« AI & LLM Testing

PDF documents

Every corpus article is also available as a plain single-column PDF, a quick way to run a basic RAG testing PDF ingestion smoke test. The three RAG testing documents below go further, each built to trip up a naive PDF-to-text pipeline in one specific, labeled way. A correct extraction should still produce clean, readable text from all of them.

• Table extraction — A real drawn table (an order log) that tests whether extraction keeps each row's cells together instead of reading columns out of order.
• Two-column layout — Two unrelated, newspaper-style essays placed side by side to test whether extraction respects column boundaries instead of reading straight across the page.
• Running headers & footers — A multi-page document with a repeated header and a page-numbered footer, testing whether extraction strips repeated boilerplate instead of injecting it mid-sentence.