Implicit Reasoning in Deep Time Series Forecasting

TL;DR

This paper investigates whether deep time series forecasting models demonstrate genuine reasoning abilities or rely on memorization, finding evidence of some reasoning capabilities in certain models through out-of-distribution generalization tests.

Contribution

It introduces an initial assessment of reasoning in deep time series models, highlighting their potential to generalize beyond memorization in OOD scenarios.

Findings

Certain models generalize effectively in OOD scenarios

Evidence of reasoning capabilities beyond memorization

Transformers and MLPs show promising generalization

Abstract

Recently, time series foundation models have shown promising zero-shot forecasting performance on time series from a wide range of domains. However, it remains unclear whether their success stems from a true understanding of temporal dynamics or simply from memorizing the training data. While implicit reasoning in language models has been studied, similar evaluations for time series models have been largely unexplored. This work takes an initial step toward assessing the reasoning abilities of deep time series forecasting models. We find that certain linear, MLP-based, and patch-based Transformer models generalize effectively in systematically orchestrated out-of-distribution scenarios, suggesting underexplored reasoning capabilities beyond simple pattern memorization.