Asked about Building the dataset
Why split an eval set into train and test?
An answer that works
Tuning against every case overfits the prompt to that set, so the score improves while real performance does not. A held-out split you never inspect gives a number that still predicts reality.
Reading an answer is not being able to give one. The section that teaches this is Building the dataset, 5 hours of Evaluation.
Learn this properlyAlso asked about Evaluation
What is an eval?Why can you not just use unit tests?How would you start evaluating an existing feature with no evals?You join a team with an LLM feature and no evals. What do you do in the first week?Why note only the first failure in a trace?Why binary judges rather than a one to five scale?Where do eval cases come from?How big should an eval set be?
All 24 questions on Evaluation