We propose TimePre, a simple framework that unifies the efficiency of Multilayer Perceptron (MLP)-based models with the distributional flexibility of Multiple Choice Learning (MCL) for Probabilistic Time-Series Forecasting (PTSF). Stabilized Instance Normalization (SIN), the core of TimePre, is a normalization layer that explicitly addresses the trade-off among accuracy, efficiency, and stability. SIN stabilizes the hybrid architecture by correcting channel-wise statistical shifts, thereby resolving the catastrophic hypothesis collapse. Extensive experiments on six benchmark datasets demonstrate that TimePre achieves state-of-the-art (SOTA) accuracy on key probabilistic metrics. Critically, TimePre achieves inference speeds that are orders of magnitude faster than sampling-based models, and is more stable than prior MCL approaches.
@article{arxiv.2511.18539,
title = {TimePre: Bridging Accuracy, Efficiency, and Stability in Probabilistic Time-Series Forecasting},
author = {Lingyu Jiang and Lingyu Xu and Peiran Li and Dengzhe Hou and Qianwen Ge and Dingyi Zhuang and Shuo Xing and Wenjing Chen and Xiangbo Gao and Ting-Hsuan Chen and Xueying Zhan and Xin Zhang and Ziming Zhang and Zhengzhong Tu and Michael Zielewski and Kazunori Yamada and Fangzhou Lin},
journal= {arXiv preprint arXiv:2511.18539},
year = {2026}
}