Reproduce “Uncertainty-Aware Prototype Learning for Open-Set Test-Time Adaptation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jinwu Hu, Langyu Bian, Jian Sun, Dongjin Yang, Chao Shuai, Yufeng Wang, Kai Zhou, Bin Xiao and 3 more
UAPA adapts visual models to unlabeled streams containing shifted familiar and unfamiliar classes. It separates samples using classifier-prototype alignment and batch-relative uncertainty, then applies different entropy and prototype-based updates to each group. On CIFAR-10-C and CIFAR-100-C, it achieves average OSCR of 70.18% versus 68.10% for UniEnt; however, csOOD rejection remains difficult in a single-class-per-batch ViT-Base setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.