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

arXiv: https://arxiv.org/abs/2607.28829 · HTML: https://mlsanity.com/paper/2607.28829/

## 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.

## Placed in

- LLM safety and security / [Federated unlearning and poisoning](https://mlsanity.com/safety/federated-unlearning-and-poisoning/)

## Builds on (most cited first, 26 of 26)

- [OpenVLA: An Open-Source Vision-Language-Action Model](https://mlsanity.com/paper/2406.09246/) (2024)
- [π0: A Vision-Language-Action Flow Model for General Robot Control](https://mlsanity.com/paper/2410.24164/) (2024)
- [Understanding Black-box Predictions via Influence Functions](https://mlsanity.com/paper/1703.04730/) (2017)
- [LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning](https://mlsanity.com/paper/2306.03310/) (2023)
- [Membership Inference Attacks From First Principles](https://mlsanity.com/paper/2112.03570/) (2021)
- [How To Backdoor Federated Learning](https://mlsanity.com/paper/1807.00459/) (2018)
- [Estimating Training Data Influence by Tracing Gradient Descent](https://mlsanity.com/paper/2002.08484/) (2020)
- [Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning](https://mlsanity.com/paper/2404.05868/) (2024)
- [Gradient Projection Memory for Continual Learning](https://mlsanity.com/paper/2103.09762/) (2021)
- [Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearning](https://mlsanity.com/paper/2410.07163/) (2024)
- [Catastrophic Failure of LLM Unlearning via Quantization](https://mlsanity.com/paper/2410.16454/) (2024)
- [Data Feedback Loops: Model-driven Amplification of Dataset Biases](https://mlsanity.com/paper/2209.03942/) (2022)
- [On Large Language Model Continual Unlearning](https://mlsanity.com/paper/2407.10223/) (2024)
- [VeriFi: Towards Verifiable Federated Unlearning](https://mlsanity.com/paper/2205.12709/) (2022)
- [Federated Unlearning: A Survey on Methods, Design Guidelines, and Evaluation Metrics](https://mlsanity.com/paper/2401.05146/) (2024)
- [SIFU: Sequential Informed Federated Unlearning for Efficient and Provable Client Unlearning in Federated Optimization](https://mlsanity.com/paper/2211.11656/) (2022)
- [Source Inference Attacks: Beyond Membership Inference Attacks in Federated Learning](https://mlsanity.com/paper/2310.00222/) (2023)
- [Edge Unlearning is Not "on Edge"! An Adaptive Exact Unlearning System on Resource-Constrained Devices](https://mlsanity.com/paper/2410.10128/) (2024)
- [Enhancing Vehicular Platooning with Wireless Federated Learning: A Resource-Aware Control Framework](https://mlsanity.com/paper/2507.00856/) (2025)
- [Forget to Improve: On-Device LLM-Agent Continual Learning via Budget-Curated Memory](https://mlsanity.com/paper/2606.25115/) (2026)
- ["X of Information'' Continuum: A Survey on AI-Driven Multi-dimensional Metrics for Next-Generation Networked Systems](https://mlsanity.com/paper/2507.19657/) (2025)
- [RELIEF: Turning Missing Modalities into Training Acceleration for Federated Learning on Heterogeneous IoT Edge](https://mlsanity.com/paper/2604.04243/) (2026)
- [Lifecycle-Aware Federated Continual Learning in Mobile Autonomous Systems](https://mlsanity.com/paper/2604.20745/) (2026)
- [Reinforcement Learning-Based Energy-Aware Coverage Path Planning for Precision Agriculture](https://mlsanity.com/paper/2601.16405/) (2026)
- [EASE: Federated Multimodal Unlearning via Entanglement-Aware Anchor Closure](https://mlsanity.com/paper/2605.00733/) (2026)
- [SCALE: Sensitivity-Aware Federated Unlearning with Information Freshness Optimization for Mobile Edge Computing](https://mlsanity.com/paper/2605.22589/) (2026)

## 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.