ICLR 2026 - Institutional Affiliations Dataset & Analysis
Summary
An end-to-end pipeline converts 5,356 ICLR 2026 papers into a clean PDF-derived institutional-affiliations dataset with a treemap visualization. It avoids author profile drift by parsing the PDF title blocks for affiliations, canonicalizes institution names, and offers multiple counting methods; the project includes steps to reproduce the pipeline and generate public datasets and charts.