Frequency Matters: When Time Series Foundation Models Fail Under Spectral Shift

• URL: http://arxiv.org/abs/2511.05619v1
• Date: Thu, 06 Nov 2025 21:08:51 GMT
• Title: Frequency Matters: When Time Series Foundation Models Fail Under Spectral Shift
• Abstract summary: Time series foundation models (TSFMs) have shown strong results on public benchmarks, prompting comparisons to a "BERT moment" for time series.<n>We examine why TSFMs often struggle to generalize and highlight spectral shift (a mismatch between the dominant frequency components in downstream tasks and those represented during pretraining) as a key factor.<n>We present evidence from an industrial-scale player engagement prediction task in mobile gaming, where TSFMs underperform domain-adapted baselines.
• Score: 6.557492981844721
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Time series foundation models (TSFMs) have shown strong results on public benchmarks, prompting comparisons to a "BERT moment" for time series. Their effectiveness in industrial settings, however, remains uncertain. We examine why TSFMs often struggle to generalize and highlight spectral shift (a mismatch between the dominant frequency components in downstream tasks and those represented during pretraining) as a key factor. We present evidence from an industrial-scale player engagement prediction task in mobile gaming, where TSFMs underperform domain-adapted baselines. To isolate the mechanism, we design controlled synthetic experiments contrasting signals with seen versus unseen frequency bands, observing systematic degradation under spectral mismatch. These findings position frequency awareness as critical for robust TSFM deployment and motivate new pretraining and evaluation protocols that explicitly account for spectral diversity.

Related papers

• FAiT: Frequency-Aware Inverted Transformer for Multivariate Time Series Forecasting [17.617858319449443]
  FAiT is a Frequency-Aware inverted Transformer that rectifies the spectral bias internally through Inverted Attention.<n>DTFM synthesizes instance-conditioned weights to adaptively re-calibrate the energy of spectral sub-bands.<n>Experiments on widely used benchmarks demonstrate that FAiT consistently outperforms state-of-the-art Transformer-based and frequency-enhanced baselines.
  arXiv  Detail & Related papers  (2026-05-31T15:51:22Z)
• FEDIN: Frequency-Enhanced Deep Interest Network for Click-Through Rate Prediction [56.59654766009799]
  We present a novel empirical observation: user attention scores exhibit distinct spectral entropy distributions when conditioned on positive or negative target items.<n>We propose the Frequency-Enhanced Deep Interest Network (FEDIN) to capture latent periodic patterns in user interests.<n>FEDIN consistently outperforms state-of-the-art sequential recommendation baselines, demonstrating superior robustness against noise.
  arXiv  Detail & Related papers  (2026-05-03T05:50:36Z)
• Diversified Scaling Inference in Time Series Foundation Models [17.268760626931517]
  This work systematically investigates two questions: how do TSFMs behave under standard sampling-based inference scaling, and can controlled sampling diversity enhance performance?<n>We first examine the properties of TSFMs under standard sampling often fail to adhere to scaling laws due to insufficient exploration of the solution space.<n>We then delve into diversified inference scaling via tailored time series perturbations to expand the generative distribution's support.
  arXiv  Detail & Related papers  (2026-01-24T08:53:42Z)
• Towards Non-Stationary Time Series Forecasting with Temporal Stabilization and Frequency Differencing [24.5925492214521]
  Time series often exhibit non-stationarity, including temporal distribution shifts and spectral variability.<n>We propose DTAF, a dual-branch framework that addresses non-stationarity in both the temporal and frequency domains.
  arXiv  Detail & Related papers  (2025-11-11T13:30:38Z)
• Synapse: Adaptive Arbitration of Complementary Expertise in Time Series Foundational Models [50.877082340479085]
  We study how different Time Series Foundational Models (TSFMs) exhibit specialized performance profiles across various forecasting settings.<n>We propose Synapse, a novel arbitration framework for TSFMs.<n>Results demonstrate that Synapse consistently outperforms other popular ensembling techniques as well as individual TSFMs.
  arXiv  Detail & Related papers  (2025-11-07T18:01:51Z)
• A Unified Frequency Domain Decomposition Framework for Interpretable and Robust Time Series Forecasting [81.73338008264115]
  Current approaches for time series forecasting, whether in the time or frequency domain, predominantly use deep learning models based on linear layers or transformers.<n>We propose FIRE, a unified frequency domain decomposition framework that provides a mathematical abstraction for diverse types of time series.<n>Fire consistently outperforms state-of-the-art models on long-term forecasting benchmarks.
  arXiv  Detail & Related papers  (2025-10-11T09:59:25Z)
• Frequency-Constrained Learning for Long-Term Forecasting [15.31488551912888]
  Real-world time series exhibit strong periodic structures arising from physical laws, human routines, or seasonal cycles.<n>Modern deep forecasting models often fail to capture these recurring patterns due to spectral bias and a lack of frequency-aware inductive priors.<n>We propose a simple yet effective method that enhances long-term forecasting by explicitly modeling periodicity.
  arXiv  Detail & Related papers  (2025-08-02T22:12:15Z)
• MFRS: A Multi-Frequency Reference Series Approach to Scalable and Accurate Time-Series Forecasting [51.94256702463408]
  Time series predictability is derived from periodic characteristics at different frequencies.<n>We propose a novel time series forecasting method based on multi-frequency reference series correlation analysis.<n> Experiments on major open and synthetic datasets show state-of-the-art performance.
  arXiv  Detail & Related papers  (2025-03-11T11:40:14Z)
• TS-RAG: Retrieval-Augmented Generation based Time Series Foundation Models are Stronger Zero-Shot Forecaster [14.512119661418522]
  We present TS-RAG, a retrieval-augmented generation framework for time series forecasting.<n>Specifically, TS-RAG leverages pre-trained time series encoders to retrieve semantically relevant segments from a dedicated knowledge base.<n>We show that TS-RAG achieves state-of-the-art zero-shot forecasting performance, outperforming the existing TSFMs by up to 6.84% across diverse domains.
  arXiv  Detail & Related papers  (2025-03-06T16:48:48Z)
• Evaluating Time Series Foundation Models on Noisy Periodic Time Series [0.0]
  This paper presents an empirical study evaluating the performance of time series foundation models (TSFMs) over two datasets constituting noisy periodic time series.<n>Our findings demonstrate that while for time series with bounded periods, TSFMs can match or outperform the statistical approaches, their forecasting abilities deteriorate with longer periods, higher noise levels, lower sampling rates and more complex shapes of the time series.
  arXiv  Detail & Related papers  (2025-01-01T16:36:21Z)
• F-SE-LSTM: A Time Series Anomaly Detection Method with Frequency Domain Information [10.113418621891281]
  We propose a new time series anomaly detection method called F-SE-LSTM.<n>This method utilizes two sliding windows and fast Fourier transform (FFT) to construct a frequency matrix.<n>We show that F-SE-LSTM exhibits better discriminative ability than ordinary time domain and frequency domain data.
  arXiv  Detail & Related papers  (2024-12-03T14:36:24Z)
• Tuning Frequency Bias of State Space Models [48.60241978021799]
  State space models (SSMs) leverage linear, time-invariant (LTI) systems to learn sequences with long-range dependencies. We find that SSMs exhibit an implicit bias toward capturing low-frequency components more effectively than high-frequency ones.
  arXiv  Detail & Related papers  (2024-10-02T21:04:22Z)
• Frequency-domain MLPs are More Effective Learners in Time Series Forecasting [67.60443290781988]
  Time series forecasting has played the key role in different industrial domains, including finance, traffic, energy, and healthcare. Most-based forecasting methods suffer from the point-wise mappings and information bottleneck. We propose FreTS, a simple yet effective architecture built upon Frequency-domains for Time Series forecasting.
  arXiv  Detail & Related papers  (2023-11-10T17:05:13Z)
• Masked Frequency Modeling for Self-Supervised Visual Pre-Training [102.89756957704138]
  We present Masked Frequency Modeling (MFM), a unified frequency-domain-based approach for self-supervised pre-training of visual models. MFM first masks out a portion of frequency components of the input image and then predicts the missing frequencies on the frequency spectrum. For the first time, MFM demonstrates that, for both ViT and CNN, a simple non-Siamese framework can learn meaningful representations even using none of the following: (i) extra data, (ii) extra model, (iii) mask token.
  arXiv  Detail & Related papers  (2022-06-15T17:58:30Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.