UPDATE: Added benchmarking results, bug fixes related to scoring & Export to PDF option for Scoring Methodology

[uchandar29]

Summary

Adds a one-click PDF export to the Scoring Methodology reference modal, fixes a bug where high label-noise rates could produce NaN instead of a floored score, and refreshes the benchmarking results data for image classification datasets.

What's included

• Export to PDF: downloads the reference document directly, no print dialog. Built with html2canvas + jsPDF instead of the browser's native print-to-PDF so it's a single click. A previous print-based version is kept as a secondary Ctrl+P option.
• Scoring fix: the label-noise formula could go negative before a fractional exponent, which is NaN in JS rather than a floor — now clamped, matching an update to SCORING_FORMULAS.md. Also guards the overall score against stray NaN/null axis values.
• Data: refreshed image classification benchmarking results; synced package-lock.json.