Erase Persona, Forget Lore: Benchmarking Multimodal Copyright Unlearning in Large Vision Language Models

June Hyoung Kwon, Jungmin Yun, Youngbin Kim


Abstract
Large Vision-Language Models (LVLMs), trained on web-scale data, risk memorizing and regenerating copyrighted visual content like characters and logos, creating significant challenges. Machine unlearning offers a path to mitigate these risks by removing specific content post-training, but evaluating its effectiveness, especially in the complex multimodal setting of LVLMs, remains an open problem. Current evaluation methods often lack robustness or fail to capture the nuances of cross-modal concept erasure. To address this critical gap, we introduce the CoVUBench benchmark, the first framework specifically designed for evaluating copyright content unlearning in LVLMs. CoVUBench utilizes procedurally generated, legally safe synthetic data coupled with systematic visual variations—spanning compositional changes and diverse domain manifestations—to ensure realistic and robust evaluation of unlearning generalization. Our comprehensive, multimodal evaluation protocol assesses both forgetting efficacy from the copyright holder’s perspective and the preservation of general model utility from the deployer’s viewpoint. By rigorously measuring this crucial trade-off, CoVUBench provides a standardized tool to advance the development of responsible and effective unlearning methods for LVLMs.
Anthology ID:
2026.lrec-1.727
Volume:
Proceedings of the Fifteenth Language Resources and Evaluation Conference
Month:
May
Year:
2026
Address:
Palma de Mallorca, Spain
Editors:
Stelios Piperidis, Núria Bel, Henk van den Heuvel, Nancy Ide, Simon Krek, Antonio Toral
Venue:
LREC
SIG:
Publisher:
ELRA Language Resource Association
Note:
Pages:
9254–9265
Language:
External URL:
https://lrec.elra.info/lrec2026-main-727
DOI:
10.63317/3zvek95uex2j
Bibkey:
Cite (ACL):
June Hyoung Kwon, Jungmin Yun, and Youngbin Kim. 2026. Erase Persona, Forget Lore: Benchmarking Multimodal Copyright Unlearning in Large Vision Language Models. In Proceedings of the Fifteenth Language Resources and Evaluation Conference, pages 9254–9265, Palma de Mallorca, Spain. ELRA Language Resource Association.
Cite (Informal):
Erase Persona, Forget Lore: Benchmarking Multimodal Copyright Unlearning in Large Vision Language Models (Kwon et al., LREC 2026)
Copy Citation: