Agentic Unlearning: When LLM Agent Meets Machine Unlearning

TL;DR

This paper introduces agentic unlearning, a unified framework for removing specific information from both model parameters and persistent memory in AI agents, addressing backflow issues and ensuring comprehensive unlearning.

Contribution

It proposes Synchronized Backflow Unlearning (SBU), a novel method that jointly unlearns from parameters and memory pathways using a dual-update protocol.

Findings

Reduces private information traces in models and memory

Maintains data utility with limited degradation

Effective on medical QA benchmarks

Abstract

In this paper, we introduce \textbf{agentic unlearning} which removes specified information from both model parameters and persistent memory in agents with closed-loop interaction. Existing unlearning methods target parameters alone, leaving two critical gaps: (i) parameter-memory backflow, where retrieval reactivates parametric remnants or memory artifacts reintroduce sensitive content, and (ii) the absence of a unified strategy that covers both parameter and memory pathways. We present Synchronized Backflow Unlearning (SBU), a framework that unlearns jointly across parameter and memory pathways. The memory pathway performs dependency closure-based unlearning that prunes isolated entities while logically invalidating shared artifacts. The parameter pathway employs stochastic reference alignment to guide model outputs toward a high-entropy prior. These pathways are integrated via a synchronized dual-update protocol, forming a closed-loop mechanism where memory unlearning and parametric suppression reinforce each other to prevent cross-pathway recontamination. Experiments on medical QA benchmarks show that SBU reduces traces of targeted private information across both pathways with limited degradation on retained data.