Improving the reproducibility of patent evidence: a repair-first pipeline for AI trust infrastructure during the USPTO Open Data Portal transition — data and code

Versioned reviewer-facing exports (v0.1 negative-result before-state; v0.2 ODP-repaired; v0.3 family/comparator), examiner-citation edge table, next-wave seed resolution table, corpus-scale repair outputs (USPTO ODP file-wrapper dates, EPO OPS INPADOC family identifiers), embedding/LLM triage outputs with the stratified calibration sample, and all repair and triage scripts accompanying the article submitted to World Patent Information. Derived metadata fields only; raw provider payloads remain subject to each provider's terms.