PBBQ: A Persian Bias Benchmark Dataset Curated with Human-AI Collaboration for Large Language Models

TL;DR

This paper introduces PBBQ, a large Persian bias benchmark dataset created with human-AI collaboration, to evaluate and address social biases in Persian language models, revealing significant biases and similarities to human stereotypes.

Contribution

The paper presents the first comprehensive Persian social bias benchmark dataset developed with social science expertise, enabling evaluation and mitigation of biases in Persian LLMs.

Findings

Current Persian LLMs exhibit significant social biases.

Models often replicate human bias patterns.

PBBQ provides a valuable resource for bias evaluation.

Abstract

With the increasing adoption of large language models (LLMs), ensuring their alignment with social norms has become a critical concern. While prior research has examined bias detection in various languages, there remains a significant gap in resources addressing social biases within Persian cultural contexts. In this work, we introduce PBBQ, a comprehensive benchmark dataset designed to evaluate social biases in Persian LLMs. Our benchmark, which encompasses 16 cultural categories, was developed through questionnaires completed by 250 diverse individuals across multiple demographics, in close collaboration with social science experts to ensure its validity. The resulting PBBQ dataset contains over 37,000 carefully curated questions, providing a foundation for the evaluation and mitigation of bias in Persian language models. We benchmark several open-source LLMs, a closed-source model, and Persian-specific fine-tuned models on PBBQ. Our findings reveal that current LLMs exhibit significant social biases across Persian culture. Additionally, by comparing model outputs to human responses, we observe that LLMs often replicate human bias patterns, highlighting the complex interplay between learned representations and cultural stereotypes.Upon acceptance of the paper, our PBBQ dataset will be publicly available for use in future work. Content warning: This paper contains unsafe content.