Interpretability Analysis of Domain Adapted Dense Retrievers

2025-01-24

AI Summary

This work investigates how unsupervised domain adaptation (UDA) alters the internal attribution mechanisms of dense retrievers in cross-domain zero-shot retrieval—specifically in finance and biomedicine. Addressing the gap in existing interpretability analyses, which neglect joint query–document modeling, we introduce Integrated Gradients to dense retrieval for the first time, proposing a novel dual-granularity attribution baseline that operates at both instance-level and ranking-level. Experiments demonstrate that UDA significantly enhances model attention to domain-critical terms (e.g., “hedge”, “corona”), validating the method’s efficacy in deconstructing black-box retrieval decisions. Our work establishes the first systematic attribution analysis framework and empirical benchmark for cross-domain interpretability in dense retrieval.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Image and Video Retrieval
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

Dense retrievers have demonstrated significant potential for neural information retrieval; however, they exhibit a lack of robustness to domain shifts, thereby limiting their efficacy in zero-shot settings across diverse domains. Previous research has investigated unsupervised domain adaptation techniques to adapt dense retrievers to target domains. However, these studies have not focused on explainability analysis to understand how such adaptations alter the model's behavior. In this paper, we propose utilizing the integrated gradients framework to develop an interpretability method that provides both instance-based and ranking-based explanations for dense retrievers. To generate these explanations, we introduce a novel baseline that reveals both query and document attributions. This method is used to analyze the effects of domain adaptation on input attributions for query and document tokens across two datasets: the financial question answering dataset (FIQA) and the biomedical information retrieval dataset (TREC-COVID). Our visualizations reveal that domain-adapted models focus more on in-domain terminology compared to non-adapted models, exemplified by terms such as"hedge,""gold,""corona,"and"disease."This research addresses how unsupervised domain adaptation techniques influence the behavior of dense retrievers when adapted to new domains. Additionally, we demonstrate that integrated gradients are a viable choice for explaining and analyzing the internal mechanisms of these opaque neural models.

Problem

• Information Retrieval
• Domain Adaptation
• Search Efficiency Analysis

Innovation

• Integrated Gradients
• Unsupervised Domain Adaptation
• Keyword Analysis