paperIEEE Xplore·1 Nov 2026·Weining Chen, Yongheng Deng, Xiang Wang, Ziyan Fu, Xiangyu Li
DWI: Efficient On-Device LLM Decoding Through Dynamic Width Selection on Nested Models
DWI speeds up on-device LLM decoding by picking the width of a nested model per step.
Extract
IEEE Transactions on Mobile Computing