F³OCUS introduces a novel approach for efficient federated fine-tuning of large vision-language models across distributed medical devices with limited resources.

• Implements layer-specific importance scoring to identify the most critical model layers for fine-tuning on each client
• Utilizes inter-client layer diversity to encourage different devices to focus on complementary parts of the model
• Employs multi-objective meta-heuristics to optimize the selection strategy across the federation
• Demonstrates significant performance gains on medical image analysis tasks while reducing computational burden

This research enables more effective deployment of advanced vision-language models in healthcare settings where data privacy is critical and computing resources are limited.

F³OCUS -- Federated Finetuning of Vision-Language Foundation Models with Optimal Client Layer Updating Strategy via Multi-objective Meta-Heuristics