arXiv 2510.25502
TempoPFN: Synthetic Pre-training of Linear RNNs for Zero-shot Time Series Forecasting
By Vladyslav Moroshan, Julien Siems, et al.
Published 2025-10-29
Discussion
Read the public discussion and references gathered around this paper.
Foundation models for zero-shot time series forecasting face challenges in efficient long-horizon prediction and reproducibility, with existing synthetic-only approaches underperforming on challenging benchmarks. This paper presents TempoPFN, a univariate time series foundation model based on linear Recurrent Neural Networks (RNNs) pre-trained exclusively on synthetic data. The model uses a GatedDeltaProduct archite…