Timer-S1: A Billion-Scale Time Series Foundation Model with Serial Scaling
This work addresses the scalability bottleneck in pre-trained time series foundation models, offering improved long-term prediction capabilities for researchers and practitioners working with large-scale time series data.
The paper introduces Timer-S1, an 8.3B parameter Mixture-of-Experts time series foundation model, designed to overcome scalability issues in existing models. It achieves state-of-the-art forecasting performance on the GIFT-Eval leaderboard, demonstrating the best MASE and CRPS scores among pre-trained models.
We introduce Timer-S1, a strong Mixture-of-Experts (MoE) time series foundation model with 8.3B total parameters, 0.75B activated parameters for each token, and a context length of 11.5K. To overcome the scalability bottleneck in existing pre-trained time series foundation models, we perform Serial Scaling in three dimensions: model architecture, dataset, and training pipeline. Timer-S1 integrates sparse TimeMoE blocks and generic TimeSTP blocks for Serial-Token Prediction (STP), a generic training objective that adheres to the serial nature of forecasting. The proposed paradigm introduces serial computations to improve long-term predictions while avoiding costly rolling-style inference and pronounced error accumulation in the standard next-token prediction. Pursuing a high-quality and unbiased training dataset, we curate TimeBench, a corpus with one trillion time points, and apply meticulous data augmentation to mitigate predictive bias. We further pioneer a post-training stage, including continued pre-training and long-context extension, to enhance short-term and long-context performance. Evaluated on the large-scale GIFT-Eval leaderboard, Timer-S1 achieves state-of-the-art forecasting performance, attaining the best MASE and CRPS scores as a pre-trained model. Timer-S1 will be released to facilitate further research.