Recognize sample accuracy while treating reliability as: Which interpretation is most defensible?
The sample supports observed accuracy, while repeated inconsistent outputs leave reliability unresolved for operational approval.
The question
A bank’s document model correctly extracts every field in 100 consecutive invoices, but produces inconsistent results when the same invoice is uploaded repeatedly. Approval requires dependable processing across ordinary operating conditions. Which interpretation is most defensible?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Recognize sample accuracy while treating reliability as unresolved. ✓Correct outputs in the sample provide limited accuracy evidence, whereas inconsistent repeated outputs directly leave dependable behavior unresolved.
- Approve the model because the consecutive-invoice sample was entirely correct.The sample supports observed correctness in those trials, but it does not resolve instability on repeated identical inputs.
- Withhold any accuracy judgment until every possible invoice and repeat run has been tested.The sample supports a limited accuracy observation; universal testing is unnecessary to recognize the separate reliability concern.
- Call the model reliable because each tested invoice was correct at least once.Reliability requires dependable behavior across relevant runs and conditions, not merely one successful result per invoice.
The trap
Separate “was correct” from “behaves dependably when conditions or runs vary.” How to remember it
The sample supports observed accuracy, while repeated inconsistent outputs leave reliability unresolved for operational approval.
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 the Foundations of AI Governance questions
- Investigate supplier-specific impacts before concluding: What conclusion best fits trustworthy-AI reasoning? →
- Use the records for transparency and provide: Which distinction best guides the response? →
- Test safe behavior in unusual conditions as well as: Which evaluation is most appropriate? →
- All 337 Understanding the Foundations of AI Governance questions →
Part of the Certsqill AIGP question bank · Understanding the Foundations of AI Governance ·
Every answer, right and wrong, comes with its own explanation.