Data and code for: Prompt-robustness tests mostly measure format compliance: evidence from 48,000 responses in a question-paired design of twelve language models

Scored per-cell data, analysis and scoring code, result tables and verification records for the ESWA pilot preprint: Prompt-robustness tests mostly measure format compliance: evidence from 48,000 responses in a question-paired design of twelve language models. This release excludes raw response databases, complete prompts and question text. Scored rows permit numerical reanalysis but not an independent end-to-end extraction audit. Code: MIT; author-generated data: CC BY 4.0. The package README specifies scope and limitations. No new paid inference was performed for this release.