Challenge: Large pre-trained language models contain societal biases and carry along these biase . Current approaches to mitigate these bias impose debiasing by updating model parameters, effectively transferring model to irreversible debiased state.
Approach: They propose to develop stand-alone debiasing functionalities separate from the model, which can be integrated into the model on-demand while keeping the core model untouched.
Outcome: The proposed approach improves or maintains effectiveness of bias mitigation, avoids catastrophic forgetting in a multi-attribute scenario, and maintains on-par task performance while granting parameter-efficiency and easy switching between the original and debiased models.

Similar Papers

Modular and On-demand Bias Mitigation with Attribute-Removal Subnetworks (2023.findings-acl)

Copied to clipboard

Challenge: Existing studies show that pre-trained language models can be used to mitigate societal biases and stereotypes.
Approach: They propose a modular bias mitigation approach that integrates debiasing modules into the core model on-demand at inference time.
Outcome: The proposed approach improves on-par with baseline finetuning on gender, race, and age protected attributes on three classification tasks with gender, age, and race as protected attributes.
PEFTDebias : Capturing debiasing information using PEFTs (2023.emnlp-main)

Copied to clipboard

Challenge: Recent research suggests that biases acquired during pretraining can propagate to downstream models, resulting in superficial text dependencies and potential implicit bias.
Approach: They propose a parameter-efficient fine-tuning approach to mitigate implicit biases within foundation models by incorporating parameters into the model and freezing them during the fine-uning process.
Outcome: The proposed method reduces biases in foundation models by incorporating parameters and freezing them during fine-tuning.
On Transferability of Bias Mitigation Effects in Language Model Fine-Tuning (2021.naacl-main)

Copied to clipboard

Challenge: PTLMs can exhibit biases against protected groups in a host of modeling tasks . but, fine-tuned LMs may propagate bias to downstream classifiers .
Approach: They propose to use upstream bias mitigation techniques to reduce bias on downstream tasks by fine-tuning an upstream model and applying it to a downstream model.
Outcome: The proposed model reduces bias on hate speech detection, toxicity detection and coreference resolution tasks over bias factors.
Effective Controllable Bias Mitigation for Classification and Retrieval using Gate Adapters (2024.eacl-long)

Copied to clipboard

Challenge: a recent study focused on learning separate modules for on-demand debiasing.
Approach: They propose a modular debiasing module with a controllable gate adapter . they demonstrate that the module can reduce the bias of search results .
Outcome: The proposed module can reduce biases on three classification tasks with four protected attributes while maintaining higher task performance.
An Empirical Analysis of Parameter-Efficient Methods for Debiasing Pre-Trained Language Models (2023.acl-long)

Copied to clipboard

Challenge: Pre-trained language models inherit more human-like biases from the training corpora, causing computationally expensive problems.
Approach: They propose parameter-efficient methods in combination with counterfactual data augmentation for bias mitigation.
Outcome: The proposed methods are effective in mitigating gender bias, prompt tuning is more suitable for GPT-2 than BERT, and less effective when it comes to racial and religious bias.
Sustainable Modular Debiasing of Language Models (2021.findings-emnlp)

Copied to clipboard

Challenge: Existing debiasing methods modify all of the PLM parameters, which is costly and leads to (catastrophic) forgetting of useful language knowledge.
Approach: They propose a modular debiasing approach based on dedicated adapters that inject adapter modules into the original PLM layers and update only the adapters.
Outcome: The proposed approach is based on dedicated adapters and retains fairness even after large-scale training.
Open-DeBias: Toward Mitigating Open-Set Bias in Language Models (2025.findings-emnlp)

Copied to clipboard

Challenge: Existing approaches to addressing harmful biases in LLMs are limited to predefined categories . a novel, data-efficient, and parameter-efficient debiasing method is proposed to mitigate existing social and stereotypical biase .
Approach: They propose an open-set bias detection and mitigation method to address harmful biases in text-based QA.
Outcome: The proposed method improves QA accuracy on Korean BBQ dataset by nearly 48% on ambiguous subsets and 6% on disambiguated ones.
Co2PT: Mitigating Bias in Pre-trained Language Models through Counterfactual Contrastive Prompt Tuning (2023.findings-emnlp)

Copied to clipboard

Challenge: Pre-trained language models can encode unfair social biases from large pre-training corpora and even amplify biase in downstream applications.
Approach: They propose a *debias-while-prompt tuning* method for mitigating biases via counterfactual contrastive prompt tuning on downstream tasks.
Outcome: The proposed method can mitigate biases on three extrinsic bias benchmarks and adapt to existing debiased language models.
When and Why Does Bias Mitigation Work? (2023.findings-emnlp)

Copied to clipboard

Challenge: Neural models exploit shallow surface features to perform language understanding tasks, rather than learning the deeper language understanding and reasoning skills that practitioners desire.
Approach: They propose to use model debiasing techniques to pressure models away from spurious features and to use them to learn useful representations instead.
Outcome: The proposed methods increase models' reliance on hidden biases instead of learning robust features that help them solve a task.
How Far Can It Go? On Intrinsic Gender Bias Mitigation for Text Classification (2023.eacl-main)

Copied to clipboard

Challenge: a growing interest in exploring how gender bias pertains in contextualized language models has been generated . intrinsic mitigation strategies and bias metrics have been proposed to mitigate gender bias in contextualised language models .
Approach: They propose to use different intrinsic bias mitigation strategies to mitigate gender bias in contextualized language models.
Outcome: The proposed probe shows that some mitigation techniques can hide gender bias . the probe also shows that not all mitigation techniques fool extrinsic bias despite their use .

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations