OFFSIDE: Benchmarking Unlearning Misinformation in Multimodal Large Language Models

TL;DR

OFFSIDE introduces a realistic multimodal unlearning benchmark for misinformation in vision-language models, centered on football transfer rumors. It provides four evaluation settings—Complete Unlearning, Selective Unlearning, Corrective Relearning, and Unimodal Unlearning—and a 15.68K VQA dataset across 80 players to probe forgetting, generalization, utility, and robustness. Across five baselines and multiple modalities, the study finds unimodal unlearning largely fails in multimodal contexts, visual rumors resist current methods, and relearning can recover forgotten content, with catastrophic forgetting driving much of the observed behavior. The results highlight the need for genuinely multimodal, continual-learning–aware unlearning techniques and establish OFFSIDE as a realistic, deployment-relevant benchmark with publicly available code.

Abstract

Advances in Multimodal Large Language Models (MLLMs) intensify concerns about data privacy, making Machine Unlearning (MU), the selective removal of learned information, a critical necessity. However, existing MU benchmarks for MLLMs are limited by a lack of image diversity, potential inaccuracies, and insufficient evaluation scenarios, which fail to capture the complexity of real-world applications. To facilitate the development of MLLMs unlearning and alleviate the aforementioned limitations, we introduce OFFSIDE, a novel benchmark for evaluating misinformation unlearning in MLLMs based on football transfer rumors. This manually curated dataset contains 15.68K records for 80 players, providing a comprehensive framework with four test sets to assess forgetting efficacy, generalization, utility, and robustness. OFFSIDE supports advanced settings like selective unlearning and corrective relearning, and crucially, unimodal unlearning (forgetting only text data). Our extensive evaluation of multiple baselines reveals key findings: (1) Unimodal methods (erasing text-based knowledge) fail on multimodal rumors; (2) Unlearning efficacy is largely driven by catastrophic forgetting; (3) All methods struggle with "visual rumors" (rumors appear in the image); (4) The unlearned rumors can be easily recovered and (5) All methods are vulnerable to prompt attacks. These results expose significant vulnerabilities in current approaches, highlighting the need for more robust multimodal unlearning solutions. The code is available at \href{https://github.com/zh121800/OFFSIDE}{https://github.com/zh121800/OFFSIDE}.

Figures (8)

• Figure 1: OFFSIDE is a comprehensive benchmark for MLLMs MU, featuring four real-world settings designed to address the removal of various rumors. Texts in red represent the target rumor, while those in green indicate successful forgetting or relearning.
• Figure 2: Overview of the OFFSIDE framework. The MLLM is first fine-tuned on the forget and retain set to obtain the vanilla model, during which it learns the rumors associated with each player. Various unlearning methods are then applied on forget set to obtain the unlearned model. After unlearning, the model is fine-tuned on the relearn set to simulate a continual learning setting. Performance is evaluated on four distinct subsets after both the unlearning and relearning stages.
• Figure 3: Results of the Unimodal Unlearning. RS, TS, FS represent retain set, test set, and forget set, respectively. CA, GS, FS refer to classification accuracy, generation score, and fact score, respectively.
• Figure 4: Illustration of experimental conclusions, observed from the OFFSIDE benchmark.
• Figure 5: Case study of four unlearning settings, each simulating a real-world MLLM unlearning scenario.