Reproduce “STAR: Boosting Time Series Foundation Models for Anomaly Detection through State-aware Adapter” 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

Hanyin Cheng, Ruitong Zhang, Yuning Lu, Peng Chen, Meng Wang, Yang Shu, Bin Yang, Chenjuan Guo

STAR uses operating states to guide time-series anomaly detection, encoding state identity, conditioning a pretrained model’s numerical processing, and matching state and value patches. Evaluated with 10% downstream data across five datasets and four models, STAR raises VUS-ROC—a threshold-independent ROC metric with tolerance around anomaly boundaries—on Genesis with DADA from 0.832 to 0.911, a result limited to that setting.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.