Reproduce “NV-Tesseract-AD: Curriculum-Enhanced Conditional Diffusion for Multivariate 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
Aditi Gautam, Manikandan Ravikiran, Saira Qureshi, Sena Fatma Ekiz, Charan Jagwani, Mahmoud Elhadidy, Sanjeev Roka
NV-Tesseract-AD trains a Transformer denoiser for timestamp-level anomaly detection in multivariate sensor streams by reconstructing observed values withheld from context. It progressively increases both the masking fraction and diffusion-noise timestep, excluding structurally missing values from training loss and anomaly scores. On TSB-UAD-M, point-adjusted F1 rises from 0.0376 to 0.1404, but preprocessing changes confound attribution to the curriculum alone.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.