Reproduce “RACE: Residual-Aware Test-Time Adaptation for Neighbor-Rich Time-Series Foundation Model 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

Hao-Nan Shi, Tong Wu, Chen-Cong Sun, Yuan Jiang, Han-Jia Ye, De-Chuan Zhan

The authors introduce RACE, which uses a frozen time-series foundation model’s past forecast errors to propose corrections from related histories. A consensus filter screens retrieved residuals, and a development-fitted Gate decides whether to apply each proposal. Across four backbones, RACE improves aggregate metrics in Web and CGM; Chronos-2’s CGM gains are 20.814% MSE, 10.677% MAE, and 9.939% CRPS, though results vary.

Summary by alphaXiv.

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