Recording datasets: Which practice best satisfies all three needs at once?
Detailed training-and-testing documentation of data, versions, parameters and results gives reproducibility, compliance evidence and a drift baseline.
The question
A regulated organization must be able to reproduce its model's results, prove compliance to an external assessor, and later compare live behavior against original test performance to detect drift. It has limited resources and can prioritize only one documentation practice for the training and testing phase. Which practice best satisfies all three needs at once?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Maintaining a public post-market monitoring plan that describes how live performance data will be collected and then periodically reviewed after deployment.Plausible but wrong: a monitoring plan governs post-release activity but does not by itself capture the training/testing baseline for reproducibility.
- Producing a concise model card summarizing intended use and headline metrics for stakeholders who consume the model's outputs downstream.Plausible but wrong: a model card communicates summary information but lacks the detailed run records needed to reproduce results or prove compliance.
- Keeping an incident register that logs failures reported after release together with their root causes and the remediation actions taken.Plausible but wrong: an incident register documents post-release events, not the training-and-testing evidence the three needs require.
- Recording datasets, versions, parameters, procedures and results of training and testing so runs are reproducible and provide a compliance baseline. ✓Correct: a complete training-and-testing record supports reproducibility, serves as compliance evidence, and gives the baseline needed to spot later drift.
The trap
Selecting a post-release or summary artifact that presupposes the detailed training-and-testing record already exists. How to remember it
Detailed training-and-testing documentation of data, versions, parameters and results gives reproducibility, compliance evidence and a drift baseline.
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
- Investigate the underrepresentation: Following good practice for managing issues and risks during training and →
- Delay the launch until conformity requirements are: What is the most defensible release-readiness decision? →
- Treat the input shift as data drift and trigger scheduled: Which maintenance action does this pattern most →
- 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.
