Do the Right Thing, Just Debias! Multi-Category Bias Mitigation Using LLMs

Amartya Roy; Danush Khanna; Devanshu Mahapatra; null Vasanthakumar; Avirup Das; Kripabandhu Ghosh

Do the Right Thing, Just Debias! Multi-Category Bias Mitigation Using LLMs

Amartya Roy, Danush Khanna, Devanshu Mahapatra, Vasanthakumar, Avirup Das, Kripabandhu Ghosh

Department of Computer Science

Research output: Preprint/Working paper › Preprint

Abstract

This paper tackles the challenge of building robust and generalizable bias mitigation models for language. Recognizing the limitations of existing datasets, we introduce ANUBIS, a novel dataset with 1507 carefully curated sentence pairs encompassing nine social bias categories. We evaluate state-of-the-art models like T5, utilizing Supervised Fine-Tuning (SFT), Reinforcement Learning (PPO, DPO), and In-Context Learning (ICL) for effective bias mitigation. Our analysis focuses on multi-class social bias reduction, cross-dataset generalizability, and environmental impact of the trained models. ANUBIS and our findings offer valuable resources for building more equitable AI systems and contribute to the development of responsible and unbiased technologies with broad societal impact.

Original language	English
Publication status	Published - 24 Sept 2024

Keywords

cs.CL

Access to Document

2409.16371v1

MCAIF: Centre for AI Fundamentals
Kaski, S. (PI), Alvarez, M. (Researcher), Pan, W. (Researcher), Mu, T. (Researcher), Rivasplata, O. (PI), Sun, M. (PI), Mukherjee, A. (PI), Caprio, M. (PI), Sonee, A. (Researcher), Leroy, A. (Researcher), Wang, J. (Researcher), Lee, J. (Researcher), Parakkal Unni, M. (Researcher), Sloman, S. (Researcher), Menary, S. (Researcher), Quilter, T. (Researcher), Hosseinzadeh, A. (PGR student), Mousa, A. (PGR student), Glover, E. (PGR student), Das, A. (PGR student), DURSUN, F. (PGR student), Zhu, H. (PGR student), Abdi, H. (PGR student), Dandago, K. (PGR student), Piriyajitakonkij, M. (PGR student), Rachman, R. (PGR student), Shi, X. (PGR student), Keany, T. (PGR student), Liu, X. (PGR student), Jiang, Y. (PGR student), Wan, Z. (PGR student), Harrison, M. (Support team), Machado, M. (Support team), Hartford, J. (PI), Kangin, D. (Researcher), Harikumar, H. (PI), Dubey, M. (PI), Parakkal Unni, M. (PI), Dash, S. P. (PGR student), Mi, X. (PGR student) & Barlas, Y. (PGR student)
1/10/21 → 30/09/26
Project: Research

Cite this

@techreport{4f77cea6638f4966afdf2b406cfd7a5c,

title = "Do the Right Thing, Just Debias! Multi-Category Bias Mitigation Using LLMs",

abstract = "This paper tackles the challenge of building robust and generalizable bias mitigation models for language. Recognizing the limitations of existing datasets, we introduce ANUBIS, a novel dataset with 1507 carefully curated sentence pairs encompassing nine social bias categories. We evaluate state-of-the-art models like T5, utilizing Supervised Fine-Tuning (SFT), Reinforcement Learning (PPO, DPO), and In-Context Learning (ICL) for effective bias mitigation. Our analysis focuses on multi-class social bias reduction, cross-dataset generalizability, and environmental impact of the trained models. ANUBIS and our findings offer valuable resources for building more equitable AI systems and contribute to the development of responsible and unbiased technologies with broad societal impact.",

keywords = "cs.CL",

author = "Amartya Roy and Danush Khanna and Devanshu Mahapatra and Vasanthakumar and Avirup Das and Kripabandhu Ghosh",

note = "17 pages, 5 Figures",

year = "2024",

month = sep,

day = "24",

language = "English",

type = "WorkingPaper",

}

TY - UNPB

T1 - Do the Right Thing, Just Debias! Multi-Category Bias Mitigation Using LLMs

AU - Roy, Amartya

AU - Khanna, Danush

AU - Mahapatra, Devanshu

AU - Vasanthakumar, null

AU - Das, Avirup

AU - Ghosh, Kripabandhu

N1 - 17 pages, 5 Figures

PY - 2024/9/24

Y1 - 2024/9/24

N2 - This paper tackles the challenge of building robust and generalizable bias mitigation models for language. Recognizing the limitations of existing datasets, we introduce ANUBIS, a novel dataset with 1507 carefully curated sentence pairs encompassing nine social bias categories. We evaluate state-of-the-art models like T5, utilizing Supervised Fine-Tuning (SFT), Reinforcement Learning (PPO, DPO), and In-Context Learning (ICL) for effective bias mitigation. Our analysis focuses on multi-class social bias reduction, cross-dataset generalizability, and environmental impact of the trained models. ANUBIS and our findings offer valuable resources for building more equitable AI systems and contribute to the development of responsible and unbiased technologies with broad societal impact.

AB - This paper tackles the challenge of building robust and generalizable bias mitigation models for language. Recognizing the limitations of existing datasets, we introduce ANUBIS, a novel dataset with 1507 carefully curated sentence pairs encompassing nine social bias categories. We evaluate state-of-the-art models like T5, utilizing Supervised Fine-Tuning (SFT), Reinforcement Learning (PPO, DPO), and In-Context Learning (ICL) for effective bias mitigation. Our analysis focuses on multi-class social bias reduction, cross-dataset generalizability, and environmental impact of the trained models. ANUBIS and our findings offer valuable resources for building more equitable AI systems and contribute to the development of responsible and unbiased technologies with broad societal impact.

KW - cs.CL

M3 - Preprint

BT - Do the Right Thing, Just Debias! Multi-Category Bias Mitigation Using LLMs

ER -

Do the Right Thing, Just Debias! Multi-Category Bias Mitigation Using LLMs

Abstract

Keywords

Access to Document

Fingerprint

Projects

MCAIF: Centre for AI Fundamentals

Cite this