mlsanityData to 14 Oct 2026

When Unlearning Fails: Reliable Data Deletion under Post-Training in Agent Networks

arXiv 2607.28829 ↗First posted 30 Jul 2026cs.NI cs.LGcited by 0

On the map: LLM safety and security → Federated unlearning and poisoning

Abstract (arXiv)

Embedded AI agents on edge devices can learn as a federation and keep improving after deployment. Each agent collects new trajectories with the current policy, and the edge server aggregates them in later rounds. In such self-improving agent networks, deleting data is harder than a one-time model repair. The data to be deleted may have already shaped later retained trajectories, so retraining or model-side unlearning leaves an influence echo that returns as the network keeps learning. We propose Muting Unlearned Trajectories' Echoes (MUTE) for reliable deletion over edge networks. MUTE traces the echo from a lightweight ledger at the edge server, without moving raw trajectories off the devices. It erases the current residue with a forget-retain update, contains high-influence retained trajectories, and audits later behavior to schedule further erasure under the uplink budget and the response deadline. On LIBERO, MUTE keeps the regenerated share of forgotten behavior at or below 0.401 for every request type, against at least 0.466 for five recent methods.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Federated unlearning and poisoning, one of 12 directions in LLM safety and security. Builds on 26 of its 47 references that are in the map. First posted on arXiv 30 Jul 2026.

Builds on

26 of 47 references.

Closest ancestors · 4

Same direction, newest first

  1. 2024Federated Unlearning: A Survey on Methods, Design Guidelines, and Evaluation Metrics10 Jan
  2. 2023Source Inference Attacks: Beyond Membership Inference Attacks in Federated Learning30 Sep
  3. 2022SIFU: Sequential Informed Federated Unlearning for Efficient and Provable Client Unlearning in Federated Optimization21 Nov
  4. 2022VeriFi: Towards Verifiable Federated Unlearning25 May

Other references · 22

  1. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,887
  2. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  3. 2017Understanding Black-box Predictions via Influence Functionscited by 1,741
  4. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,410
  5. 2021Membership Inference Attacks From First Principlescited by 878
  6. 2018How To Backdoor Federated Learningcited by 765
  7. 2020Estimating Training Data Influence by Tracing Gradient Descentcited by 580
  8. 2024Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearningcited by 504
  9. 2021Gradient Projection Memory for Continual Learningcited by 348
  10. 2024Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearningcited by 172
  11. 2024Catastrophic Failure of LLM Unlearning via Quantizationcited by 81
  12. 2022Data Feedback Loops: Model-driven Amplification of Dataset Biasescited by 60
10 more
  1. 2024On Large Language Model Continual Unlearningcited by 52
  2. 2024Edge Unlearning is Not "on Edge"! An Adaptive Exact Unlearning System on Resource-Constrained Devicescited by 9
  3. 2025Enhancing Vehicular Platooning with Wireless Federated Learning: A Resource-Aware Control Frameworkcited by 9
  4. 2026Forget to Improve: On-Device LLM-Agent Continual Learning via Budget-Curated Memorycited by 9
  5. 2025"X of Information'' Continuum: A Survey on AI-Driven Multi-dimensional Metrics for Next-Generation Networked Systemscited by 7
  6. 2026RELIEF: Turning Missing Modalities into Training Acceleration for Federated Learning on Heterogeneous IoT Edgecited by 7
  7. 2026Lifecycle-Aware Federated Continual Learning in Mobile Autonomous Systemscited by 7
  8. 2026Reinforcement Learning-Based Energy-Aware Coverage Path Planning for Precision Agriculturecited by 6
  9. 2026EASE: Federated Multimodal Unlearning via Entanglement-Aware Anchor Closurecited by 5
  10. 2026SCALE: Sensitivity-Aware Federated Unlearning with Information Freshness Optimization for Mobile Edge Computingcited by 5

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Lethe: Principled Dual-Stream Update for Persistent Knowledge Erasure in Federated Unlearning
  2. 2026FUPareto: Bridging the Forgetting-Utility Gap in Federated Unlearning via Pareto Augmented Optimization
  3. 2026Forgetting Without Restarting: Execution-State Unlearning for Stateful LLM Agents
  4. 2026K-Bench: A Benchmark for LLM Unlearning in Agentic Deployments
  5. 2026Agentic Unlearning: When LLM Agent Meets Machine Unlearning
  6. 2026TOUR: A Trajectory-Level Unlearning Benchmark for Offline Reinforcement Learning
  7. 2026FedCARE: Federated Unlearning with Conflict-Aware Projection and Relearning-Resistant Recovery
  8. 2025NoT: Federated Unlearning via Weight Negation