LLM Bias

Large language models (LLMs) often exhibit biases reflecting societal prejudices, leading to unfair or discriminatory outputs. Current research focuses on developing methods to detect and mitigate these biases, encompassing both implicit and explicit forms across various protected attributes like race, gender, and age, using techniques such as prompt engineering, attention mechanism analysis, and counterfactual evaluations applied to models like GPT-3.5 and others. Understanding and addressing LLM bias is crucial for ensuring fairness and ethical deployment of these powerful technologies, impacting both the development of responsible AI and the avoidance of harmful societal consequences.

Papers

October 31, 2024

Desert Camels and Oil Sheikhs: Arab-Centric Red Teaming of Frontier LLMs
Muhammed Saeed, Elgizouli Mohamed, Mukhtar Mohamed, Shaina Raza, Shady Shehata, Muhammad Abdul-Mageed
Large Language Model Language Model Absolute Stance Bias Bias Mitigation Red Teaming Electric SHEEP LLM Bias

October 28, 2024

LLMs are Biased Evaluators But Not Biased for Retrieval Augmented Generation
Yen-Shan Chen, Jing Jin, Peng-Ting Kuo, Chao-Wei Huang, Yun-Nung Chen
Large Language Model Faithful Generation Retrieval Augmented LLM Bias Fact Verification Task

October 21, 2024

How Can We Diagnose and Treat Bias in Large Language Models for Clinical Decision-Making?
Kenza Benkirane, Jackie Kay, Maria Perez-Ortiz
Bias Mitigation Bias Evaluation Racial Bias Clinical Language LLM Bias

October 20, 2024

October 17, 2024

LLMs are Biased Teachers: Evaluating LLM Bias in Personalized Education
Iain Weissburg, Sathvika Anand, Sharon Levy, Haewon Jeong
Large Language Model Inherent Bias LLM Bias Bias Score Source Bias Personalized Education

October 14, 2024

Assessing Bias in Metric Models for LLM Open-Ended Generation Bias Benchmarks
Nathaniel Demchak, Xin Guan, Zekun Wu, Ziyi Xu, Adriano Koshiyama, Emre Kazim
New Benchmark Social Bias Stereotype Content Bias Evaluation Inherent Bias Demographic Attribute Bias Benchmark LLM Bias

October 10, 2024

Human and LLM Biases in Hate Speech Annotations: A Socio-Demographic Analysis of Annotators and Targets
Tommaso Giorgi, Lorenzo Cima, Tiziano Fagni, Marco Avvenuti, Stefano Cresci
Hate Speech Human Generated Individual Annotator Socio Demographic LLM Bias

July 15, 2024

An Actionable Framework for Assessing Bias and Fairness in Large Language Model Use Cases
Dylan Bouchard
Large Language Model Procedural Fairness Absolute Stance Bias Fairness Metric LLM Integrated Application LLM Bias Fair Credit Practical Framework Language Model Application

July 14, 2024

BiasAlert: A Plug-and-play Tool for Social Bias Detection in LLMs
Zhiting Fan, Ruizhe Chen, Ruiling Xu, Zuozhu Liu
Large Language Model Bias Mitigation Social Bias Plug and Play BIAS Detection LLM Bias

July 1, 2024

The African Woman is Rhythmic and Soulful: Evaluation of Open-ended Generation for Implicit Biases
Serene Lim
Global Evaluation Implicit Bias Bias Metric LLM Bias Open Ended Generation African American

June 23, 2024

Found in the Middle: Calibrating Positional Attention Bias Improves Long Context Utilization
Cheng-Yu Hsieh, Yung-Sung Chuang, Chun-Liang Li, Zifeng Wang, Long T. Le, Abhishek Kumar, James Glass, Alexander Ratner, Chen-Yu Lee, Ranjay Krishna, Tomas Pfister
Large Language Model Low Priority Input Context LLM Bias Attention Bias Long Input Context

May 31, 2024

May 29, 2024

Quantitative Certification of Bias in Large Language Models
Isha Chaudhary, Qian Hu, Manoj Kumar, Morteza Ziyadi, Rahul Gupta, Gagandeep Singh
Absolute Stance Bias Bias Label LLM Bias Novel Certification

May 7, 2024

The Silicon Ceiling: Auditing GPT's Race and Gender Biases in Hiring
Lena Armstrong, Abbey Liu, Stephen MacNeil, Danaë Metaxa
Gender Bias Social Bias LLM Bias Digital Divide Resume Screening

April 4, 2024

The Impact of Unstated Norms in Bias Analysis of Language Models
Farnaz Kohankhaki, D. B. Emerson, Jacob-Junqi Tian, Laleh Seyyed-Kalantari, Faiza Khan Khattak
Large Language Model Language Model Global Impact Absolute Stance Bias Stereotype Content Bias Evaluation Learning Norm LLM Bias Bias Measurement

February 18, 2024

Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinement
Wenda Xu, Guanglei Zhu, Xuandong Zhao, Liangming Pan, Lei Li, William Yang Wang
Large Language Model Text Generation Self Feedback Self Refinement LLM Bias Self Critique Pipeline

February 6, 2024

Measuring Implicit Bias in Explicitly Unbiased Large Language Models
Xuechunzi Bai, Angelina Wang, Ilia Sucholutsky, Thomas L. Griffiths
Social Bias Implicit Bias LLM Bias

December 24, 2023

A Group Fairness Lens for Large Language Models
Guanqun Bi, Lei Shen, Yuqiang Xie, Yanan Cao, Tiangang Zhu, Xiaodong He
Large Language Model Chain of Thought Group Fairness LLM Bias

LLM Bias

Papers

Desert Camels and Oil Sheikhs: Arab-Centric Red Teaming of Frontier LLMs

LLMs are Biased Evaluators But Not Biased for Retrieval Augmented Generation

How Can We Diagnose and Treat Bias in Large Language Models for Clinical Decision-Making?

Hey GPT, Can You be More Racist? Analysis from Crowdsourced Attempts to Elicit Biased Content from Generative AI

A Novel Interpretability Metric for Explaining Bias in Language Models: Applications on Multilingual Models from Southeast Asia

LLMs are Biased Teachers: Evaluating LLM Bias in Personalized Education

Assessing Bias in Metric Models for LLM Open-Ended Generation Bias Benchmarks

Human and LLM Biases in Hate Speech Annotations: A Socio-Demographic Analysis of Annotators and Targets

An Actionable Framework for Assessing Bias and Fairness in Large Language Model Use Cases

BiasAlert: A Plug-and-play Tool for Social Bias Detection in LLMs

The African Woman is Rhythmic and Soulful: Evaluation of Open-ended Generation for Implicit Biases

Found in the Middle: Calibrating Positional Attention Bias Improves Long Context Utilization

UniBias: Unveiling and Mitigating LLM Bias through Internal Attention and FFN Manipulation

Generative AI Voting: Fair Collective Choice is Resilient to LLM Biases and Inconsistencies

Quantitative Certification of Bias in Large Language Models

The Silicon Ceiling: Auditing GPT's Race and Gender Biases in Hiring

The Impact of Unstated Norms in Bias Analysis of Language Models

Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinement

Measuring Implicit Bias in Explicitly Unbiased Large Language Models

A Group Fairness Lens for Large Language Models