mlsanityData to 7 Oct 2026

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

arXiv 2602.00620 ↗First posted 31 Jan 2026cs.LG cs.AIcited by 3

Not on the map. Why?

Abstract (arXiv)

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 for time series classification.The source code is publicly available at https://github.com/fangjuntao/TIC-FM.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 28 of its 36 references that are in the map. First posted on arXiv 31 Jan 2026.

Builds on

28 of 36 references.

References · 28

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  2. 2018Representation Learning with Contrastive Predictive Codingcited by 7,662
  3. 2016Layer Normalizationcited by 5,579
  4. 2016Gaussian Error Linear Units (GELUs)cited by 3,431
  5. 2021An Explanation of In-context Learning as Implicit Bayesian Inferencecited by 915
  6. 2024Chronos: Learning the Language of Time Seriescited by 712
  7. 2023A decoder-only foundation model for time-series forecastingcited by 650
  8. 2022TabPFN: A Transformer That Solves Small Tabular Classification Problems in a Secondcited by 643
  9. 2024Unified Training of Universal Time Series Forecasting Transformerscited by 588
  10. 2018Deep learning for time series classification: a reviewcited by 575
  11. 2019Are Transformers universal approximators of sequence-to-sequence functions?cited by 404
  12. 2025TabICL: A Tabular Foundation Model for In-Context Learning on Large Datacited by 267
16 more
  1. 2024Foundation Models for Time Series Analysis: A Tutorial and Surveycited by 234
  2. 2023One Step of Gradient Descent is Provably the Optimal In-Context Learner with One Layer of Linear Self-Attentioncited by 168
  3. 2025Chronos-2: From Univariate to Universal Forecastingcited by 158
  4. 2023Looped Transformers are Better at Learning Learning Algorithmscited by 135
  5. 2023The Learnability of In-Context Learningcited by 126
  6. 2025TiRex: Zero-Shot Forecasting Across Long and Short Horizons with Enhanced In-Context Learningcited by 107
  7. 2024Can Looped Transformers Learn to Implement Multi-step Gradient Descent for In-context Learning?cited by 60
  8. 2024Is In-Context Learning in Large Language Models Bayesian? A Martingale Perspectivecited by 51
  9. 2024In-Context Fine-Tuning for Time-Series Foundation Modelscited by 31
  10. 2024In-Context Learning of a Linear Transformer Block: Benefits of the MLP Component and One-Step GD Initializationcited by 30
  11. 2019Universal approximations of permutation invariant/equivariant functions by deep neural networkscited by 29
  12. 2024In-context Time Series Predictorcited by 21
  13. 2023NuTime: Numerically Multi-Scaled Embedding for Large-Scale Time-Series Pretrainingcited by 20
  14. 2023Data Augmentation for Time-Series Classification: An Extensive Empirical Study and Comprehensive Surveycited by 9
  15. 2025Pre-trained Forecasting Models: Strong Zero-Shot Feature Extractors for Time Series Classificationcited by 9
  16. 2025The Initialization Determines Whether In-Context Learning Is Gradient Descentcited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Forecasting Is Not Attribution: Localizing Decoder Bypass in Graph-Based Neural Marketing Mix Modelscited by 0
  2. 2026TimEE: End-to-end Time Series Classification via In-Context Learningcited by 0
  3. 2026ChorusTIC: Training-Free Multivariate Time Series Classification via Chorus In-Context Learningcited by 0