Representation-Aware Unlearning via Activation Signatures: From Suppression to Entity-Signature Erasure

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mahmood, Syed Naveed, Bhuiyan, Md. Rezaur Rahman, Zaman, Tasfia, Khondaker, Jareen Tasneem, Sakib, Md. Sameer, Wadith, K. M. Shadman, Tasnim, Nazia, Sadeque, Farig
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911719169720320
author Mahmood, Syed Naveed
Bhuiyan, Md. Rezaur Rahman
Zaman, Tasfia
Khondaker, Jareen Tasneem
Sakib, Md. Sameer
Wadith, K. M. Shadman
Tasnim, Nazia
Sadeque, Farig
author_facet Mahmood, Syed Naveed
Bhuiyan, Md. Rezaur Rahman
Zaman, Tasfia
Khondaker, Jareen Tasneem
Sakib, Md. Sameer
Wadith, K. M. Shadman
Tasnim, Nazia
Sadeque, Farig
contents Entity-level unlearning is usually evaluated by what a model says: whether it stops naming the target, refuses a query, or shifts a Truth Ratio distribution. These output-level tests, however, do not show whether a subject's internal representation has been attenuated. We introduce the Entity Representation Unlearning Framework (ERUF), a representation-aware framework that mines subject-specific activation signatures, suppresses the corresponding activation direction, and distills the behavior into LoRA parameters. Among evaluated baselines, ERUF is the only method that jointly achieves surface-level suppression, internal attenuation, and utility preservation. On TOFU forget10, ERUF achieves FQ = 0.99 and MU = 0.62, matching reported oracle utility while approaching oracle forget quality. Across most standard foundation-model settings, ERUF maintains low leakage and low internal target activation, with SMR between 0.00% and 1.10%, EL10 below 0.06, and utility drift below 3%. On Llama-3.1-8B, adversarial entity recovery falls from 63.89% to 20.15%, while name-agnostic recovery decreases by 72.7% to 77.4%. Joint surface/internal diagnostics further reveal scale-dependent behavior in reasoning-prior models that surface metrics alone would miss. We interpret these results as operational evidence of representation-level attenuation, not as a formal guarantee of irreversible deletion.
format Preprint
id arxiv_https___arxiv_org_abs_2601_10566
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Representation-Aware Unlearning via Activation Signatures: From Suppression to Entity-Signature Erasure
Mahmood, Syed Naveed
Bhuiyan, Md. Rezaur Rahman
Zaman, Tasfia
Khondaker, Jareen Tasneem
Sakib, Md. Sameer
Wadith, K. M. Shadman
Tasnim, Nazia
Sadeque, Farig
Computation and Language
Machine Learning
I.2.7; I.2.6
Entity-level unlearning is usually evaluated by what a model says: whether it stops naming the target, refuses a query, or shifts a Truth Ratio distribution. These output-level tests, however, do not show whether a subject's internal representation has been attenuated. We introduce the Entity Representation Unlearning Framework (ERUF), a representation-aware framework that mines subject-specific activation signatures, suppresses the corresponding activation direction, and distills the behavior into LoRA parameters. Among evaluated baselines, ERUF is the only method that jointly achieves surface-level suppression, internal attenuation, and utility preservation. On TOFU forget10, ERUF achieves FQ = 0.99 and MU = 0.62, matching reported oracle utility while approaching oracle forget quality. Across most standard foundation-model settings, ERUF maintains low leakage and low internal target activation, with SMR between 0.00% and 1.10%, EL10 below 0.06, and utility drift below 3%. On Llama-3.1-8B, adversarial entity recovery falls from 63.89% to 20.15%, while name-agnostic recovery decreases by 72.7% to 77.4%. Joint surface/internal diagnostics further reveal scale-dependent behavior in reasoning-prior models that surface metrics alone would miss. We interpret these results as operational evidence of representation-level attenuation, not as a formal guarantee of irreversible deletion.
title Representation-Aware Unlearning via Activation Signatures: From Suppression to Entity-Signature Erasure
topic Computation and Language
Machine Learning
I.2.7; I.2.6
url https://arxiv.org/abs/2601.10566