Wang, Xiaodong
Chen, Ying
Liu, Xiaohong
Qiu, Cen
Tang, Hong
Huang, Tinggui
Guo, Siqi
Ma, Sainan
Cai, Mengjiao
Sun, Qingyun
Chang, Zichen
Liu, Jinge
Wang, Xiongjun
Li, Jinda
Qian, Wulei
Wang, Biyu
Zhang, Boan
Bai, Chenguang
Shi, Min
Zhang, Xinlei
Li, Meng
Wang, Jiahai
Wang, Bin
Ma, Jinlu
Ai, Lirong
Yu, Shaoqing
Wang, Liming
Feng, Ninghan
Liu, Xiyang
Yu, Guanzhen
Funding for this research was provided by:
National Natural Science Foundation of China (82103037)
National Natural Science Foundation of China (82173306)
National Natural Science Foundation of China (82370777)
National Natural Science Foundation of China (82172860)
National Natural Science Foundation of China (82473085)
Article History
Received: 23 December 2025
Accepted: 9 March 2026
First Online: 21 March 2026
Competing interests
: The authors declare the following competing interests: X.L., L.W., X.W., G.Y., Y.C., and T.H. are inventors of a pending Chinese patent application (Application No. CN202511104724.5) filed by Xidian University. This patent relates to the uncertainty-aware module and the uncertainty-guided active learning framework for model training and annotation described in this manuscript. The patent is currently in the “Publication of Invention Patent Application” status. G.Y. is the founder of Jingguan Biotech, which may have a potential commercial interest in future independent applications related to this work. The UPATHLN platform described in this study is an academic research prototype; Jingguan Biotech did not fund this study and had no role in it. All other authors declare no competing interests.