Before Forgetting, Learn to Remember: Revisiting Foundational Learning Failures in LVLM Unlearning Benchmarks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kwon, JuneHyoung, Kim, MiHyeon, Lee, Eunju, Yun, JungMin, Lim, Byeonggeuk, Kim, YoungBin
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917462215229440
author Kwon, JuneHyoung
Kim, MiHyeon
Lee, Eunju
Yun, JungMin
Lim, Byeonggeuk
Kim, YoungBin
author_facet Kwon, JuneHyoung
Kim, MiHyeon
Lee, Eunju
Yun, JungMin
Lim, Byeonggeuk
Kim, YoungBin
contents While Large Vision-Language Models (LVLMs) offer powerful capabilities, they pose privacy risks by unintentionally memorizing sensitive personal information. Current unlearning benchmarks attempt to mitigate this using fictitious identities but overlook a critical stage 1 failure: models fail to effectively memorize target information initially, rendering subsequent unlearning evaluations unreliable. Diagnosing under-memorization and the multi-hop curse as root causes, we introduce ReMem, a Reliable Multi-hop and Multi-image Memorization Benchmark. ReMem ensures robust foundational learning through principled data scaling, reasoning-aware QA pairs, and diverse visual contexts. Additionally, we propose a novel Exposure metric to quantify the depth of information erasure from the model's internal probability distribution. Extensive experiments demonstrate that ReMem provides a rigorous and trustworthy framework for diagnosing both learning and unlearning behaviors in LVLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2605_03759
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Before Forgetting, Learn to Remember: Revisiting Foundational Learning Failures in LVLM Unlearning Benchmarks
Kwon, JuneHyoung
Kim, MiHyeon
Lee, Eunju
Yun, JungMin
Lim, Byeonggeuk
Kim, YoungBin
Computer Vision and Pattern Recognition
Artificial Intelligence
While Large Vision-Language Models (LVLMs) offer powerful capabilities, they pose privacy risks by unintentionally memorizing sensitive personal information. Current unlearning benchmarks attempt to mitigate this using fictitious identities but overlook a critical stage 1 failure: models fail to effectively memorize target information initially, rendering subsequent unlearning evaluations unreliable. Diagnosing under-memorization and the multi-hop curse as root causes, we introduce ReMem, a Reliable Multi-hop and Multi-image Memorization Benchmark. ReMem ensures robust foundational learning through principled data scaling, reasoning-aware QA pairs, and diverse visual contexts. Additionally, we propose a novel Exposure metric to quantify the depth of information erasure from the model's internal probability distribution. Extensive experiments demonstrate that ReMem provides a rigorous and trustworthy framework for diagnosing both learning and unlearning behaviors in LVLMs.
title Before Forgetting, Learn to Remember: Revisiting Foundational Learning Failures in LVLM Unlearning Benchmarks
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2605.03759