Reproduce “AnchorMTS: Multi-View Signal Representation and Forecast-Anchored Fusion for Long-Term Time-Series Forecasting” 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
Gao Wu, Tingting Xu, Changning Wu, Tengfei Long, Yong Liu, Jiabin Jin, Kexin Zhang
AnchorMTS uses visual and textual descriptions of observed histories to refine a complete numerical forecast. A numerical backbone produces an anchor, while auxiliary evidence proposes a correction capped at 25% of each variable’s anchor forecast norm. Across six benchmarks and 24 dataset–horizon cells, it reports macro MSE/MAE of 0.336/0.367 versus TimePro’s 0.348/0.374; the bound is exact, with zero evaluated violations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.