A Mechanistic Analysis of Gender Sensitivity in Dense Retrieval Models
Catherine Chen, Maarten de Rijke, Carsten Eickhoff
Abstract
While gender bias in dense retrieval models is well documented, with prior work showing that models often score male-gendered documents higher than female or neutral variants, the internal mechanisms producing these disparities are poorly understood. In this paper, we mechanistically analyze bi-encoder models to localize gender sensitivity, finding that the signal originates in input embeddings and propagates through a small set of late-layer attention heads that carry both gender and term-matching signals. Guided by these findings, we test steering interventions at both identified points and find distinct effects: embedding-level steering non-specifically neutralizes score differences, while attention-level steering produces directional shifts. Our findings provide a mechanistic basis for targeted debiasing and highlight the challenge of disentangling gender from relevance signals in shared model components.
Create a lesson
Related papers
SURF: Subtractive Updates for Recommender Forgetting
Filippo Betello, Antonio Purificato, Nicola Tonellotto et al.
Exploring LLMs and RAG for Plausible and Explainable Material Prediction of Vehicle Components
Frederik Wagner, Annerose Eichel, Sabine Schulte im Walde
One-Step Retrieval Framework for Real-Time Sponsored Search Ads Using Hierarchical Text Representations
Tongtong Liu, Renyu Zhang, Jiayu Ding et al.
Quanta: A Self-Contained Python Library for Hybrid Retrieval over Quantised Embeddings, Lexical Indexes, and Knowledge Graphs
Ioannis E. Livieris
Single-Token Expected-Value Scoring for Cold-Start Candidate Ranking
Qihang Wang, Jinwei Tan, Mengyuan Shi et al.
PageRecall: Measuring Page Selection in Literature-Grounded Question Answering
Aaditya Chauhan