mlsanityData to 7 Oct 2026

AVEC: Bootstrapping Privacy for Local LLMs

arXiv 2509.10561 ↗First posted 10 Sep 2025cs.CR cs.AIcited by 1

On the map: Synthetic data and self-improvement for LLMs → Privacy-preserving tabular data synthesis

Abstract (arXiv)

This position paper presents AVEC (Adaptive Verifiable Edge Control), a framework for bootstrapping privacy for local language models by enforcing privacy at the edge with explicit verifiability for delegated queries. AVEC introduces an adaptive budgeting algorithm that allocates per-query differential privacy parameters based on sensitivity, local confidence, and historical usage, and uses verifiable transformation with on-device integrity checks. We formalize guarantees using R'enyi differential privacy with odometer-based accounting, and establish utility ceilings, delegation-leakage bounds, and impossibility results for deterministic gating and hash-only certification. Our evaluation is simulation-based by design to study mechanism behavior and accounting; we do not claim deployment readiness or task-level utility with live LLMs. The contribution is a conceptual architecture and theoretical foundation that chart a pathway for empirical follow-up on privately bootstrapping local LLMs.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Privacy-preserving tabular data synthesis, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 13 of its 24 references that are in the map. First posted on arXiv 10 Sep 2025.

Builds on

13 of 24 references.

References · 13

  1. 2017Attention Is All You Needcited by 59,932
  2. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  3. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,592
  4. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  5. 2016Deep Learning with Differential Privacycited by 2,682
  6. 2023A Survey on Large Language Model based Autonomous Agentscited by 2,203
  7. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,081
  8. 2019R'enyi Differential Privacy of the Sampled Gaussian Mechanismcited by 199
  9. 2023LLM in a flash: Efficient Large Language Model Inference with Limited Memorycited by 170
  10. 2024On-Device Language Models: A Comprehensive Reviewcited by 101
  11. 2016Privacy Odometers and Filters: Pay-as-you-Go Compositioncited by 32
  12. 2025A Survey of Zero-Knowledge Proof Based Verifiable Machine Learningcited by 21
1 more
  1. 2024SCALM: Towards Semantic Caching for Automated Chat Services with Large Language Modelscited by 16

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Sensitive-Topic Leakage Through LLM Routing Metadata: Measurement and Mitigationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Privacy-preserved LLM Cascade via CoT-enhanced Policy Learning
  2. 2026Securing LLMs in the Wild: Privacy and Security Challenges at the Edge
  3. 2025AlignDP: Hybrid Differential Privacy with Rarity-Aware Protection for LLMs
  4. 2025Privacy in Action: Towards Realistic Privacy Mitigation and Evaluation for LLM-Powered Agents
  5. 2026LLM-Redactor: An Empirical Evaluation of Eight Techniques for Privacy-Preserving LLM Requests
  6. 2026P2Skill: Privacy Preserving Skill Distillation for Cloud-Local LLM Inference Systems
  7. 2026DPrivBench: Benchmarking LLMs' Reasoning for Differential Privacy
  8. 2026OCELOT: Inference-Leakage Budgets for Privacy-Preserving LLM Agents