# Locating Demographic Bias at the Attention-Head Level in CLIP's Vision Encoder

arXiv: https://arxiv.org/abs/2603.11793 · HTML: https://mlsanity.com/paper/2603.11793/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 24 references that are in the map.
- First posted on arXiv 12 Mar 2026.

## Builds on (most cited first, 17 of 17)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)](https://mlsanity.com/paper/1711.11279/) (2017)
- [Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings](https://mlsanity.com/paper/1607.06520/) (2016)
- [Model Cards for Model Reporting](https://mlsanity.com/paper/1810.03993/) (2018)
- [Towards Automated Circuit Discovery for Mechanistic Interpretability](https://mlsanity.com/paper/2304.14997/) (2023)
- [Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints](https://mlsanity.com/paper/1707.09457/) (2017)
- [Diffusion Models already have a Semantic Latent Space](https://mlsanity.com/paper/2210.10960/) (2022)
- [Multimodal datasets: misogyny, pornography, and malignant stereotypes](https://mlsanity.com/paper/2110.01963/) (2021)
- [Interpreting CLIP's Image Representation via Text-Based Decomposition](https://mlsanity.com/paper/2310.05916/) (2023)
- [Evaluating CLIP: Towards Characterization of Broader Capabilities and Downstream Implications](https://mlsanity.com/paper/2108.02818/) (2021)
- [Finetuning Text-to-Image Diffusion Models for Fairness](https://mlsanity.com/paper/2311.07604/) (2023)
- [Self-Discovering Interpretable Diffusion Latent Directions for Responsible Text-to-Image Generation](https://mlsanity.com/paper/2311.17216/) (2023)
- [FACET: Fairness in Computer Vision Evaluation Benchmark](https://mlsanity.com/paper/2309.00035/) (2023)
- [Balancing Act: Distribution-Guided Debiasing in Diffusion Models](https://mlsanity.com/paper/2402.18206/) (2024)
- [Dissecting and Mitigating Diffusion Bias via Mechanistic Interpretability](https://mlsanity.com/paper/2503.20483/) (2025)

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

- [Efficient bias mitigation in T2I diffusion models using Concept Graphs](https://mlsanity.com/paper/2607.03397/) (2026)
- [Mechanism-Level Evaluation for Vision-Language Models: Controlled Activation-Replacement Diagnosis of Gender Bias](https://mlsanity.com/paper/2609.16651/) (2026)

## Abstract (arXiv)

Standard fairness audits of foundation models quantify that a model is biased, but not where inside the network the bias resides. We propose a mechanistic fairness audit that combines projected residual-stream decomposition, zero-shot Concept Activation Vectors, and bias-augmented TextSpan analysis to locate demographic bias at the level of individual attention heads in vision transformers. As a feasibility case study, we apply this pipeline to the CLIP ViT-L-14 encoder on 42 profession classes of the FACET benchmark, auditing both gender and age bias. For gender, the pipeline identifies four terminal-layer heads whose ablation reduces global bias (Cramer's V: 0.381 -> 0.362) while marginally improving accuracy (+0.42%); a layer-matched random control confirms that this effect is specific to the identified heads. A single head in the final layer contributes to the majority of the reduction in the most stereotyped classes, and class-level analysis shows that corrected predictions shift toward the correct occupation. For age, the same pipeline identifies candidate heads, but ablation produces weaker and less consistent effects, suggesting that age bias is encoded more diffusely than gender bias in this model. These results provide preliminary evidence that head-level bias localisation is feasible for discriminative vision encoders and that the degree of localisability may vary across protected attributes. keywords: Bias . CLIP . Mechanistic Interpretability . Vision Transformer . Fairness