser_input (question), response (answer), retrieved_contexts (chunks), and, for ground-truth metrics, reference
- DThe model weights and tokenizer
Answers, scoring, hints, and the full battle stay sealed.
// SKILL TAGS
ragasragevaluatemetrics