Your invoice extraction uses tool use with strict JSON schemas. JSON syntax errors never occur, but 12% of…

Question

Your invoice extraction uses tool use with strict JSON schemas. JSON syntax errors never occur, but 12% of extractions fail semantic validation — for example, line item amounts don't match the extracted total, or vendor IDs don't match valid formats. These failures currently route to manual review. What's the most effective approach to reduce manual review volume while maintaining accuracy?

Accepted Answer

C. When validation fails, make a follow-up request with the document, extraction, and validation errors for model correction.

Answer

A. Retry the extraction up to 3 times when validation fails, accepting the first result that passes validation.

Answer

B. Implement post-processing logic that automatically corrects common errors, such as recalculating totals from line items when sums don't match.

Answer

D. Add stricter schema constraints with detailed field descriptions to prevent the model from generating invalid values initially.

Question 219