Replication package for "Gender Attribution Under Workplace Role-and-Register Prompts in Contemporary Large Language Models"

This reproducibility package accompanies the manuscript “Gender Attribution Under Workplace Role-and-Register Prompts in Contemporary Large Language Models.” It contains 11,250 archived live API responses from nine large-language-model configurations, a reconstructed processed dataset, deterministic correction audit, collection and analysis code, an executed notebook, machine-readable statistical tables, and publication figures. The experimental grid comprises 25 workplace phrases and 50 stateless generations per model–phrase cell. The package supports the reported gender-attribution distributions, model-level risk differences, heterogeneity analyses, dependence-aware robustness checks, control-prompt audit, and name-diversity diagnostics. Name-based race classification is included only as a measurement-reliability analysis and should not be interpreted as verified racial or ethnic identity. The experimental contrast bundles workplace role content with communicative register and does not causally isolate occupational status.