Challenge: Recent research has shown that large-scale pretrained language models exhibit issues relating to racism, sexism, religion bias, and toxicity in general.
Approach: They propose a gray-box method for debiasing pretrained masked language models using partitioned contrastive gradient unlearning (PCGU) aims to optimize only the weights that contribute most to a specific domain of bias by computing a first-order approximation based on the gradients of contrastive sentence pairs.
Outcome: The proposed method is low-cost and can pinpoint the sources of social bias in large pretrained language models.

Similar Papers

Can Machine Unlearning Reduce Social Bias in Language Models? (2024.emnlp-industry)

Copied to clipboard

Challenge: Existing methods for mitigating bias in language models are expensive and time-consuming . comparative studies have not evaluated their respective advantages and disadvantages .
Approach: They propose to use Partitioned Contrastive Gradient Unlearning and Negation via Task Vector to reduce social biases in open-source language models.
Outcome: The proposed methods outperform PCGU and DPO in debiasing models . the proposed methods can be easily tuned to balance the trade-off between bias reduction and generation quality .
Auto-Debias: Debiasing Masked Language Models with Automated Biased Prompts (2022.acl-long)

Copied to clipboard

Challenge: Existing methods to mitigate human-like biases in pretrained language models are based on external corpora and require a distribution alignment loss to mitigate them.
Approach: They propose an automatic method to mitigate biases in pretrained language models by searching for biased prompts such that cloze-style completions are the most different with respect to different demographic groups.
Outcome: The proposed method reduces biases in pretrained language models, including gender and racial bias, and improves fairness of the models.
Projective Methods for Mitigating Gender Bias in Pre-trained Language Models (2024.lrec-main)

Copied to clipboard

Challenge: Mitigating gender bias in NLP has a long history tied to debiasing static word embeddings.
Approach: They propose a masked language modelling task where content is developed around known social stereotypes and a projective debiasing method is used to reduce bias.
Outcome: The proposed methods reduce intrinsic bias and mitigat observed bias in a downstream setting, but the two outcomes are not necessarily correlated.
From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trails of Political Biases Leading to Unfair NLP Models (2023.acl-long)

Copied to clipboard

Challenge: Hundreds of studies have highlighted ethical issues in NLP models .
Approach: They propose to measure media biases in LMs trained on diverse data sources . they focus on hate speech and misinformation detection .
Outcome: The proposed methods quantify the fairness of downstream NLP models trained on politically biased LMs.
Mitigating Biases in Language Models via Bias Unlearning (2025.emnlp-main)

Copied to clipboard

Challenge: Recent debiasing approaches target different demographic groups, harming fairness and discrimination.
Approach: They propose a model debiasing framework which targets stereotypes by unlearning stereotype forgetting and anti-stereotype retention.
Outcome: The proposed framework outperforms existing methods in mitigating bias while retaining language modeling capabilities.
Upstream Mitigation Is Not All You Need: Testing the Bias Transfer Hypothesis in Pre-Trained Language Models (2022.acl-long)

Copied to clipboard

Challenge: Large language models and other massively pre-trained "foundation" models can easily adapt to a wide variety of downstream tasks in a process called finetuning.
Approach: They propose to use the bias transfer hypothesis to reduce social biases internalized by large language models during pre-training into harmful task-specific behavior after fine-tuning.
Outcome: The bias transfer hypothesis is the theory that social biases internalized by large language models during pre-training transfer into harmful task-specific behavior after fine-tuning.
StereoSet: Measuring stereotypical bias in pretrained language models (2021.acl-long)

Copied to clipboard

Challenge: Existing literature on stereotypical biases in language models is limited . current evaluations focus on measuring bias without considering language modeling ability .
Approach: They propose to measure stereotypical biases in four domains: gender, profession, race, and religion . they compare stereotypical and language modeling ability of popular models like BERT, GPT-2, RoBERTa and XLnet .
Outcome: The proposed model shows strong stereotypical biases in gender, profession, race, and religion domains.
Debiasing Large Language Models with Structured Knowledge (2024.findings-acl)

Copied to clipboard

Challenge: Existing methods to reduce biases in pre-training models are hampered by their performance.
Approach: They propose a method that utilizes structured knowledge to mitigate bias in LLMs . their method obviates the need for training from scratch, thus offering enhanced scalability .
Outcome: The proposed method outperforms state-of-the-art (SOTA) baselines in the debiasing ability.
Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP (2021.tacl-1)

Copied to clipboard

Challenge: Pretrained language models pick up and reproduce undesirable biases when trained on large, unfiltered crawls from the Internet.
Approach: They propose a decoding algorithm that, given only a textual description of the undesired behavior, reduces the probability of a language model producing problematic text.
Outcome: The proposed approach reduces the probability of a language model producing problematic text by giving only a textual description of the undesired behavior.
Toward Inclusive Language Models: Sparsity-Driven Calibration for Systematic and Interpretable Mitigation of Social Biases in LLMs (2025.findings-emnlp)

Copied to clipboard

Challenge: a new method to mitigate stereotypical bias in large language models is needed . inherent biases from training on vast Internet datasets can amplify harmful stereotypes .
Approach: They propose a method to identify stereotypical bias in decoder-only transformer models . they apply a localization mechanism that correlates internal activations with a new Context Influence score .
Outcome: The proposed method reduces stereotypical biases on BBQ, StereoSet, and CrowS-Pairs while improving reasoning performance on MMLU by 10%.

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