TY  - RPRT
TI  - Don't Stop Long-Context Continual Pre-Training Too Early: An Empirical Scaling Law
AU  - Ziqing Qiao
AU  - Xiaoyue Xu
AU  - Sikui Zhang
AU  - Yinuo Xu
AU  - Bingxiang He
AU  - Chaojun Xiao
AU  - Xu Han
PY  - 2026
UR  - https://arxiv.org/abs/2610.29352
ID  - 2610.29352
ER  -