Cross-Lingual Consistency Benchmark: Saudi PDPL and GDPR (Arabic–English)

A bilingual benchmark of 90 closed-form legal questions on the Saudi Personal Data Protection Law (PDPL) and the EU General Data Protection Regulation (GDPR), each posed in parallel Arabic and English, for measuring cross-lingual answer consistency of large language models: whether a model gives the same answer to the same legal question in both languages.

The record includes the questions with reference answers and legal basis, the evaluation script with the exact system prompts, raw outputs of three freely accessible models (Gemini 2.5 Flash, GPT-OSS-120B, Nemotron-3-Super-120B), and the scored results and summaries used in the paper. See README.md for file descriptions, settings, and instructions for reproducing the results.

This dataset accompanies: N. S. Alrajeh and M. A. Almasre, "Measuring What Accuracy Misses: Cross-Lingual Consistency of LLMs on the Saudi PDPL and the GDPR," MCAIT 2026, Madinah, Saudi Arabia (forthcoming in IEEE Xplore).