KoSBi: A Dataset for Mitigating Social Bias Risks Towards Safer Large Language Model Application

Hwaran Lee, Seokhee Hong, Joonsuk Park, Takyoung Kim, Gunhee Kim +1 more
5/28/2023
cs.CL

No official code found. Best implementation: naver-ai/korean-safety-benchmarks (253 stars, MIT, last commit 2023-06-29).

Abstract

Large language models (LLMs) learn not only natural text generation abilities but also social biases against different demographic groups from real-world data. This poses a critical risk when deploying LLM-based applications. Existing research and resources are not readily applicable in South Korea due to the differences in language and culture, both of which significantly affect the biases and targeted demographic groups. This limitation requires localized social bias datasets to ensure the safe and effective deployment of LLMs. To this end, we present KO SB I, a new social bias dataset of 34k pairs of contexts and sentences in Korean covering 72 demographic groups in 15 categories. We find that through filtering-based moderation, social biases in generated content can be reduced by 16.47%p on average for HyperCLOVA (30B and 82B), and GPT-3.

View on arXivView PDF

Code Implementations(1)

naver-ai/korean-safety-benchmarksCommunity100%
MIT

Official datasets and pytorch implementation repository of SQuARe and KoSBi (ACL 2023)

25318PythonMay 12, 20233 years agoMIT
ailanguage-modelnlpsafetysocial-bias

Similar Papers

Bias and Fairness in Large Language Models: A Survey

90% match1 repo1622023

ROBBIE: Robust Bias Evaluation of Large Generative Language Models

89% match2023

Toward Fairness in Text Generation via Mutual Information Minimization based on Importance Sampling

88% match2023

KoBBQ: Korean Bias Benchmark for Question Answering

88% match2023

SocialStigmaQA: A Benchmark to Uncover Stigma Amplification in Generative Language Models

88% match2023

Evaluating LLMs for Demographic-Targeted Social Bias Detection: A Comprehensive Benchmark Study

88% match2025

Cite this paper

@article{lee2023kosbi,
  title  = {KoSBi: A Dataset for Mitigating Social Bias Risks Towards Safer Large Language Model Application},
  author = {Hwaran Lee and Seokhee Hong and Joonsuk Park and Takyoung Kim and Gunhee Kim and Jung-Woo Ha},
  year   = {2023},
  eprint = {2305.17701},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2305.17701v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2305.17701.svg)](https://codeofpaper.com/paper/2305.17701)

Discussion