Reproduce “STRIDE: Reasoning-Aware Training for 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

Md Atik Ahamed, Mihir Parmar, Palash Goyal, Chun-Liang Li, Qiang Cheng, Tomas Pfister, Jinsung Yoon

STRIDE combines language-model reasoning with forecasting. It trains a small language model to distill reasoning about historical data, then projects its internal representation into a numerical forecaster that predicts quantiles. On GIFT-Eval, adding STRIDE to Synapse lowers MASE from 0.699 to 0.625 and CRPS from 0.479 to 0.423; these ensemble results do not establish the same gains for STRIDE alone.

Summary by alphaXiv.

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