Extended Thinking & Think Tags

Scenario: Structured Data Extraction You are building a structured data extraction system using Claude. The system extracts information from unstructured documents, validates output using JSON schemas, and maintains high accuracy. It must handle edge cases gracefully and integrate with downstream systems. Your extractor captures sample_size , study_design , and citation_source from journal articles. It works on papers with explicit Methods and References sections, but misses the same fields when sample details appear in an abstract footnote, study design appears in an appendix table, or citations are inline only. Validation passes because those fields are nullable, but downstream analytics show systematic under-extraction. What change would most effectively improve extraction quality?