I have updated the documentation of a reproducible review pipeline for submitted cheat-sheet drafts in my introductory statistics course.

The project starts from a practical teaching problem: students submit handwritten cheat-sheet drafts in very heterogeneous formats, including PDFs, scans, photos, ZIP archives, and explanatory notes. To make the review process transparent, the pipeline standardizes the files, separates content pages from explanatory pages, measures page usage, flags identical or very similar submissions, and estimates topic coverage with a vision-capable language model.

The documentation is intentionally process-oriented. It treats the early cheat-sheet submission as a formative learning event: students need to select, structure, condense, and justify course material several weeks before the exam. The analysis is not meant to automatically decide about individual students. It provides an auditable basis for manual review, and the published report only uses aggregate results.

This is also an example of what I mean by public practice: a real teaching and assessment workflow made visible as a reproducible data pipeline, including its operational constraints, safeguards, and documentation.