Overview of the NLPCC 2025 Shared Task: Gender Bias Mitigation Challenge

2025-06-14

AI Summary

To address the scarcity of Chinese gender-bias corpora and the susceptibility of existing models to biased data, this work introduces CORGI-PM—the first fine-grained Chinese benchmark for gender bias probing and mitigation (32.9k sentences, including 5.2k annotated bias sentence pairs with human-corrected counterparts). We propose a novel three-task evaluation framework—encompassing bias detection, bias classification, and generative debiasing—that jointly ensures cultural adaptability and end-to-end modeling capability. Methodologically, we integrate pretrained language model–driven bias detection with controllable rewriting, enhanced by multi-stage supervised fine-tuning and human verification. In public evaluations, our best-performing system achieves an F1 score of 89.2% on bias detection; post-mitigation sentences attain semantic fidelity and linguistic fluency exceeding 92%, respectively. This benchmark and framework establish a new standard for rigorous, culturally grounded gender-bias assessment and mitigation in Chinese NLP.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Ethics, Bias, and Fairness

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

As natural language processing for gender bias becomes a significant interdisciplinary topic, the prevalent data-driven techniques, such as pre-trained language models, suffer from biased corpus. This case becomes more obvious regarding those languages with less fairness-related computational linguistic resources, such as Chinese. To this end, we propose a Chinese cOrpus foR Gender bIas Probing and Mitigation (CORGI-PM), which contains 32.9k sentences with high-quality labels derived by following an annotation scheme specifically developed for gender bias in the Chinese context. It is worth noting that CORGI-PM contains 5.2k gender-biased sentences along with the corresponding bias-eliminated versions rewritten by human annotators. We pose three challenges as a shared task to automate the mitigation of textual gender bias, which requires the models to detect, classify, and mitigate textual gender bias. In the literature, we present the results and analysis for the teams participating this shared task in NLPCC 2025.

Problem

• Mitigate gender bias in Chinese NLP models
• Detect and classify gender-biased Chinese text
• Create corpus for gender bias probing in Chinese

Innovation

• Chinese corpus for gender bias probing
• High-quality labeled sentences annotation
• Automated gender bias detection and mitigation