Lune

EMNLP Findings2025

No Free Lunch: Retrieval-Augmented Generation Undermines Fairness in LLMs, Even for Vigilant Users

Mengxuan Hu, Hongyi Wu, Ronghang Zhu, Zihan Guan, Dongliang Guo, Daiqing Qi, Sheng Li

Abstract

Retrieval-Augmented Generation (RAG) is widely adopted for its effectiveness and costefficiency in mitigating hallucinations and enhancing the domain-specific generation capabilities of large language models (LLMs). However, is this effectiveness and cost-efficiency truly a free lunch? In this study, we comprehensively investigate the fairness costs associated with RAG by proposing a practical threelevel threat model from the perspective of user awareness of fairness. Specifically, varying levels of user fairness awareness result in different degrees of fairness censorship on the external dataset. We examine the fairness implications of RAG using uncensored, partially censored, and fully censored datasets. Our experiments demonstrate that fairness alignment can be easily undermined through RAG without the need for fine-tuning or retraining. Even with fully censored and supposedly unbiased external datasets, RAG can lead to biased outputs. Our findings underscore the limitations of current alignment methods in the context of RAG-based LLMs and highlight the urgent need for new strategies to ensure fairness. We propose potential mitigations and call for further research to develop robust fairness safeguards in RAG-based LLMs.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

5 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • The Crowded Embedding Space: A Mean-Field Mechanism for Emergent Marginalization in Retrieval-Augmented Agents cites Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
  • The Crowded Embedding Space: A Mean-Field Mechanism for Emergent Marginalization in Retrieval-Augmented Agents cites Retrieval Augmented Language Model Pre-Training
  • Alignment-Weighted DPO: A principled reasoning approach to improve safety alignment cites Finetuned Language Models are Zero-Shot Learners
  • Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection cites Finetuned Language Models are Zero-Shot Learners
  • Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection cites Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
  • Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection cites Retrieval Augmented Language Model Pre-Training
  • Your RAG is Unfair: Exposing Fairness Vulnerabilities in Retrieval-Augmented Generation via Backdoor Attacks cites Language Models are Few-Shot Learners
  • Language Models are Few-Shot Learners cites Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
  • Finetuned Language Models are Zero-Shot Learners cites Language Models are Few-Shot Learners
  • Information Retrieval Induced Safety Degradation in AI Agents cites Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers5

  1. Information Retrieval Induced Safety Degradation in AI AgentsCheng Yu, Benedikt Stroebl, Diyi Yang, Orestis PapakyriakopoulosNeurIPS 2025 · 5 citations
  2. Alignment-Weighted DPO: A principled reasoning approach to improve safety alignmentMengxuan Hu, Vivek V. Datla, Anoop Kumar, Zihan Guan et al.ICLR 2026 · 3 citations
  3. The Crowded Embedding Space: A Mean-Field Mechanism for Emergent Marginalization in Retrieval-Augmented AgentsShwan Ashrafi, Dan RothICML 2026
  4. BalancEdit: Dynamically Balancing the Generality-Locality Trade-off in Multi-modal Model EditingDongliang Guo, Mengxuan Hu, Zihan Guan, Thomas Hartvigsen et al.ICML 2025
  5. Your RAG is Unfair: Exposing Fairness Vulnerabilities in Retrieval-Augmented Generation via Backdoor AttacksGaurav Bagwe, Saket S. Chaturvedi, Xiaolong Ma, Xiaoyong Yuan et al.EMNLP 2025

Builds on26

  1. Language Models are Few-Shot LearnersTom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah et al.NeurIPS 2020 · 64,255 citations
  2. Retrieval-Augmented Generation for Knowledge-Intensive NLP TasksPatrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni et al.NeurIPS 2020 · 19,162 citations
  3. Finetuned Language Models are Zero-Shot LearnersJason Wei, Maarten Bosma, Vincent Y. Zhao, Kelvin Guu et al.ICLR 2022 · 4,966 citations
  4. Retrieval Augmented Language Model Pre-TrainingKelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat et al.ICML 2020 · 2,937 citations
  5. Self-RAG: Learning to Retrieve, Generate, and Critique through Self-ReflectionAkari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil et al.ICLR 2024 · 1,798 citations

Related papers

  1. LLM-Generated Text May Harm Your Retrieval! A Robust Detection Strategy for Retrieval-Augmented GenerationZhaoheng Huang, Yutao Zhu, Ji-Rong Wen, Zhicheng DouACL 2026
  2. Benchmarking Large Language Models in Retrieval-Augmented GenerationJiawei Chen, Hongyu Lin, Xianpei Han, Le SunAAAI 2024 · 531 citations
  3. C-RAG: Certified Generation Risks for Retrieval-Augmented Language ModelsMintong Kang, Nezihe Merve Gürel, Ning Yu, Dawn Song et al.ICML 2024 · 33 citations
  4. Understand What LLM Needs: Dual Preference Alignment for Retrieval-Augmented GenerationGuanting Dong, Yutao Zhu, Chenghao Zhang, Zechen Wang et al.WWW 2025 · 44 citations
  5. Fine-Tuning or Retrieval? Comparing Knowledge Injection in LLMsOded Ovadia, Menachem Brief, Moshik Mishaeli, Oren ElishaEMNLP 2024 · 89 citations
Dusk over the sea between two cliffs drawn in fine vertical lines