Ask Without Telling: Local SLMs Consult Cloud LLMs Without Revealing Task Intent
PriCon, an end-to-end framework that transforms the task itself through recoverable mathematical reformulation rather than hiding it among alternatives, is proposed, which reduces cloud-side task-intent inference Hit@1 to nearly 0, and preserves cloud-assisted utility.
Yan-Meng Wang, Yun-Xuan Li, Shi-Long Fan et al.
· 0 citations