mlsanityData to 14 Oct 2026

Exploratory As-Analyzed No-Detection of Culturally-Marked Predicate-Triggered PII Amplification in a Synthetic-English RAG Probe: A Predicate-Resource-Confounded Audit

arXiv 2608.20351 ↗First posted 16 Jun 2026cs.CL cs.CY cs.LGcited by 0

On the map: Retrieval-augmented generation → RAG robustness and adaptive retrieval

Abstract (arXiv)

We ask whether stereotype-loaded queries about culturally marked people leak more personal information from a retrieval-augmented generation (RAG) system than otherwise-equivalent neutral queries. We pre-register a four-culture audit (en-Anglo, es-LATAM, Arabic, Hindi) on a synthetic English PII corpus, comparing five query arms we call the Stereotype-Trigger Leakage Delta (STLD). Two caveats up front. Our locked confirmatory estimator was never run, so every test in the paper is exploratory or sensitivity, with all plan deviations listed in the appendix. And the name-leakage metric is contaminated by a prompt-echo artifact: the model often just re-emits the name we asked about, which inflates apparent leakage without any retrieval at all. On the cleaner channels (email, phone, ssn-like, address), we find no stereotype-driven amplification on any of the four cultures after multiple-comparison correction. Because our sample is only powered for mid-sized effects, and because the culturally marked probes mix stereotype content with cultural markers and heritage practices, we present this as no detection, not evidence of no effect, of culturally marked predicate leakage that is confounded with the underlying resource.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in RAG robustness and adaptive retrieval, one of 15 directions in Retrieval-augmented generation. Builds on 28 of its 47 references that are in the map. First posted on arXiv 16 Jun 2026.

Builds on

28 of 47 references.

Closest ancestors · 2

Same direction, newest first

  1. 2026Does RAG Know When Retrieval Is Wrong? Diagnosing Context Compliance under Knowledge Conflict14 May
  2. 2025Evaluating the Effect of Retrieval Augmentation on Social Biases24 Feb

Other references · 26

  1. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  2. 2025Qwen2.5-VL Technical Reportcited by 5,587
  3. 2024Qwen2.5 Technical Reportcited by 4,942
  4. 2024M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillationcited by 1,073
  5. 2020StereoSet: Measuring stereotypical bias in pretrained language modelscited by 924
  6. 2020CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Modelscited by 710
  7. 2022Deduplicating Training Data Mitigates Privacy Risks in Language Modelscited by 314
  8. 2024The Good and The Bad: Exploring Privacy Issues in Retrieval-Augmented Generation (RAG)cited by 159
  9. 2020Intrinsic Bias Metrics Do Not Correlate with Application Biascited by 148
  10. 2023KoBBQ: Korean Bias Benchmark for Question Answeringcited by 71
  11. 2023SeeGULL: A Stereotype Benchmark with Broad Geo-Cultural Coverage Leveraging Generative Modelscited by 66
  12. 2025The State of Multilingual LLM Safety Research: From Measuring the Language Gap to Mitigating Itcited by 45
14 more
  1. 2024MBBQ: A Dataset for Cross-Lingual Comparison of Stereotypes in Generative LLMscited by 35
  2. 2023Quantifying Association Capabilities of Large Language Models and Its Implications on Privacy Leakagecited by 28
  3. 2024SeeGULL Multilingual: a Dataset of Geo-Culturally Situated Stereotypescited by 24
  4. 2026SafetyRepro: Configuration-Conditional Rank Instability on Alignment Benchmarkscited by 15
  5. 2026Relevant Is Not Warranted: Evidence-Force Calibration for Cited RAGcited by 14
  6. 2025Understanding and Mitigating Cross-lingual Privacy Leakage via Language-specific and Universal Privacy Neuronscited by 13
  7. 2025DAIQ: Auditing Demographic Attribute Inference from Question in LLMscited by 10
  8. 2024Underestimated Privacy Risks for Minority Populations in Large Language Model Unlearningcited by 9
  9. 2025Your RAG is Unfair: Exposing Fairness Vulnerabilities in Retrieval-Augmented Generation via Backdoor Attackscited by 8
  10. 2026Retrieval--Reasoning Processes for Multi-hop Question Answering: A Four-Axis Design Framework and Empirical Trendscited by 8
  11. 2026Pre-Registering the Detectable Effect: A Paired-MDE Budget for 4-bit Quantization Benchmarks, with a Pilot Auditcited by 8
  12. 2025A Comprehensive Framework to Operationalize Social Stereotypes for Responsible AI Evaluationscited by 7
  13. 2026Auditing Reasoning-Trace Memorization Claims after Unlearning with Head-Conditioned Canariescited by 7
  14. 2026Do LLMs Really Memorize Personally Identifiable Information? Revisiting PII Leakage with a Cue-Controlled Memorization Frameworkcited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Where Privacy Risk Lives in English-Source Multilingual RAG: A Stage-Decomposed Audit Across Five Query Languages
  2. 2026Retrieval Sensitivity to Identity Signals in Queries
  3. 2026QuantiBias: Benchmarking Quantization-Induced Bias in LLMs
  4. 2025Reading Between the Prompts: How Stereotypes Shape LLM's Implicit Personalization
  5. 2026Dialect vs Demographics: Quantifying LLM Bias from Implicit Linguistic Signals vs. Explicit User Profiles
  6. 2026ImplicitBBQ: Benchmarking Implicit Bias in Large Language Models through Characteristic Based Cues
  7. 2026Evaluating Social Bias in RAG Systems: When External Context Helps and Reasoning Hurts
  8. 2025Mitigating Social Bias in English and Urdu Language Models Using PRM-Guided Candidate Selection and Sequential Refinement