# Weight Pruning Amplifies Bias: A Multi-Method Study of Compressed LLMs for Edge AI

arXiv: https://arxiv.org/abs/2605.08137 · HTML: https://mlsanity.com/paper/2605.08137/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 26 of its 36 references that are in the map.
- First posted on arXiv 2 May 2026.

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

- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding](https://mlsanity.com/paper/1510.00149/) (2015)
- [The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks](https://mlsanity.com/paper/1803.03635/) (2018)
- [Learning both Weights and Connections for Efficient Neural Networks](https://mlsanity.com/paper/1506.02626/) (2015)
- [SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot](https://mlsanity.com/paper/2301.00774/) (2023)
- [A Simple and Effective Pruning Approach for Large Language Models](https://mlsanity.com/paper/2306.11695/) (2023)
- [BBQ: A Hand-Built Bias Benchmark for Question Answering](https://mlsanity.com/paper/2110.08193/) (2021)
- [Bias and Fairness in Large Language Models: A Survey](https://mlsanity.com/paper/2309.00770/) (2023)
- [Large Language Models: A Survey](https://mlsanity.com/paper/2402.06196/) (2024)
- [DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models](https://mlsanity.com/paper/2306.11698/) (2023)
- [A Survey on Model Compression for Large Language Models](https://mlsanity.com/paper/2308.07633/) (2023)
- [Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications](https://mlsanity.com/paper/2402.05162/) (2024)
- [Efficient Large Language Models: A Survey](https://mlsanity.com/paper/2312.03863/) (2023)
- [What Do Compressed Deep Neural Networks Forget?](https://mlsanity.com/paper/1911.05248/) (2019)
- [Characterising Bias in Compressed Models](https://mlsanity.com/paper/2010.03058/) (2020)
- [A Fast Post-Training Pruning Framework for Transformers](https://mlsanity.com/paper/2204.09656/) (2022)
- [Decoding Compressed Trust: Scrutinizing the Trustworthiness of Efficient LLMs Under Compression](https://mlsanity.com/paper/2403.15447/) (2024)
- [Pruning has a disparate impact on model accuracy](https://mlsanity.com/paper/2205.13574/) (2022)
- [Sustainable LLM Inference for Edge AI: Evaluating Quantized LLMs for Energy Efficiency, Output Accuracy, and Inference Latency](https://mlsanity.com/paper/2504.03360/) (2025)
- [Beyond Perplexity: Multi-dimensional Safety Evaluation of LLM Compression](https://mlsanity.com/paper/2407.04965/) (2024)
- [Accuracy is Not All You Need](https://mlsanity.com/paper/2407.09141/) (2024)
- [Wanda++: Pruning Large Language Models via Regional Gradients](https://mlsanity.com/paper/2503.04992/) (2025)
- [Investigating the Impact of Quantization Methods on the Safety and Reliability of Large Language Models](https://mlsanity.com/paper/2502.15799/) (2025)
- [Bias in Pruned Vision Models: In-Depth Analysis and Countermeasures](https://mlsanity.com/paper/2304.12622/) (2023)
- [Investigating Social Bias Changes in Quantized Language Models](https://mlsanity.com/paper/2602.06181/) (2026)
- [Less Is More? Examining Fairness in Pruned Large Language Models for Summarising Opinions](https://mlsanity.com/paper/2508.17610/) (2025)

## Built on by (most cited first, 2 of 2)

- [The Asymmetric Harms of LLM Compression](https://mlsanity.com/paper/2608.19670/) (2026)
- [The Asymmetric Effects of Knowledge Distillation on Bias in Small Language Models](https://mlsanity.com/paper/2607.28639/) (2026)

## Abstract (arXiv)

Weight pruning is widely advocated for deploying Large Language Models on resource-constrained IoT and edge devices, yet its impact on model fairness remains poorly understood. We conduct a controlled empirical study of three instruction-tuned models (Gemma-2-9b-it, Mistral-7B-Instruct-v0.3, Phi-3.5-mini-instruct) across three pruning methods (Random, Magnitude, Wanda) at four sparsity levels (10-70%) on 12,148 BBQ bias benchmark items with 5 random seeds, totaling 2,368,860 inference records. Our results reveal a Smart Pruning Paradox: activation-aware pruning (Wanda) preserves perplexity nearly perfectly (just 3.5% increase at 50% sparsity for Mistral-7B), yet produces the highest bias amplification, with Stereotype Reliance Score increasing 83.7% and 47-59% of previously unbiased items developing new stereotypical behaviors at 70% sparsity. Random pruning destroys language capability entirely (perplexity exceeding 104 and reaching 108) but produces only random-chance bias. We further show that unstructured pruning provides zero storage savings and zero inference latency reduction on real edge hardware, undermining the primary motivation for its use in IoT deployment. Of 180 dense-vs-pruned comparisons, 141 (78.3%) are significant (p < 0.05) with mean |h| = 0.305. Published quantization studies report up to 21% of responses flipping between biased and unbiased states; our pruning results show transition rates nearly three times higher (47-59%), suggesting pruning poses a categorically greater risk to alignment than quantization. These findings demonstrate that perplexity-based evaluation provides false assurance of behavioral equivalence, and that IoT deployment pipelines require bias-aware validation before deploying pruned models at the edge.