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

arXiv: https://arxiv.org/abs/2512.02833 · HTML: https://mlsanity.com/paper/2512.02833/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 21 of its 31 references that are in the map.
- First posted on arXiv 2 Dec 2025.

## Builds on (most cited first, 21 of 21)

- [Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting](https://mlsanity.com/paper/2012.07436/) (2020)
- [DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks](https://mlsanity.com/paper/1704.04110/) (2017)
- [Chronos: Learning the Language of Time Series](https://mlsanity.com/paper/2403.07815/) (2024)
- [Time-LLM: Time Series Forecasting by Reprogramming Large Language Models](https://mlsanity.com/paper/2310.01728/) (2023)
- [A decoder-only foundation model for time-series forecasting](https://mlsanity.com/paper/2310.10688/) (2023)
- [Unified Training of Universal Time Series Forecasting Transformers](https://mlsanity.com/paper/2402.02592/) (2024)
- [MOMENT: A Family of Open Time-series Foundation Models](https://mlsanity.com/paper/2402.03885/) (2024)
- [SCINet: Time Series Modeling and Forecasting with Sample Convolution and Interaction](https://mlsanity.com/paper/2106.09305/) (2021)
- [Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Experts](https://mlsanity.com/paper/2409.16040/) (2024)
- [Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Series](https://mlsanity.com/paper/2401.03955/) (2024)
- [ForecastPFN: Synthetically-Trained Zero-Shot Forecasting](https://mlsanity.com/paper/2311.01933/) (2023)
- [ChatTime: A Unified Multimodal Time Series Foundation Model Bridging Numerical and Textual Data](https://mlsanity.com/paper/2412.11376/) (2024)
- [Moirai-MoE: Empowering Time Series Foundation Models with Sparse Mixture of Experts](https://mlsanity.com/paper/2410.10469/) (2024)
- [This Time is Different: An Observability Perspective on Time Series Foundation Models](https://mlsanity.com/paper/2505.14766/) (2025)
- [Frequency Adaptive Normalization For Non-stationary Time Series Forecasting](https://mlsanity.com/paper/2409.20371/) (2024)
- [Towards Neural Scaling Laws for Time Series Foundation Models](https://mlsanity.com/paper/2410.12360/) (2024)
- [LightGTS: A Lightweight General Time Series Forecasting Model](https://mlsanity.com/paper/2506.06005/) (2025)
- [The Effectiveness of Discretization in Forecasting: An Empirical Study on Neural Time Series Models](https://mlsanity.com/paper/2005.10111/) (2020)
- [Out-of-Distribution Generalization in Time Series: A Survey](https://mlsanity.com/paper/2503.13868/) (2025)
- [Nonstationary Time Series Forecasting via Unknown Distribution Adaptation](https://mlsanity.com/paper/2402.12767/) (2024)
- [VSFormer: Value and Shape-Aware Transformer with Prior-Enhanced Self-Attention for Multivariate Time Series Classification](https://mlsanity.com/paper/2412.16515/) (2024)

## Built on by (most cited first, 1 of 1)

- [Does Normalization Choice Matter for Causal Large Time-Series Models?](https://mlsanity.com/paper/2606.09954/) (2026)

## Abstract (arXiv)

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 across four architecturally diverse TSFMs, we empirically establish REVIN as the most efficient approach, reducing zero-shot MASE by 89% relative to an un-normalized baseline and by 44% versus other normalization methods, while matching the best in-domain accuracy (0.84 MASE) without any dataset-level preprocessing -- yielding the highest accuracy-efficiency trade-off. Yet its effect utilization depends on architectural design choices and optimization objective, particularly with respect to training loss scale sensitivity and model type (probabilistic, point-forecast, or LLM-based models).