This research introduces a novel Federated Prompt Learning (FPL) approach that enables personalized multimodal AI systems while preserving user privacy—crucial for sensitive customer support applications.

• Combines pre-trained multimodal LLMs with federated learning to create personalized AI systems
• Addresses the critical balance between personalization, model quality, and privacy protection
• Implements differential privacy techniques to safeguard user data
• Demonstrates viable performance while maintaining strong privacy guarantees

For support teams, this breakthrough enables personalized customer experiences across text, image, and audio modalities without compromising sensitive data—potentially revolutionizing how organizations deliver secure, tailored support.

Privacy-Preserving Personalized Federated Prompt Learning for Multimodal Large Language Models