OFFICINA / ERRORBARSFIELD TOOL 01
Evaluation / uncertainty / decisions
Is that eval gain real?
Compare per-question scores with paired tests, cluster-aware uncertainty, and a correction for looking at several models. Every number is inspectable.
Runs in this browser. Selected files and text stay in this tab. The first visit downloads Python from jsDelivr. Methods & limits
Clears inputs, results, downloads and the Python worker.