S6 · scenario95 questions · 18 free
Structured data extraction (S6)
Extract from unstructured docs, validate with JSON schema, handle edge cases.
The Structured Data Extraction scenario (S6) is pulling clean, validated data out of messy documents: schema-enforced output, validation, and the edge cases that break naive extractors. The exam tests precision and recovery, not just a lucky parse.
Expect questions on enforcing structured output with JSON schemas, designing prompts with explicit criteria to reduce false positives, building validation, retry, and feedback loops, and batching extraction efficiently at volume.
This scenario spans 12 subtopic areas, covered by 95 practice questions across 26 easy, 46 medium, and 23 hard items.
Sample question · free
Your extraction system uses a JSON schema to validate output but still flags too many borderline fields as present. The system prompt says 'be conservative with low-confidence extractions.' Precision remains poor. What change will most directly reduce false positives?
What's covered
Subtopic areas in Structured data extraction, drawn from the exam blueprint:
4.1Design prompts with explicit criteria to improve precision and reduce false positives134.2Apply few-shot prompting to improve output consistency and quality124.3Enforce structured output using tool use and JSON schemas124.4Implement validation, retry, and feedback loops for extraction quality104.5Design efficient batch processing strategies114.6Design multi-instance and multi-pass review architectures105.1Manage conversation context to preserve critical information across long interactions45.2Design effective escalation and ambiguity resolution patterns55.3Implement error propagation strategies across multi-agent systems55.4Manage context effectively in large codebase exploration45.5Design human review workflows and confidence calibration35.6Preserve information provenance and handle uncertainty in multi-source synthesis6