Biases in Edge Language Models: Detection, Analysis, and Mitigation
Not on the map. Why?
Abstract (arXiv)
The integration of large language models (LLMs) on low-power edge devices such as Raspberry Pi, known as edge language models (ELMs), has introduced opportunities for more personalized, secure, and low-latency language intelligence that is accessible to all. However, the resource constraints inherent in edge devices and the lack of robust ethical safeguards in language models raise significant concerns about fairness, accountability, and transparency in model output generation. This paper conducts a comparative analysis of text-based bias across language model deployments on edge, cloud, and desktop environments, aiming to evaluate how deployment settings influence model fairness. Specifically, we examined an optimized Llama-2 model running on a Raspberry Pi 4; GPT 4o-mini, Gemini-1.5-flash, and Grok-beta models running on cloud servers; and Gemma2 and Mistral models running on a MacOS desktop machine. Our results demonstrate that Llama-2 running on Raspberry Pi 4 is 43.23% and 21.89% more prone to showing bias over time compared to models running on the desktop and cloud-based environments. We also propose the implementation of a feedback loop, a mechanism that iteratively adjusts model behavior based on previous outputs, where predefined constraint weights are applied layer-by-layer during inference, allowing the model to correct bias patterns, resulting in 79.28% reduction in model bias.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 17 of its 45 references that are in the map. First posted on arXiv 17 Feb 2025.
Builds on
17 of 45 references.
References · 17
- 2024Gemini 1.5: Unlocking multimodal understanding across millions of tokens of contextcited by 3,224
- 2016Pruning Filters for Efficient ConvNetscited by 1,143
- 2024Justice or Prejudice? Quantifying Biases in LLM-as-a-Judgecited by 368
- 2023Fine-Tuning Language Models with Just Forward Passescited by 345
- 2024Break the Sequential Dependency of LLM Inference Using Lookahead Decodingcited by 298
- 2022Easily Accessible Text-to-Image Generation Amplifies Demographic Stereotypes at Large Scalecited by 262
- 2023"Kelly is a Warm Person, Joseph is a Role Model": Gender Biases in LLM-Generated Reference Letterscited by 199
- 2023LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructionscited by 124
- 2023Is ChatGPT Fair for Recommendation? Evaluating Fairness in Large Language Model Recommendationcited by 114
- 2024A Review on Edge Large Language Models: Design, Execution, and Applicationscited by 100
- 2024Systematic Biases in LLM Simulations of Debatescited by 99
- 2024On Protecting the Data Privacy of Large Language Models (LLMs): A Surveycited by 96
5 more
- 2023Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMscited by 50
- 2024CFaiRLLM: Consumer Fairness Evaluation in Large-Language Model Recommender Systemcited by 18
- 2024Empirical Guidelines for Deploying LLMs onto Resource-constrained Edge Devicescited by 17
- 2024Designing Efficient LLM Accelerators for Edge Devicescited by 13
- 2020Addressing the Memory Bottleneck in AI Model Trainingcited by 4
Built on by
Citing papers per year: 2025: 1.
1 paper.