Evaluate once on a held-out December-like set: Which evidence would most directly address whether performance
Evaluate the finalized model on a held-out, December-like set to test generalization beyond tuning data.
The question
A retailer trained a demand model on January–October sales and used November records for tuning. The team now reports strong November accuracy and seeks approval for December deployment. Which evidence would most directly address whether performance generalizes beyond tuning data?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Report November accuracy with additional decimal placesMore precise reporting does not change the fact that November records influenced tuning and are not held out.
- Ask planners whether November forecasts seemed reasonablePlanner judgments may supplement evaluation, but cannot replace a held-out test of performance under deployment-like conditions.
- Evaluate once on a held-out December-like set ✓A separate held-out set resembling December deployment conditions provides independent evidence after training and tuning are complete.
- Tune the model further on November recordsFurther tuning on November records increases exposure to the same data and cannot provide an independent generalization check.
The trap
Check whether any reported evaluation cases influenced model fitting or tuning; if so, they are not held out. How to remember it
Evaluate the finalized model on a held-out, December-like set to test generalization beyond tuning data.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding How to Govern AI Development questions
- Compare model outcomes with and without distance: Which analysis is most informative? →
- Create a deduplicated held-out clip set: Which evidence should reviewers request? →
- Use subgroup acceptance criteria and investigate: What evidence should determine approval? →
- All 426 Understanding How to Govern AI Development questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Development ·
Every answer, right and wrong, comes with its own explanation.