Position: Universal Time Series Foundation Models Rest on a Category Error
Not on the map. Why?
Abstract (arXiv)
This position paper argues that the pursuit of "Universal Foundation Models for Time Series" rests on a fundamental category error, mistaking a structural Container for a semantic Modality. We contend that because time series hold incompatible generative processes (e.g., finance vs. fluid dynamics), monolithic models degenerate into expensive "Generic Filters" that fail to generalize under distributional drift. To address this, we introduce the "Autoregressive Blindness Bound," a theoretical limit proving that history-only models cannot predict intervention-driven regime shifts. We advocate replacing universality with a Causal Control Agent paradigm, where an agent leverages external context to orchestrate a hierarchy of specialized solvers, from frozen domain experts to lightweight Just-in-Time adaptors. We conclude by calling for a shift in benchmarks from "Zero-Shot Accuracy" to "Drift Adaptation Speed" to prioritize robust, control-theoretic systems.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 57 of its 78 references that are in the map. First posted on arXiv 5 Feb 2026.
Builds on
57 of 78 references.
References · 57
- 2017Attention Is All You Needcited by 59,932
- 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
- 2023GPT-4 Technical Reportcited by 18,132
- 2020Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecastingcited by 2,141
- 2021Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecastingcited by 1,654
- 2022Are Transformers Effective for Time Series Forecasting?cited by 1,400
- 2022FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecastingcited by 1,200
- 2024Chronos: Learning the Language of Time Seriescited by 712
- 2023Time-LLM: Time Series Forecasting by Reprogramming Large Language Modelscited by 680
- 2023A decoder-only foundation model for time-series forecastingcited by 650
- 2023One Fits All:Power General Time Series Analysis by Pretrained LMcited by 615
- 2024Unified Training of Universal Time Series Forecasting Transformerscited by 588
38 more
- 2023Large Language Models Are Zero-Shot Time Series Forecasterscited by 491
- 2024MOMENT: A Family of Open Time-series Foundation Modelscited by 453
- 2022Non-stationary Transformers: Exploring the Stationarity in Time Series Forecastingcited by 385
- 2024TimeMixer: Decomposable Multiscale Mixing for Time Series Forecastingcited by 333
- 2023Long-term Forecasting with TiDE: Time-series Dense Encodercited by 327
- 2024Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Expertscited by 280
- 2024Foundation Models for Time Series Analysis: A Tutorial and Surveycited by 234
- 2023Neural General Circulation Models for Weather and Climatecited by 232
- 2024GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluationcited by 188
- 2025Sundial: A Family of Highly Capable Time Series Foundation Modelscited by 168
- 2025Chronos-2: From Univariate to Universal Forecastingcited by 158
- 2023Stabilizing Transformer Training by Preventing Attention Entropy Collapsecited by 154
- 2023TimeGPT-1cited by 149
- 2023UniTime: A Language-Empowered Unified Model for Cross-Domain Time Series Forecastingcited by 139
- 2024From News to Forecast: Integrating Event Analysis in LLM-Based Time Series Forecasting with Reflectioncited by 93
- 2024Context is Key: A Benchmark for Forecasting with Essential Textual Informationcited by 78
- 2019Shape and Time Distortion Loss for Training Deep Time Series Forecasting Modelscited by 66
- 2024AutoTimes: Autoregressive Time Series Forecasters via Large Language Modelscited by 63
- 2025Moirai 2.0: When Less Is More for Time Series Forecastingcited by 59
- 2025fev-bench: A Realistic Benchmark for Time Series Forecastingcited by 52
- 2024Position: What Can Large Language Models Tell Us about Time Series Analysiscited by 47
- 2025DBLoss: Decomposition-based Loss Function for Time Series Forecastingcited by 44
- 2023CUTS: Neural Causal Discovery from Irregular Time-Series Datacited by 32
- 2025Battling the Non-stationarity in Time Series Forecasting via Test-time Adaptationcited by 32
- 2024Towards Neural Scaling Laws for Time Series Foundation Modelscited by 30
- 2024Learning Universal Predictorscited by 29
- 2023Self-Interpretable Time Series Prediction with Counterfactual Explanationscited by 24
- 2025TimeSeriesScientist: A General-Purpose AI Agent for Time Series Analysiscited by 18
- 2023When Model Meets New Normals: Test-time Adaptation for Unsupervised Time-series Anomaly Detectioncited by 17
- 2024Intervention-Aware Forecasting: Breaking Historical Limits from a System Perspectivecited by 16
- 2025One Surrogate to Fool Them All: Universal, Transferable, and Targeted Adversarial Attacks with CLIPcited by 15
- 2025TimeSeriesGym: A Scalable Benchmark for (Time Series) Machine Learning Engineering Agentscited by 14
- 2024Online Test-Time Adaptation of Spatial-Temporal Traffic Flow Forecastingcited by 11
- 2023Counterfactual Explanations for Time Series Forecastingcited by 10
- 2025Rethinking Evaluation in the Era of Time Series Foundation Models: (Un)known Information Leakage Challengescited by 10
- 2025CLIP-Guided Backdoor Defense through Entropy-Based Poisoned Dataset Separationcited by 9
- 2025From Samples to Scenarios: A New Paradigm for Probabilistic Forecastingcited by 8
- 2025Breaking the Stealth-Potency Trade-off in Clean-Image Backdoors with Generative Trigger Optimizationcited by 7
50 most cited of 57 All 57 (JSON)
Built on by
Citing papers per year: 2026: 1.
1 paper.