78. A model produces the required table columns and valid data types. What additional evaluation is needed?
EasyMachine learning basics
Technical
Show model answer
Check whether values are supported by the input, calculations are correct and rows represent the requested population. Structural validity is useful, but it does not establish factual accuracy, completeness or the absence of fabricated records.
Use independently verified examples and business invariants, such as totals reconciling to permitted source data. Score structural and semantic failures separately so improvements in formatting cannot disguise deterioration in substantive correctness.
What interviewers listen for
- Separate structural and semantic validity
- Reconcile values with source evidence
- Score different failure types separately