TY  - RPRT
TI  - CoVSpec: Efficient Device-Edge Co-Inference for Vision-Language Models via Speculative Decoding
AU  - Yuanyuan Jia
AU  - Shunpu Tang
AU  - Qianqian Yang
PY  - 2026
UR  - https://arxiv.org/abs/2605.02218
ID  - 2605.02218
ER  -