Toward Understanding Unlearning Difficulty: A Mechanistic Perspective and Circuit-Guided Difficulty Metric

TL;DR

This paper introduces a circuit-based metric called CUD to quantify and understand the difficulty of unlearning specific samples in language models, revealing mechanistic insights into why some data are harder to erase.

Contribution

It proposes a novel circuit-guided difficulty metric for unlearning, linking unlearning challenges to model internal mechanisms and pathways.

Findings

CUD reliably distinguishes easy and hard-to-unlearn samples.

Easy samples are associated with shorter, earlier circuit interactions.

Hard samples involve longer, deeper circuit pathways near late-stage computation.

Abstract

Machine unlearning is becoming essential for building trustworthy and compliant language models. Yet unlearning success varies considerably across individual samples: some are reliably erased, while others persist despite the same procedure. We argue that this disparity is not only a data-side phenomenon, but also reflects model-internal mechanisms that encode and protect memorized information. We study this problem from a mechanistic perspective based on model circuits--structured interaction pathways that govern how predictions are formed. We propose Circuit-guided Unlearning Difficulty (CUD), a {\em pre-unlearning} metric that assigns each sample a continuous difficulty score using circuit-level signals. Extensive experiments demonstrate that CUD reliably separates intrinsically easy and hard samples, and remains stable across unlearning methods. We identify key circuit-level patterns that reveal a mechanistic signature of difficulty: easy-to-unlearn samples are associated with shorter, shallower interactions concentrated in earlier-to-intermediate parts of the original model, whereas hard samples rely on longer and deeper pathways closer to late-stage computation. Compared to existing qualitative studies, CUD takes a first step toward a principled, fine-grained, and interpretable analysis of unlearning difficulty; and motivates the development of unlearning methods grounded in model mechanisms.