IslamTrust Benchmark: Multilingual LLM Evaluation for Islamic Values Alignment

A multilingual benchmark dataset designed to evaluate the alignment of large language models with general Islamic values and principles. Created by Abderraouf000 and last updated on December 21, 2025, it contains questions categorized by evaluation type, such as 'Other faiths', with multiple-choice answers and a correct choice index.

Use Cases

• Benchmarking LLM alignment with Islamic principles based on categorized questions.
• Evaluating multilingual model performance on ethical reasoning tasks based on the described question-answer structure.
• Fine-tuning language models for culturally-aware responses based on the provided correct answers.

Strengths

• Dataset was 'carefully' created according to the description.
• Provides a structured evaluation framework with categories like 'Other faiths'.

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

huggingface

Freshness

Last updated 2025-12-21 08:11:26