Collaborative Edge-to-Server Inference for Vision-Language Models

Soochang Song, Yongjune Kim

We propose a collaborative edge-to-server inference framework for vision-language models (VLMs) that reduces communication cost while maintaining inference accuracy. In typical dep…