A Comparative Study on how Data Normalization Affects Zero-Shot Generalization in Time Series Foundation Models

We investigate input normalization methods for Time Series Foundation Models (TSFMs). While normalization is well-studied in dataset-specific time series models, it remains overlooked in TSFMs where generalization is critical. Time series data, unlike text or images, exhibits significant scale variation across domains and channels, coupled with non-stationarity, can undermine TSFM performance regardless of architectural complexity. Through systematic evaluation using four architecturally diverse TSFMs, we empirically establish RevIN as the most efficient approach, reducing zero-shot MASE by 89% relative to raw baseline and by 44% in comparison to other normalization methods, while matching the best in-domain accuracy without any dataset-level preprocessing—yielding the highest accuracy-efficiency trade-off. Yet its effective utilization depends on architectural design choices and optimization objective, particularly with respect to training loss scale sensitivity and model type (point-forecast-, distribution- or LLM-based models).