# Small-Data Virtual Metrology Replay Model Card

Version: 2026-08-21-v80

## Purpose

This browser lab teaches how a semiconductor engineer should read virtual metrology when physical measurements are sparse. The useful question is not "can the model predict?" in isolation; it is "when should we trust, guardband, sample more, or hold for review?"

## Dataset

- Rows: 2880
- Module contexts: cmp_thickness_polish, lithography_cd_overlay, plasma_etch_cd_bias, implant_sheet_resistance, ald_dielectric_thickness
- Sample plans: dense_baseline, periodic_skip, adaptive_uncertainty, excursion_confirm
- Data ages: fresh, aging, drifted
- Sensor health states: stable, noisy, shifted
- Physical metrology fractions: 0.1, 0.25, 0.5, 1
- Model families: mean_shift_baseline, ridge_surrogate, gaussian_process_proxy, conformal_guardband_proxy

## Public Boundary

- No wafer data, private metrology, tool log, recipe output, APC workflow, model training service, or quality disposition is exposed.
- No process recipe, sampling policy, release rule, foundry claim, product claim, or qualification decision.
- All Cpk, error, uncertainty, risk, and cost values are normalized educational proxies.

## Source Anchors

- NIST virtual metrology dynamic sampling: https://www.nist.gov/publications/comparative-study-semiconductor-virtual-metrology-methods-and-novel-algorithmic
- IRDS / NIST virtual metrology white paper: https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=924090
- NIST metrology and process-control uncertainty: https://www.nist.gov/publications/metrology-and-process-control-dealing-measurement-uncertainty
- NIST CHIPS Metrology Program: https://www.nist.gov/chips/research-development-programs/metrology-program
- NIST/SEMATECH process monitoring handbook: https://www.nist.gov/publications/nistsematech-engineering-statistics-handbook-chapter-6-process-or-product-monitoring
- NIST process capability handbook: https://www.itl.nist.gov/div898/handbook/pmc/section1/pmc16.htm