# Interpreting Bias in Large Language Models: A Feature-Based Approach

arXiv: https://arxiv.org/abs/2406.12347 · HTML: https://mlsanity.com/paper/2406.12347/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 29 references that are in the map.
- First posted on arXiv 18 Jun 2024.

## Builds on (most cited first, 19 of 19)

- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings](https://mlsanity.com/paper/1607.06520/) (2016)
- [Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small](https://mlsanity.com/paper/2211.00593/) (2022)
- [StereoSet: Measuring stereotypical bias in pretrained language models](https://mlsanity.com/paper/2004.09456/) (2020)
- [Towards Automated Circuit Discovery for Mechanistic Interpretability](https://mlsanity.com/paper/2304.14997/) (2023)
- [CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Models](https://mlsanity.com/paper/2010.00133/) (2020)
- [The Woman Worked as a Babysitter: On Biases in Language Generation](https://mlsanity.com/paper/1909.01326/) (2019)
- [Dissecting Recall of Factual Associations in Auto-Regressive Language Models](https://mlsanity.com/paper/2304.14767/) (2023)
- [BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generation](https://mlsanity.com/paper/2101.11718/) (2021)
- [On Measuring Social Biases in Sentence Encoders](https://mlsanity.com/paper/1903.10561/) (2019)
- [Does Circuit Analysis Interpretability Scale? Evidence from Multiple Choice Capabilities in Chinchilla](https://mlsanity.com/paper/2307.09458/) (2023)
- [Bias Out-of-the-Box: An Empirical Analysis of Intersectional Occupational Biases in Popular Generative Language Models](https://mlsanity.com/paper/2102.04130/) (2021)
- [UnQovering Stereotyping Biases via Underspecified Questions](https://mlsanity.com/paper/2010.02428/) (2020)
- [Identifying and Controlling Important Neurons in Neural Machine Translation](https://mlsanity.com/paper/1811.01157/) (2018)
- [AtP*: An efficient and scalable method for localizing LLM behaviour to components](https://mlsanity.com/paper/2403.00745/) (2024)
- [Sparse Interventions in Language Models with Differentiable Masking](https://mlsanity.com/paper/2112.06837/) (2021)
- [A Trip Towards Fairness: Bias and De-Biasing in Large Language Models](https://mlsanity.com/paper/2305.13862/) (2023)
- [Causally Testing Gender Bias in LLMs: A Case Study on Occupational Bias](https://mlsanity.com/paper/2212.10678/) (2022)
- [Identifying and Adapting Transformer-Components Responsible for Gender Bias in an English Language Model](https://mlsanity.com/paper/2310.12611/) (2023)

## Built on by (most cited first, 4 of 4)

- [Analysing Moral Bias in Finetuned LLMs through Mechanistic Interpretability](https://mlsanity.com/paper/2510.12229/) (2025)
- [Large Language Models Develop Novel Social Biases Through Adaptive Exploration](https://mlsanity.com/paper/2511.06148/) (2025)
- [UGID: Unified Graph Isomorphism for Debiasing Large Language Models](https://mlsanity.com/paper/2603.19144/) (2026)
- [Thin Evidence, Thick Priors: How Language Models Substitute Identity for Missing Financial Facts](https://mlsanity.com/paper/2610.07798/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) such as Mistral and LLaMA have showcased remarkable performance across various natural language processing (NLP) tasks. Despite their success, these models inherit social biases from the diverse datasets on which they are trained. This paper investigates the propagation of biases within LLMs through a novel feature-based analytical approach. Drawing inspiration from causal mediation analysis, we hypothesize the evolution of bias-related features and validate them using interpretability techniques like activation and attribution patching. Our contributions are threefold: (1) We introduce and empirically validate a feature-based method for bias analysis in LLMs, applied to LLaMA-2-7B, LLaMA-3-8B, and Mistral-7B-v0.3 with templates from a professions dataset. (2) We extend our method to another form of gender bias, demonstrating its generalizability. (3) We differentiate the roles of MLPs and attention heads in bias propagation and implement targeted debiasing using a counterfactual dataset. Our findings reveal the complex nature of bias in LLMs and emphasize the necessity for tailored debiasing strategies, offering a deeper understanding of bias mechanisms and pathways for effective mitigation.