Changes in version 1.2.0 (2026-09-23) A correctness, speed and usability release. Detection changes were measured article by article on every labeled set (data-raw/benchmark/snapshot.R and evaluate.R); behavior-preserving changes were verified to give identical predictions on 3124 cached articles. Bug fixes - Plain-text results now have the same columns as XML results. rt_fund() returned is_funded_pred/funding_text, so rt_summary() silently dropped funding from plain-text output; it now returns is_fund_pred/fund_text (the old names remain as deprecated copies for one release). pmid is NA when the file name has no PMID (it used to be the whole path), article is always the file name, rt_coi() returns a tibble, and rt_data_code() returns the PMID and the extracted links. - Namespaced XML. With the default remove_ns = FALSE, an article whose
element declared a default namespace (as OAI-PMH records do) gave all-FALSE indicators with is_success = TRUE. Namespaces are now always removed; remove_ns is accepted and ignored. - PMC identifiers. Current PMC XML tags the PMCID as pub-id-type = "pmcid"; only the older "pmc" was read, so pmcid_pmc and pmcid_uid were empty for every current article. Both are read now. - rt_all_pmc() agrees with the standalone detectors. It removed citation markers and tables from the XML before the AI, data/code and reporting detectors ran, and all_meta = TRUE removed /