Large language models (LLMs) have achieved impressive performance on various natural language generation tasks. Nonetheless, they suffer from generating negative and harmful contents that are biased against certain demographic groups (e.g., female), raising severe fairness concerns. As remedies, prior works intervened the generation by removing attitude or demographic information, inevitably degrading the generation quality and resulting in notable \textit{fairness-fluency} trade-offs. However, it is still under-explored to what extent the fluency \textit{has to} be affected in order to achieve a desired level of fairness. In this work, we conduct the first formal study from an information-theoretic perspective. We show that previous approaches are excessive for debiasing and propose LIDAO, a general framework to debias a (L)LM at a better fluency provably. We further robustify LIDAO in adversarial scenarios, where a carefully-crafted prompt may stimulate LLMs exhibiting instruction-following abilities to generate texts with fairness issue appears only when the prompt is also taken into account. Experiments on three LMs ranging from 0.7B to 7B parameters demonstrate the superiority of our method.
Information Gain-Guided Causal Intervention for Autonomous Debiasing Large Language Models
Research Center for Social Computing and Interactive Robotics（社会计算与交互机器人研究中心）；Harbin Institute of Technology（哈尔滨工业大学）；Beijing Academy of Artificial Intelligence（北京人工智能研究院）
Can AI Debias the News? LLM Interventions Improve Cross-Partisan Receptivity but LLMs Overestimate Their Own Effectiveness
Department of Experimental Psychology, University of Oxford（心理学实验系，牛津大学）；Uehiro Oxford Institute, University of Oxford（牛津大学欧胡罗研究所）；Psychology Programme, School of Social Sciences, Nanyang Technological University（南洋理工大学社会科学学院心理学项目）；Department of Communication and Culture, BI Norwegian Business School（BI挪威商学院传播与文化系）
Latent space bias directions in LLMs capture confidence, not fairness
Institute of Astronomy & Kavli Institute for Cosmology, University of Cambridge（剑桥大学天文学研究所及卡夫利宇宙学研究所）；Risk and Security AI Lab, Visa Inc.（Visa公司风险与安全人工智能实验室）
Deep learning models often learn to make predictions that rely on sensitive social attributes like gender and race, which poses significant fairness risks, especially in societal applications, e.g., hiring, banking, and criminal justice. Existing work tackles this issue by minimizing information about social attributes in models for debiasing. However, the high correlation between target task and social attributes makes bias mitigation incompatible with target task accuracy. Recalling that model bias arises because the learning of features in regard to bias attributes (i.e., bias features) hel…
Stony Brook University（石溪大学）；Los Altos High School（洛斯阿尔托斯高中）
Understanding the cumulative effect of multiple fairness enhancing interventions at different stages of the machine learning (ML) pipeline is a critical and underexplored facet of the fairness literature. Such knowledge can be valuable to data scientists/ML practitioners in designing fair ML pipelines. This paper takes the first step in exploring this area by undertaking an extensive empirical study comprising 60 combinations of interventions, 9 fairness metrics, 2 utility metrics (Accuracy and F1 Score) across 4 benchmark datasets. We quantitatively analyze the experimental data to measure th…
How Does Alignment Tuning Shape Representations of Sycophancy and Related Cue-Induced Biases in LLMs?
University of Michigan（密歇根大学）；Jinesis Lab, University of Toronto & Vector Institute（多伦多大学Jinesis实验室和向量研究所）；Max-Planck Institute for Intelligent Systems（马克斯·普朗克智能系统研究所）；ELLIS Institute Tübingen（图宾根ELLIS研究所）
University of Chinese Academy of Sciences（中国科学院大学）；Institute of Software, Chinese Academy of Sciences（中国科学院软件研究所）；Beijing Academy of Artificial Intelligence（北京智源人工智能研究院）
Prompt-based probing has been widely used in evaluating the abilities of pretrained language models (PLMs). Unfortunately, recent studies have discovered such an evaluation may be inaccurate, inconsistent and unreliable. Furthermore, the lack of understanding its inner workings, combined with its wide applicability, has the potential to lead to unforeseen risks for evaluating and applying PLMs in real-world applications. To discover, understand and quantify the risks, this paper investigates the prompt-based probing from a causal view, highlights three critical biases which could induce biased…
Mitigating Social Bias in Large Language Models: A Multi-Objective Approach within a Multi-Agent Framework
School of Software Engineering, Sun Yat-sen University（中山大学软件工程学院）；School of Physics and Astronomy, Sun Yat-sen University（中山大学物理与天文学院）；School of Cyber Science and Technology, Zhejiang University（浙江大学网络空间学院）；Department of Computer Science, City University of Hong Kong（香港城市大学计算机科学系）
Breaking the Likelihood Trap: Variance-Calibrated Modulation for Large Language Model Decoding
School of Computer and Information Engineering, Henan University（河南大学计算机与信息工程学院）；Department of Statistics, LMU Munich（慕尼黑大学统计系）；Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）
