Reproduce “LLM-Assisted Logic Rule Learning: Scaling Human Expertise for Time Series Anomaly Detection” 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
Haoting Zhang, Shekhar Jain
The authors use multimodal LLMs and expert review to turn stock-health examples into explicit rules for detecting weekly product anomalies. The rules combine qualitative patterns with quantitative time-series features and are iteratively refined against precision, recall, and F1. On 10,000 ASINs, they score 92.01% F1 in 4.27 seconds, versus Claude Sonnet 4’s 95.98%; labels are LLM-generated and expert-reviewed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.