December 4, 2025 · papers
SLED presented at ACM/IEEE Symposium on Edge Computing
SLED brings speculative decoding to edge serving, where the draft model and the target model may sit on different machines with a slow link between them. Presented at the Tenth ACM/IEEE Symposium on Edge Computing in Washington, D.C., led by Xiangchen Li with Dimitris Spatharakis, Saeid Ghafouri and Jiakun Fan, and collaborators at Queen's University Belfast and University College Dublin.