Weicheng Ma, Henry Scheible, Brian Wang, Goutham Veeramachaneni, Pratim Chowdhary, Alan Sun, Andrew Koulogeorge, Lili Wang, Diyi Yang, Soroush Vosoughi
| Challenge: | Current approaches for examining stereotypes in PLMs require intricate human knowledge about these stereotypes and entail careful manual curation of examples. |
| Approach: | They propose a framework for examining stereotype-encoding behavior of PLMs using model probing and textual analyses. |
| Outcome: | The proposed approach can debiase PLMs without compromising their language modeling capabilities or performance. |
Similar Papers
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%. |
Debiasing Pretrained Text Encoders by Paying Attention to Paying Attention (2022.emnlp-main)
Copied to clipboard
| Challenge: | Recent research has exposed text encoders for replicating discriminatory social biases which may cause unintended and undesired model behaviors with respect to social groups. |
| Approach: | They propose a method to reduce social stereotypes by redistributing attention scores of a text encoder so it forgets any preference to historically advantaged groups and attends to all social classes with the same intensity. |
| Outcome: | The proposed method reduces stereotypes and inflicts no semantic damage on pre-trained encoders. |
Stepmothers are mean and academics are pretentious: What do pretrained language models learn about you? (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies on "gender bias" and "racial bias" focus on stereotypical attributes of word representations . a new method to elicit stereotypical information is proposed to capture stereotypical traits in language models . |
| Approach: | They propose a method to elicit stereotypical information from pretrained language models . they use fine-tuning on news sources to study their emotional effects . |
| Outcome: | The proposed method can be used to analyze emotion and stereotype shifts due to linguistic experience using fine-tuning on news sources. |
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. |
Quantifying Stereotypes in Language (2024.eacl-long)
Copied to clipboard
| Challenge: | Existing studies define a sentence as stereotypical and anti-stereotypical, but they lack a fine-grained quantification of stereotypes. |
| Approach: | They quantify stereotypes in language by annotating a dataset to quantify stereotype of sentences. |
| Outcome: | The proposed models validate the findings of the current studies. |
When Debiasing Backfires: Counterintuitive Side Effects of Preprocessing-Based Stereotype Mitigation (2026.findings-acl)
Copied to clipboard
| Challenge: | Preprocessing-based methods for stereotype mitigation are widely used in NLP . preprocessing methods cause unintended shifts in attention flow, authors say . |
| Approach: | They propose to use preprocessing-based methods to reduce stereotypes for targeted groups . they find that stereotyping or counter-stereotyping can increase for other demographics . |
| Outcome: | The proposed methods often induce unintended shifts across demographics, the authors show . they show that such side effects are not accompanied by large changes in attention flow . |
Causal-Debias: Unifying Debiasing in Pretrained Language Models and Fine-tuning via Causal Invariant Learning (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods to remove unwanted stereotypical associations from pretrained language models (PLMs) are often focused on removing unwanted stereotypes from PLMs. |
| Approach: | They propose a framework to remove unwanted stereotypical associations in pretrained language models . they propose bias-relevant factors are causal, while labelrelevant factors causal . |
| Outcome: | The proposed framework reduces stereotypical associations after PLMs are fine-tuned . the proposed framework mitigates bias from a causal invariant perspective . |
StereoMap: Quantifying the Awareness of Human-like Stereotypes in Large Language Models (2023.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have been observed to encode harmful associations present in the training data. |
| Approach: | They propose a framework to map LLMs' perceptions of how demographic groups have been viewed by society using the dimensions of Warmth and Competence. |
| Outcome: | The proposed framework maps LLMs’ perceptions of social groups using the dimensions of Warmth and Competence. |
A Study of Implicit Bias in Pretrained Language Models against People with Disabilities (2022.coling-1)
Copied to clipboard
| Challenge: | Pretrained language models exhibit sociodemographic biases, such as against gender and race, raising concerns of downstream biase in language technologies. |
| Approach: | They propose to use word embedding-based and transformer-based PLMs to test for the presence of biases against people with disabilities (PWDs) |
| Outcome: | The proposed models favor ableist language, despite their sociodemographic biases against race and gender. |
Analyzing Stereotypes in Generative Text Inference Tasks (2021.findings-acl)
Copied to clipboard
| Challenge: | Social psychology studies how social stereotypes are shared as part of cultural knowledge . |
| Approach: | They study how stereotypes manifest when potential targets are situated in neutral contexts . they collect human judgments on the presence of stereotypes in generated inferences based on annotator positionality . |
| Outcome: | The results show that the annotators' positions differ depending on the type of inferences they generate . |