mlsanityData to 7 Oct 2026

Private Anytime Selective-Risk Certification for Federated Retrieval-Augmented Generation: Guarantees and Empirical Limits

arXiv 2608.07913 ↗First posted 8 Aug 2026cs.CR cs.AIcited by 0

On the map: Retrieval-augmented generation → Privacy and security of RAG

Abstract (arXiv)

Selective-risk certificates promise that accepted outputs meet a declared error target. We develop Fed-SRC, a score-agnostic certificate for federated, differentially private, adaptively monitored retrieval-augmented generation. Clients release only Gaussian-perturbed score and loss histograms. Record-indexed and noise-variance-indexed martingales jointly bound target-risk contrast and accepted mass over all registered thresholds and rounds, permitting predictable recruitment, dropout, threshold selection, and optional stopping. A range-one total-variation term transfers the calibration mixture to a declared deployment mixture. The contribution is this private, federated, anytime combination, rather than the contrast statistic or acceptance floor individually. Empirically, no simultaneous-bound violation occurs in any evaluated cell, privacy level, or policy. Operational power depends on the score and population: the primary target r*=0.10 never certifies, and on RAGTruth the secondary target r*=0.20 never certifies either, whereas on HaluEval question answering it certifies in all 200 non-private trials, with held-out risk below the target. Naively privatized non-private certificates violate their bounds in 146 to 198 of 200 trials. As an exploratory comparison, we also evaluate a private betting-capital heuristic for which we do not establish e-process validity. This heuristic stops certifying at epsilon <= 4, where Fed-SRC still certifies. Certification nevertheless consumes roughly 30 times more stream events than unique calibration items.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Privacy and security of RAG, one of 15 directions in Retrieval-augmented generation. Builds on 19 of its 38 references that are in the map. First posted on arXiv 8 Aug 2026.

Builds on

19 of 38 references.

References · 19

  1. 2023FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generationcited by 1,259
  2. 2016Concentrated Differential Privacy: Simplifications, Extensions, and Lower Boundscited by 339
  3. 2021Distribution-Free, Risk-Controlling Prediction Setscited by 278
  4. 2023RAGTruth: A Hallucination Corpus for Developing Trustworthy Retrieval-Augmented Language Modelscited by 272
  5. 2021Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Controlcited by 256
  6. 2018Time-uniform, nonparametric, nonasymptotic confidence sequencescited by 220
  7. 2018Improving the Gaussian Mechanism for Differential Privacy: Analytical Calibration and Optimal Denoisingcited by 216
  8. 2022Game-theoretic statistics and safe anytime-valid inferencecited by 145
  9. 2024ConU: Conformal Uncertainty in Large Language Models with Correctness Coverage Guaranteescited by 62
  10. 2024Mitigating LLM Hallucinations via Conformal Abstentioncited by 56
  11. 2023Federated Conformal Predictors for Distributed Uncertainty Quantificationcited by 46
  12. 2023Conformal Prediction for Federated Uncertainty Quantification Under Label Shiftcited by 29
7 more
  1. 2024C-RAG: Certified Generation Risks for Retrieval-Augmented Language Modelscited by 27
  2. 2022Fully Adaptive Composition in Differential Privacycited by 22
  3. 2022Nonparametric extensions of randomized response for private confidence setscited by 6
  4. 2025Selective Risk Certification for LLM Outputs via Information-Lift Statistics: PAC-Bayes, Robustness, and Skeleton Designcited by 5
  5. 2025Anytime-Valid Answer Sufficiency Certificates for LLM Generation via Sequential Information Liftcited by 3
  6. 2026Optimal Rates for Differentially Private Hypothesis Testing with E-valuescited by 1
  7. 2026Certified Adaptive Refresh: Anytime-Valid Monitoring for Federated Conformal RAGcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Privacy-Preserving Federated Embedding Learning for Localized Retrieval-Augmented Generation
  2. 2026Byzantine-Robust Federated RAG via Aligned Calibration and Fixed-Membership Conformal Prediction
  3. 2024FRAG: Toward Federated Vector Database Management for Collaborative and Secure Retrieval-Augmented Generation
  4. 2025Federated Retrieval-Augmented Generation: A Systematic Mapping Study
  5. 2024C-FedRAG: A Confidential Federated Retrieval-Augmented Generation System
  6. 2026Security and Privacy in Retrieval-Augmented Generation: Architectures, Threats, Defenses, and Future Directions for Building Trustworthy Systems
  7. 2026Quantization Enables Private Dense Retrieval against Malicious Service Providers
  8. 2024Privacy-Preserving Retrieval-Augmented Generation with Differential Privacy