Skip to content

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

Jul 2026 · arXiv.org · Vol abs/2607.28829 · 0 citations · 47 references
Computer Science

TL;DR

Experiments on LIBERO with two vision-language-action backbones, three deletion granularities, and a physical Jetson-based edge testbed show that MUTE keeps behavioral leakage and influence regeneration low while preserving task utility and using much less communication than full retraining.

Abstract

Self-improving federated agent networks keep training after deployment by collecting new trajectories with the current policy and feeding them back into later rounds. This closed loop makes unlearning harder than a one-time model repair. When a data owner requests deletion, the target data may have already shaped later retained trajectories, so retraining or model-side unlearning can leave an influence echo that returns as the network continues to operate. We show that this echo survives retained-data retraining, grows with the amount of forget-shaped retained data, and can be traced from deployment, collection, and aggregation records. To address this problem, we propose MUTE, a Muting Unlearned Trajectories'Echoes method for reliable deletion in self-improving federated agent networks. MUTE estimates downstream influence from a lightweight server ledger, removes the current residue through a forget-retain update, contains high-influence retained trajectories through quarantine or down-weighting, and audits later behavior to schedule additional erasure under an uplink budget. Experiments on LIBERO with two vision-language-action backbones, three deletion granularities, and a physical Jetson-based edge testbed show that MUTE keeps behavioral leakage and influence regeneration low while preserving task utility and using much less communication than full retraining.

View source

Similar papers

#machine learning Preprint Oct 2026

ASCENT: Online Test-Time Training of Long-Horizon Agents via Self-Distillation of Verified Experience

ASCENT improves task success and interaction efficiency as experience accumulates, outperforms online adaptation methods, and transfers to held-out scenes, showing that an agent can consolidate verified experience into its weights without a separate training phase or memory retrieval.

Hao-Dong Lu, Dong Gong · 2 citations
#artificial intelligence Preprint Oct 2026

Self-Generated Feedback Destabilizes Test-Time Training: A Causal Decomposition of Long-Horizon Adaptation

Test-time training (TTT) lets a model store information in its weights during inference. When the model learns from its own output, however, each update also changes the model that generates the next training example. Across 128K-token streams, retaining generated-text updates worsens prediction on independent human-wr...

Cheng-Zhi Luo, Bing Li, Bernard Ghanem · 0 citations
#artificial intelligence Preprint Oct 2026

Follow the Winners: Conservative Policy Improvement with the Cross-Entropy Method for Critic-Free RFT

Critic-free reinforcement fine-tuning (RFT) for agentic large language models is often done through GRPO-style methods, which compute a group baseline over repeated rollouts to reduce target variance. However, this setup is ill-suited to agents acting in stateful environments such as live services or security sandboxes...

Joery A. de Vries, Neil D. Lawrence, Zhen-Wen Dai · 0 citations
#artificial intelligence Preprint Oct 2026

Asynchronous Is Nearly Free for Evolution Strategies on Long-Horizon Agentic Tasks

LLM-based long-horizon agentic post-training is often bottlenecked by rollout generation: trajectories span many interaction turns, completion times vary substantially, and synchronous update barriers leave faster workers waiting for stragglers. Asynchronous reinforcement learning which has been adopted in LLM post-tra...

William Hoy, Jing-Xuan Fan, Nurcin Celik et al. · 0 citations
Preprint Oct 2026

Self-Repairing Recurrent Ensembles for Real-Time Recovery from Distribution Shift

This work presents a method that lets a policy recover from such shifts online and without supervision, and recovers close to the original performance after a sensor shift, while ensembles that see the full observation are unable to recover.

Julian Lemmel, P. Wendel-Garcia, Taisuke Kobayashi et al. · 0 citations

Related blog posts

MIT News · Artificial Intelligence Nov 17, 2026

Learning the language of machine learning

By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.