Assign candidate examples to training, evaluation, or rejection based on what they teach and measure.
Sort each candidate example by its role in the fine-tuning workflow. Train Eval Reject Clean common success path with approved output train Rare edge case leadership wants measured eval Duplicate of another easy example reject Prompt contains an answer users never provide Fresh failure mode with reviewer agreement Ambiguous case reviewers still dispute Split fine-tuning data You gave each row a job before training. Ask whether the row should teach the model, test the model, or stay out.
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in