The foundation models possess cognitive computing capabilities to handle complex general multi-media tasks and are typically deployed in the cloud. Correspondingly, specialized models are designed to be lightweight, goal-oriented, capable of quick responses, and adaptive iterations, making them potentially suitable for deployment on resource-constrained end devices. However, mainstream cloud-centric learning paradigms exhibit deficiencies in real-time performance, personalization, load cost, and privacy security, often neglecting the capabilities of device-based specialized models. In recent years, an increasing number of researchers have focused on the collaboration and evolution between multi-media cloud-based foundation models and device-based specialized models to propose innovative solutions that will ultimately benefit various applications.

This workshop is relevant to ACM Multimedia Asia, as the collaboration and evolution of multi-media foundation models and specialized models can be applied to numerous multimedia and visual application problems (e.g., autonomous driving, video rendering, robots). This workshop would be crucial as it will emphasize that, despite significant breakthroughs in recent years, more work is needed to support the stable and efficient deployment of collaborative AI models in heterogeneous, complex, and dynamic application scenarios. 

The main goal of this workshop is to encourage researchers from academia and industry to gather, share their encountered challenges, and discuss potential solutions. We hope this workshop will serve as a platform to promote the development of research on the collaboration and evolution of foundation and specialized models. 

We especially welcome BRAVE NEW IDEAs, i.e., papers without complete methods or experiments.

CEFSW Workshop will be held on Dec 3, 2024.