Skip to interview prep
Apexed

80. An answer was correct when an evaluation was created but is wrong after a policy changes. How would you maintain the benchmark?

MediumMachine learning basics

Technical
Show model answer

Record the reference date, authoritative source and intended time scope for each time-sensitive task. Decide whether the task asks about the historical situation or requires the latest valid information.

Update or retire affected references through a reviewed process while preserving benchmark versions for comparison. Separate a stale reference from a model error, and avoid silently changing labels midway through a candidate comparison.

What interviewers listen for

  • Version dates and authoritative references
  • Distinguish historical from current tasks
  • Review label changes between comparisons
An answer was correct when an evaluation was created but is wrong after a policy changes. How would you maintain the benchmark — Data & analytics interview ques