AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

GeWu: A Culturally-Grounded Chinese Benchmark for Multi-Stage Social Bias Evaluation in Large Language Models

Yi Lin, Ziyi Zhou, Jiashi Gao, Xinwei Guo, Jiaxin Zhang, Haiyan Wu, Xin Yao, Xuetao Wei

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Bias measurement and mitigation · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
enhance CBBQ ambiguous contexts with LLM-generated cultural detail (GeWu)11 Chinese LLMs · Stage I multiple choicebias score: bias scores more than double on aggregate for most modelsImproved
enhance context for Qwen2.5-1.5B-Instruct (Age)Qwen2.5-1.5B-Instruct · Stage I Agebias score: 0.90 -> 0.97Improved
enhance context for Qwen3-1.7B (Religion)Qwen3-1.7B · Stage I Religionbias score: 0.75 -> 0.94Improved
enhanced contexts diversify questionsGeWu vs CBBQ · question pairsaverage ROUGE-L: 0.8825 -> 0.8025Improved
increase model size to 660B (DeepSeek-V3)Chinese LLMs · Stage I GeWubias scoreNo clear effect
switch to reasoning mode / reasoning-distilled modelsQwen3-14B, DeepSeek-R1-Distill-Qwen-14B · Stage I GeWubias scoreHurt
add second dialogue turn with enhanced cultural context11 LLMs on GeWu-1K · Stage II multi-turnbias score (DeepSeek-V3 classifier)Hurt
add second turn context for household registrationLLMs on GeWu-1K · Stage II Household registrationbias scoreHurt
small model scale (Qwen2.5-1.5B, Qwen3-1.7B) vs medium scaleQwen families · Stage I GeWubias score: Qwen3-1.7B approaching 100%Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy