cs.IRMar 25, 2026

Who Benefits from RAG? The Role of Exposure, Utility and Attribution Bias

Authors: Mahdi Dehghan, Graham McDonald

Organizations: University of Glasgow, Glasgow, UK

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Large Language Models (LLMs) enhanced with Retrieval-Augmented Generation (RAG) have achieved substantial improvements in accuracy by grounding their responses in external documents that are relevant to the user's query. However, relatively little work has investigated the impact of RAG in terms of fairness. Particularly, it is not yet known if queries that are associated with certain groups within a fairness category systematically receive higher accuracy, or accuracy improvements in RAG systems compared to LLM-only, a phenomenon we refer to as query group fairness. In this work, we conduct extensive experiments to investigate the impact of three key factors on query group fairness in RAG, namely: Group exposure, i.e., the proportion of documents from each group appearing in the retrieved set, determined by the retriever; Group utility, i.e., the degree to which documents from each group contribute to improving answer accuracy, capturing retriever-generator interactions; and Group attribution, i.e., the extent to which the generator relies on documents from each group when producing responses. We examine group-level average accuracy and accuracy improvements disparities across four fairness categories using three datasets derived from the TREC 2022 Fair Ranking Track for two tasks: article generation and title generation. Our findings show that RAG systems suffer from the query group fairness problem and amplify disparities in terms of average accuracy across queries from different groups, compared to an LLM-only setting. Moreover, group utility, exposure, and attribution can exhibit strong positive or negative correlations with average accuracy or accuracy improvements of queries from that group, highlighting their important role in fair RAG. Our data and code are publicly available from Github.

Figures & tables

Fairness CategoryGroups
Age of the Topic (AoT)Unk, Pre-1900s, 20th century, 21st century
Popularity (Pop)Low, Medium-Low, Medium-High, High
Age of Article (AoA)2001–2006, 2007–2011, 2012–2016, 2017–2022
Alphabetical (Alp)a–d, e–k, l–r, s–z
Table 1: Fairness categories and their groups.
Figure 1: Prompt templates used in RAG systems.
Topic#Q#DocsAvg. Body Len.Avg. Title Len.
Cities159117,36295.42.3
Geography14480,27365.32.2
Military History245109,169251.83.2
Table 2: Dataset statistics.
Table 3: Average ROUGE-L scores in the RAG and LLM-only (N) settings. The best LLM-retriever combination is highlighted bold for each task and dataset.
Figure 2: Range of query group accuracy gains, i.e., RΔR^{\Delta} , for all fairness categories across all topics, tasks, retrievers and LLMs.
Figure 3: Range of query group accuracy in a RAG setting, i.e., RragR^{rag} , for all fairness categories across all topics, tasks, retrievers and LLMs.
Table 4: Comparison of the range of query group accuracy in the RAG and LLM-only (indicated as N) settings, i.e., RragR^{rag} vs RllmR^{llm} .
Figure 4: Spearman correlation between query group accuracy in RAG setting (i.e., ACrag(gi)AC^{rag}(g_{i}) ) and corresponding group utility, exposure and attribution for all fairness categories across topics, task, and LLMs.
Figure 5: Spearman correlation between query group accuracy gain (i.e., Δ(AC(gi))\Delta(AC(g_{i})) ) and corresponding group utility, exposure and attribution for all fairness categories across topics, tasks, and LLMs.
More in this topic

Group Fairness

Group fairness measures and reduces disparities in how algorithms predict, rank, allocate resources, and make decisions for different groups.

See all 140 papers

Explore similar work

CardsList
Oct 26, 2026

Topic-Level Fairness in Retrieval-Augmented Generation

Group fairness in retrieval-augmented generation (RAG) systems is typically evaluated using metrics aggregated across queries. However, such metrics can conceal topic-level disparities: a system may favor women on questions about human rights and men on questions about actors and filmmakers, appearing nearly neutral overall even though neither topic is fair. We therefore study topic-level fairness, which requires group fairness to hold within each topic rather than only in aggregate. To reduce the topic-level disparity, we adapt the reranker separately for each topic with a Bradley-Terry objective under direct supervision from the generated output: preference pairs built from the generator's own output carry the direction that topic requires. With preference pairs built from RoMQA alone, the adaptation reduces the mean absolute per-topic disparity by 51% on RoMQA and by 40% on TREC Fair Ranking while retaining accuracy. No baseline reaches the reduction, even when modified to treat each topic separately rather than apply one correction to every topic. Neither does our method with one reranker for all topics, which follows the direction that most topics require.

Jaeyoung Park, Hantian Zhang, Kexin Rong +1

KAIST · Meta · Georgia Institute of Technology

Group FairnessRetrieval-Augmented GenerationRAG EvaluationLearning to Rank
AskPDF
Mar 23, 2026

CatRAG: Functor-Guided Structural Debiasing with Retrieval Augmentation for Fair LLMs

Large Language Models (LLMs) are deployed in high-stakes settings but can show demographic, gender, and geographic biases that undermine fairness and trust. Prior debiasing methods, including embedding-space projections, prompt-based steering, and causal interventions, often act at a single stage of the pipeline, resulting in incomplete mitigation and brittle utility trade-offs under distribution shifts. We propose CatRAG Debiasing, a dual-pronged framework that integrates functor with Retrieval-Augmented Generation (RAG) guided structural debiasing. The functor component leverages category-theoretic structure to induce a principled, structure-preserving projection that suppresses bias-associated directions in the embedding space while retaining task-relevant semantics. On the Bias Benchmark for Question Answering (BBQ) across three open-source LLMs (Meta Llama-3, OpenAI GPT-OSS, and Google Gemma-3), CatRAG achieves state-of-the-art results, improving accuracy by up to 40% over the corresponding base models and by more than 10% over prior debiasing methods, while reducing bias scores to near zero (from 60% for the base models) across gender, nationality, race, and intersectional subgroups.

Ravi Ranjan, Utkarsh Grover, Mayur Akewar +2

Florida International University Miami, USA · University of South Florida Tampa, USA

Language Model DebiasingSocial Bias in Language ModelsAlgorithmic BiasDemographic Bias in Language ModelsLanguage Model Bias Evaluation
AskPDF
Aug 26, 2026

When RAG Fails to Equalize: Geo-bias in Factual Question Answering over Public Companies

Retrieval-augmented generation (RAG) is widely assumed to mitigate factual errors in large language models (LLMs), but it remains unclear whether retrieval uniformly compensates for missing knowledge. We study this question in a controlled factual QA setting over public companies, constructing a benchmark of approximately 2,000 firms across global equity indices. We evaluate six LLMs on four atomic attributes under four conditions: no-context, perfect context, misleading context, and distraction context. We find strong geographic disparities in no-context accuracy, indicating uneven parametric knowledge. While perfect context improves performance, it does not eliminate these gaps: gains are correlated with baseline accuracy, suggesting retrieval effectiveness is coupled to internal representations. Under misleading context, models frequently copy incorrect information. Larger models improve overall performance but do not remove these structural effects. These results challenge the view of RAG as a universal corrective and highlight the interaction between model knowledge, context quality, and entity representation.

Abhinav Havaldar, Enrico Santus

Bloomberg

Jurisdictional Bias in Language ModelsRAG RobustnessFactual Knowledge in Language ModelsLanguage Model Bias EvaluationKnowledge Conflicts in Language Models
AskPDF