Reproduce “Training on the Future: A Delay-Aware Audit of Test-Time Adaptation 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
Mohamed Readh Fentazi, Mazene Ameur, Adlen Ksentini
The authors audit time-series test-time adaptation under delayed label release, modeling when forecast windows become usable and allowing additional pipeline lag. They compare published methods with closed-form correctors across five benchmarks. On one ETTh2/DLinear stream, apparent adapter gains over persistence at early updates reverse at causal delay; results elsewhere are mixed, with all audited methods improving on ETTm1.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.