Towards Transfer Unlearning: Empirical Evidence of Cross-Domain Bias Mitigation
- URL: http://arxiv.org/abs/2407.16951v1
- Date: Wed, 24 Jul 2024 02:37:42 GMT
- Title: Towards Transfer Unlearning: Empirical Evidence of Cross-Domain Bias Mitigation
- Authors: Huimin Lu, Masaru Isonuma, Junichiro Mori, Ichiro Sakata,
- Abstract summary: We study an unlearning-based approach to debiasing in large language models (LLMs)
We propose a mask language modeling unlearning technique, which unlearns the harmful part of the text.
Experimental results demonstrate the effectiveness of our approach in diminishing bias while maintaining the language modeling abilities.
- Score: 18.150899267807965
- License: http://creativecommons.org/licenses/by/4.0/
- Abstract: Large language models (LLMs) often inherit biases from vast amounts of training corpora. Traditional debiasing methods, while effective to some extent, do not completely eliminate memorized biases and toxicity in LLMs. In this paper, we study an unlearning-based approach to debiasing in LLMs by performing gradient ascent on hate speech against minority groups, i.e., minimizing the likelihood of biased or toxic content. Specifically, we propose a mask language modeling unlearning technique, which unlearns the harmful part of the text. This method enables LLMs to selectively forget and disassociate from biased and harmful content. Experimental results demonstrate the effectiveness of our approach in diminishing bias while maintaining the language modeling abilities. Surprisingly, the results also unveil an unexpected potential for cross-domain transfer unlearning: debiasing in one bias form (e.g. gender) may contribute to mitigating others (e.g. race and religion).
Related papers
- Are Large Language Models Really Bias-Free? Jailbreak Prompts for Assessing Adversarial Robustness to Bias Elicitation [0.0]
Large Language Models (LLMs) have revolutionized artificial intelligence, demonstrating remarkable computational power and linguistic capabilities.
These models are inherently prone to various biases stemming from their training data.
This study explores the presence of these biases within the responses given by the most recent LLMs, analyzing the impact on their fairness and reliability.
arXiv Detail & Related papers (2024-07-11T12:30:19Z) - Evaluating Implicit Bias in Large Language Models by Attacking From a Psychometric Perspective [66.34066553400108]
We conduct a rigorous evaluation of Large Language Models' implicit bias towards certain groups by attacking them with carefully crafted instructions to elicit biased responses.
We propose three attack approaches, i.e., Disguise, Deception, and Teaching, based on which we built evaluation datasets for four common bias types.
arXiv Detail & Related papers (2024-06-20T06:42:08Z) - Likelihood-based Mitigation of Evaluation Bias in Large Language Models [39.77680080235204]
Large Language Models (LLMs) are widely used to evaluate natural language generation tasks as automated metrics.
It is possible that there might be a likelihood bias if LLMs are used for evaluation.
arXiv Detail & Related papers (2024-02-25T04:52:02Z) - Cognitive Bias in High-Stakes Decision-Making with LLMs [19.87475562475802]
We develop a framework designed to uncover, evaluate, and mitigate cognitive bias in large language models (LLMs)
Inspired by prior research in psychology and cognitive science, we develop a dataset containing 16,800 prompts to evaluate different cognitive biases.
We test various bias mitigation strategies, amidst proposing a novel method utilising LLMs to debias their own prompts.
arXiv Detail & Related papers (2024-02-25T02:35:56Z) - Self-Debiasing Large Language Models: Zero-Shot Recognition and
Reduction of Stereotypes [73.12947922129261]
We leverage the zero-shot capabilities of large language models to reduce stereotyping.
We show that self-debiasing can significantly reduce the degree of stereotyping across nine different social groups.
We hope this work opens inquiry into other zero-shot techniques for bias mitigation.
arXiv Detail & Related papers (2024-02-03T01:40:11Z) - GPTBIAS: A Comprehensive Framework for Evaluating Bias in Large Language
Models [83.30078426829627]
Large language models (LLMs) have gained popularity and are being widely adopted by a large user community.
The existing evaluation methods have many constraints, and their results exhibit a limited degree of interpretability.
We propose a bias evaluation framework named GPTBIAS that leverages the high performance of LLMs to assess bias in models.
arXiv Detail & Related papers (2023-12-11T12:02:14Z) - Diagnosing and Debiasing Corpus-Based Political Bias and Insults in GPT2 [0.0]
Training large language models (LLMs) on extensive, unfiltered corpora sourced from the internet is a common and advantageous practice.
Recent research shows that generative pretrained transformer (GPT) language models can recognize their own biases and detect toxicity in generated content.
This study investigates the efficacy of the diagnosing-debiasing approach in mitigating two additional types of biases: insults and political bias.
arXiv Detail & Related papers (2023-11-17T01:20:08Z) - Mitigating Biases for Instruction-following Language Models via Bias Neurons Elimination [54.865941973768905]
We propose a novel and practical bias mitigation method, CRISPR, to eliminate bias neurons of language models in instruction-following settings.
CRISPR automatically determines biased outputs and categorizes neurons that affect the biased outputs as bias neurons using an explainability method.
Experimental results demonstrate the effectiveness of our method in mitigating biases under zero-shot instruction-following settings without losing the model's task performance and existing knowledge.
arXiv Detail & Related papers (2023-11-16T07:16:55Z) - Exploring the Jungle of Bias: Political Bias Attribution in Language Models via Dependency Analysis [86.49858739347412]
Large Language Models (LLMs) have sparked intense debate regarding the prevalence of bias in these models and its mitigation.
We propose a prompt-based method for the extraction of confounding and mediating attributes which contribute to the decision process.
We find that the observed disparate treatment can at least in part be attributed to confounding and mitigating attributes and model misalignment.
arXiv Detail & Related papers (2023-11-15T00:02:25Z) - Mitigating Biases in Toxic Language Detection through Invariant
Rationalization [70.36701068616367]
biases toward some attributes, including gender, race, and dialect, exist in most training datasets for toxicity detection.
We propose to use invariant rationalization (InvRat), a game-theoretic framework consisting of a rationale generator and a predictor, to rule out the spurious correlation of certain syntactic patterns.
Our method yields lower false positive rate in both lexical and dialectal attributes than previous debiasing methods.
arXiv Detail & Related papers (2021-06-14T08:49:52Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.