Systematic review demonstrates personalized federated intelligence framework for foundation models, highlighting private adaptation across decentralized edge devices.
Key Points
To introduce personalized federated intelligence (PFI) as a framework that integrates federated learning with foundation models to achieve artificial personalized intelligence without compromising user privacy.
Conducted a comprehensive literature review synthesizing recent developments at the intersection of federated learning and foundation models.
Characterized the core pipeline stages of PFI, focusing on edge-level personalization, trustworthy model adaptation, and retrieval-augmented generation.
Identified privacy concerns, computational costs, and large model scales as primary bottlenecks limiting user-level customization of centralized foundation models.
Formulated the PFI architectural paradigm to enable decentralized, privacy-preserving fine-tuning and adaptive refinement on edge devices.
Outlined open challenges and critical future research trajectories necessary to transition from general foundation models to personalized edge intelligence.