Reproduce “JudgeCast: Time Series Forecasting with Experience-Informed Covariate Judgements” 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
Donguk Kwon, Wooseok Jeong, Dongha Lee
JudgeCast uses frozen time-series foundation model forecasts and a frozen large language model to judge contextual covariates and revise forecasts using experience. It tests alternative judgments against forecast residuals, retaining experience only when adjustment improves on no adjustment. Across seven datasets, it had lowest MSE and MAE, with average reductions of 16.7% and 6.5%; this measures forecast improvement, not causal accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.