ArabicCulturalQA: Cross-Dialectal Cultural Question Answering Benchmark

ArabicCulturalQA is a cross-dialectal Arabic cultural question-answering benchmark with parallel multiple-choice and open-ended formats. It covers Modern Standard Arabic, English, Egyptian, Levantine, Gulf, and Maghrebi dialects. The dataset accompanies an LREC 2026 paper and was created by QCRI.

Use Cases

• Benchmarking Arabic language models on cultural knowledge based on the described multiple-choice question format.
• Training models for open-ended cultural question answering based on the described open-ended question format.
• Evaluating model performance across different Arabic dialects based on the parallel Modern Standard Arabic, Egyptian, Levantine, Gulf, and Maghrebi versions.
• Studying cross-lingual transfer between English and Arabic for cultural QA tasks.

Strengths

• The MCQ test set has been reviewed and post-edited by native speakers of each dialect.
• Provides parallel data across six language/dialect variants: Modern Standard Arabic, English, Egyptian, Levantine, Gulf, and Maghrebi.
• Includes both multiple-choice and open-ended question formats for comprehensive evaluation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Native-speaker post-editing of the open-ended question test set is described as ongoing.

Provenance

Source

QCRI

Collection Method

Likely created as a research benchmark for an LREC 2026 paper.

Freshness

Last updated 2026-05-10 09:27:05; freshness should be verified.