Rethinking Zero-Shot Time Series Classification: From Task-specific Classifiers to In-Context Inference

TL;DR

This paper introduces TIC-FM, a zero-shot time series classification framework that uses in-context learning with a Transformer, eliminating the need for task-specific classifiers and improving accuracy especially with limited labels.

Contribution

The paper proposes TIC-FM, a novel in-context learning approach for zero-shot time series classification that subsumes traditional classifiers and enhances transferability without training.

Findings

Strong accuracy on 128 UCR datasets

Consistent gains in low-label scenarios

Training-free transfer performance

Abstract

The zero-shot evaluation of time series foundation models (TSFMs) for classification typically uses a frozen encoder followed by a task-specific classifier. However, this practice violates the training-free premise of zero-shot deployment and introduces evaluation bias due to classifier-dependent training choices. To address this issue, we propose TIC-FM, an in-context learning framework that treats the labeled training set as context and predicts labels for all test instances in a single forward pass, without parameter updates. TIC-FM pairs a time series encoder and a lightweight projection adapter with a split-masked latent memory Transformer. We further provide theoretical justification that in-context inference can subsume trained classifiers and can emulate gradient-based classifier training within a single forward pass. Experiments on 128 UCR datasets show strong accuracy, with consistent gains in the extreme low-label situation, highlighting training-free transfer