The Shibboleth Effect: LLM Cross-Lingual Bias Audit Data

Replication data for a study auditing the cross-lingual distributional skew of large language models. The repository contains a complete computational pipeline, scenario briefings, simulation logs, and analysis scripts. Hakan Mehmetcik published the data under a CC-BY-4.0 license on figshare in May 2026.

Use Cases

• Auditing cross-lingual bias in LLMs based on scenario briefings and simulation logs
• Reproducing computational social science experiments based on the provided pipeline and scripts
• Analyzing distributional skew in model outputs based on the simulation logs

Strengths

• Includes a complete computational pipeline for full reproducibility
• Published under a permissive CC-BY-4.0 license
• Contains multiple file formats (MD, PY, TXT, CSV, JSON, JSONL) for different components

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download
• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

figshare

Collection Method

Likely generated from computational simulations and experiments as part of a research study.

Freshness

Last updated 2026-05-24 00:42:35