A team preparing a production rollout needs to avoid a situation where reviewers cannot tell what data trained a given model. What governance practice addresses this directly?
Select an answer to reveal the explanation.
Short Explanation
Throughput and compute don't tell a reviewer where the training data came from. You need the factsheet's data lineage record so the evidence trail shows what trained the model.
Full Explanation
Recording data lineage as part of the model's factsheet and evidence record is correct because factsheets are built to record the data a model was trained and evaluated on, which is precisely the traceability a reviewer needs. Increasing the model's inference throughput before release is incorrect because throughput is a performance characteristic unrelated to data provenance. Adding more compute nodes to the training cluster is incorrect because compute capacity does not document what data was used. Shortening the model's release notes to a single sentence is incorrect because shortening release notes reduces documentation rather than supplying the missing lineage. For the exam, tie data traceability directly to factsheet and evidence record content.