# Understanding the Effect of Model Compression on Social Bias in Large Language Models

arXiv: https://arxiv.org/abs/2312.05662 · HTML: https://mlsanity.com/paper/2312.05662/

## Key facts

- Cited by 19 papers in the map (data through 14 Oct 2026).
- Builds on 26 of its 37 references that are in the map.
- First posted on arXiv 9 Dec 2023.

## Builds on (most cited first, 26 of 26)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://mlsanity.com/paper/1910.01108/) (2019)
- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling](https://mlsanity.com/paper/2304.01373/) (2023)
- [StereoSet: Measuring stereotypical bias in pretrained language models](https://mlsanity.com/paper/2004.09456/) (2020)
- [CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Models](https://mlsanity.com/paper/2010.00133/) (2020)
- [Null It Out: Guarding Protected Attributes by Iterative Nullspace Projection](https://mlsanity.com/paper/2004.07667/) (2020)
- [Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP](https://mlsanity.com/paper/2103.00453/) (2021)
- [Measuring Bias in Contextualized Word Representations](https://mlsanity.com/paper/1906.07337/) (2019)
- [The Lottery Ticket Hypothesis for Pre-trained BERT Networks](https://mlsanity.com/paper/2007.12223/) (2020)
- [Depth-Adaptive Transformer](https://mlsanity.com/paper/1910.10073/) (2019)
- [Measuring and Reducing Gendered Correlations in Pre-trained Models](https://mlsanity.com/paper/2010.06032/) (2020)
- [What Do Compressed Deep Neural Networks Forget?](https://mlsanity.com/paper/1911.05248/) (2019)
- [Towards Debiasing Sentence Representations](https://mlsanity.com/paper/2007.08100/) (2020)
- [Characterising Bias in Compressed Models](https://mlsanity.com/paper/2010.03058/) (2020)
- [Debiasing Pre-trained Contextualised Embeddings](https://mlsanity.com/paper/2101.09523/) (2021)
- [Beyond Distillation: Task-level Mixture-of-Experts for Efficient Inference](https://mlsanity.com/paper/2110.03742/) (2021)
- [Lost in Pruning: The Effects of Pruning Neural Networks beyond Test Accuracy](https://mlsanity.com/paper/2103.03014/) (2021)
- [Unmasking the Mask -- Evaluating Social Biases in Masked Language Models](https://mlsanity.com/paper/2104.07496/) (2021)
- [Intriguing Properties of Quantization at Scale](https://mlsanity.com/paper/2305.19268/) (2023)
- [Debiasing isn't enough! -- On the Effectiveness of Debiasing MLMs and their Social Biases in Downstream Tasks](https://mlsanity.com/paper/2210.02938/) (2022)
- [The Low-Resource Double Bind: An Empirical Study of Pruning for Low-Resource Machine Translation](https://mlsanity.com/paper/2110.03036/) (2021)
- [Beyond Preserved Accuracy: Evaluating Loyalty and Robustness of BERT Compression](https://mlsanity.com/paper/2109.03228/) (2021)
- [Intriguing Properties of Compression on Multilingual Models](https://mlsanity.com/paper/2211.02738/) (2022)

## Built on by (most cited first, 19 of 19)

- [Building Guardrails for Large Language Models](https://mlsanity.com/paper/2402.01822/) (2024)
- [On the social bias of speech self-supervised models](https://mlsanity.com/paper/2406.04997/) (2024)
- [Safety Alignment via Constrained Knowledge Unlearning](https://mlsanity.com/paper/2505.18588/) (2025)
- [What's Not Said Still Hurts: A Description-Based Evaluation Framework for Measuring Social Bias in LLMs](https://mlsanity.com/paper/2502.19749/) (2025)
- [FLEX: A Benchmark for Evaluating Robustness of Fairness in Large Language Models](https://mlsanity.com/paper/2503.19540/) (2025)
- [Quantization Undoes Alignment: Bias Emergence in Compressed LLMs Across Models and Precision Levels](https://mlsanity.com/paper/2605.15208/) (2026)
- [Model Compression Techniques in Biometrics Applications: A Survey](https://mlsanity.com/paper/2401.10139/) (2024)
- [Mitigating Biases of Large Language Models in Stance Detection with Counterfactual Augmented Calibration](https://mlsanity.com/paper/2402.14296/) (2024)
- [Investigating Social Bias Changes in Quantized Language Models](https://mlsanity.com/paper/2602.06181/) (2026)
- [How Quantization Shapes Bias in Large Language Models](https://mlsanity.com/paper/2508.18088/) (2025)
- [WINA: Weight Informed Neuron Activation for Accelerating Large Language Model Inference](https://mlsanity.com/paper/2505.19427/) (2025)
- [Surgical Feature-Space Decomposition of LLMs: Why, When and How?](https://mlsanity.com/paper/2405.13039/) (2024)
- [Investigating Hallucination in Conversations for Low Resource Languages](https://mlsanity.com/paper/2507.22720/) (2025)
- [Preserving Fairness and Safety in Quantized LLMs Through Critical Weight Protection](https://mlsanity.com/paper/2601.12033/) (2026)
- [Stress-Testing Efficient Responsible-AI Evaluation: When Compute Savings Change Benchmark Conclusions](https://mlsanity.com/paper/2608.31108/) (2026)
- [Fair-GPTQ: Bias-Aware Quantization for Large Language Models](https://mlsanity.com/paper/2509.15206/) (2025)
- [Debiasing Large Language Models in Thai Political Stance Detection via Counterfactual Calibration](https://mlsanity.com/paper/2509.21946/) (2025)
- [Downsized and Compromised?: Assessing the Faithfulness of Model Compression](https://mlsanity.com/paper/2510.06125/) (2025)
- [Knowing Bias, Doing Better: Mitigating Social Bias in LLMs via Know-Bias Neuron Enhancement](https://mlsanity.com/paper/2601.21864/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) trained with self-supervision on vast corpora of web text fit to the social biases of that text. Without intervention, these social biases persist in the model's predictions in downstream tasks, leading to representational harm. Many strategies have been proposed to mitigate the effects of inappropriate social biases learned during pretraining. Simultaneously, methods for model compression have become increasingly popular to reduce the computational burden of LLMs. Despite the popularity and need for both approaches, little work has been done to explore the interplay between these two. We perform a carefully controlled study of the impact of model compression via quantization and knowledge distillation on measures of social bias in LLMs. Longer pretraining and larger models led to higher social bias, and quantization showed a regularizer effect with its best trade-off around 20% of the original pretraining time.