Benchmark measures cultural values in large language models reliably
GYROval: A Robust Benchmark for Cultural Value Orientation in Large Language Models
Human-Computer InteractionArtificial IntelligenceComputers and Society
Summary
This work addresses how well large language models reflect cultural values by creating a new test called GYROval. The test uses paired scenarios that don’t have a right answer but reflect different cultural attitudes, measuring where models tend to lean on two cultural value scales. The authors tested twenty models, including versions in English and Russian, to check how stable and consistent these cultural leanings are across languages and settings. This benchmark is publicly available for others to use.
What this means in practice
- •For language model developers: Assess and compare cultural biases in language models to improve global usability and fairness.
- •For cross-cultural nlp teams: Evaluate how well large language models reflect cultural values across languages and domains for better localization.
Authors
Alexander Didenko, Anna Shabanova, Vladislav Zapylikhin, Alexander Antipov, Ruslana Raemgulova
Abstract
We present a robust benchmark for measuring cultural value orientation in large language models on the two Inglehart-Welzel axes over several domains and roles (hence GYROval - Gridded Yielding of Robust value Orientation), together with the results of administering it to twenty models. Items are binary contrastive scenarios in the sense introduced by CDEval: both options are legitimate courses of action, neither is correct, there is no answer key, and a model's score on an axis is the proportion of its responses falling on the counted pole. Eleven of the twenty models were additionally administered a paired Russian translation of the identical items and a second sampling temperature. The instrument is publicly released in both languages. Stability was assessed by treating the vignette as the unit of analysis, ranking the models within the levels of each perturbation factor, and summarising the agreement between levels by tie-corrected Kendall's \emph{W} against an empirical permutation null.