TY  - RPRT
TI  - WinQ: Accelerating Quantization-Aware Training of Language Models Around Saddle Points
AU  - Dongyue Li
AU  - Zechun Liu
AU  - Kai Yi
AU  - Zhenshuo Zhang
AU  - Changsheng Zhao
AU  - Raghuraman Krishnamoorthi
AU  - Harshit Khaitan
AU  - Hongyang R. Zhang
AU  - Steven Li
PY  - 2026
UR  - https://arxiv.org/abs/2605.17471
ID  - 2605.17471
ER  -