AI RESEARCH
Retrieval Mechanisms Surpass Long-Context Scaling in Time Series Forecasting
arXiv CS.LG
•
ArXi:2605.08217v1 Announce Type: new Time Series Foundation Models (TSFMs) have borrowed the long context paradigm from natural language processing under the premise that feeding history into the model improves forecast quality. But in stochastic domains, distant history is often just high-frequency noise, not signal. Hence, the proposed work tests whether this premise actually holds by running continuous context architectures (PatchTST included) through the ETTh1 benchmark.