Abstract
Model editing offers a fast, targeted approach to correct post-deployment mistakes in medical vision-language models (VLMs) without costly retraining. However, existing multimodal model editing benchmarks focus on general-purpose tasks and fail to reflect realistic clinical domain requirements and variability. To address this, we introduce M3Bench, a clinically grounded benchmark for multimodal model editing that evaluates whether an edit remains reliable, precise, and generalizable under the challenges of image and text variation, modality and protocol shifts, clinical knowledge composition, and temporal progression.
M3Bench contains 16,276 questions spanning diverse anatomy, modalities, and specialties, and supports both single and sequential edits. By evaluating 4 representative editors across 6 medical and general VLMs, we find that no method excels across all criteria. Gradient-based editors achieve strong transfer but suffer from catastrophic locality violations, whereas memory-based methods preserve locality but lack compositional generality and exhibit high backbone-dependent hyperparameter sensitivity. We further attribute these failures to the latent space geometry of VLMs and how different editing methods shift its landscape.
Overall, M3Bench establishes a rigorous clinical stress test for multimodal model editing and offers actionable guidance for safer post-deployment adaptation. The benchmark is publicly available at M3Bench GitHub.
Blogger's Review: The introduction of M3Bench provides a significant evaluation standard for multimodal model editing in the medical field, especially regarding clinical adaptability. By deeply analyzing the pros and cons of different editing methods, researchers can better understand model performance in real-world applications, thereby enhancing the safety and reliability of medical AI.