mlsanityData to 7 Oct 2026

Sundial: A Family of Highly Capable Time Series Foundation Models

arXiv 2502.00816 ↗First posted 2 Feb 2025cs.LGcited by 168

Not on the map. Why?

Abstract (arXiv)

We introduce Sundial, a family of native, flexible, and scalable time series foundation models. To predict the next-patch's distribution, we propose a TimeFlow Loss based on flow-matching, which facilitates native pre-training of Transformers on continuous-valued time series without discrete tokenization. Conditioned on arbitrary-length time series, our models are pre-trained without specifying any prior distribution and can generate multiple probable predictions, achieving more flexibility in representation learning than using parametric densities. Towards time series foundation models, we leverage minimal but crucial adaptations of Transformers and curate TimeBench with one trillion time points, comprising mostly real-world datasets and synthetic data. By mitigating mode collapse via TimeFlow Loss, we pre-train a family of Sundial models on TimeBench, which achieve unprecedented model capacity and generalization performance. In addition to excellent scalability, Sundial achieves state-of-the-art results on both point and probabilistic forecasting benchmarks with a just-in-time inference speed, i.e., making zero-shot predictions within a few milliseconds. We believe that Sundial's pioneering generative forecasting capability can improve model reliability in real-world decision-making. Code is available at: https://github.com/thuml/Sundial.

Key facts. Cited by 168 papers in the map (data through 7 Oct 2026). Builds on 48 of its 64 references that are in the map. First posted on arXiv 2 Feb 2025.

Builds on

48 of 64 references.

References · 48

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  3. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  4. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  5. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  6. 2022Scalable Diffusion Models with Transformerscited by 5,931
  7. 2022Flow Matching for Generative Modelingcited by 5,426
  8. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,581
  9. 2021Zero-Shot Text-to-Image Generationcited by 4,276
  10. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,094
  11. 2021On the Opportunities and Risks of Foundation Modelscited by 3,961
  12. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,747
36 more
  1. 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awarenesscited by 3,611
  2. 2023A Survey of Large Language Modelscited by 2,537
  3. 2022A Time Series is Worth 64 Words: Long-term Forecasting with Transformerscited by 1,981
  4. 2019Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelismcited by 1,829
  5. 2021Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecastingcited by 1,654
  6. 2018An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modelingcited by 1,447
  7. 2022TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysiscited by 1,030
  8. 2017DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networkscited by 1,004
  9. 2020On Layer Normalization in the Transformer Architecturecited by 906
  10. 2023iTransformer: Inverted Transformers Are Effective for Time Series Forecastingcited by 906
  11. 2023Improving and generalizing flow-based generative models with minibatch optimal transportcited by 835
  12. 2019Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecastingcited by 748
  13. 2019N-BEATS: Neural basis expansion analysis for interpretable time series forecastingcited by 744
  14. 2024Chronos: Learning the Language of Time Seriescited by 712
  15. 2023A decoder-only foundation model for time-series forecastingcited by 650
  16. 2024Autoregressive Image Generation without Vector Quantizationcited by 644
  17. 2024Unified Training of Universal Time Series Forecasting Transformerscited by 588
  18. 2021CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputationcited by 577
  19. 2022Efficiently Scaling Transformer Inferencecited by 529
  20. 2023Large Language Models Are Zero-Shot Time Series Forecasterscited by 491
  21. 2024Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Modelscited by 467
  22. 2024MOMENT: A Family of Open Time-series Foundation Modelscited by 453
  23. 2022Non-stationary Transformers: Exploring the Stationarity in Time Series Forecastingcited by 385
  24. 2023Long-term Forecasting with TiDE: Time-series Dense Encodercited by 327
  25. 2024Flow Matching Guide and Codecited by 291
  26. 2024Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Expertscited by 280
  27. 2024Foundation Models for Time Series Analysis: A Tutorial and Surveycited by 234
  28. 2018Adaptive Input Representations for Neural Language Modelingcited by 226
  29. 2017A Multi-Horizon Quantile Recurrent Forecastercited by 211
  30. 2024Timer: Generative Pre-trained Transformers Are Large Time Series Modelscited by 198
  31. 2024GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluationcited by 188
  32. 2023Koopa: Learning Non-stationary Time Series Dynamics with Koopman Predictorscited by 110
  33. 2024Timer-XL: Long-Context Transformers for Unified Time Series Forecastingcited by 83
  34. 2023Non-autoregressive Conditional Diffusion Models for Time Series Predictioncited by 77
  35. 2024Flow Matching with Gaussian Process Priors for Probabilistic Time Series Forecastingcited by 40
  36. 2024Scaling Law for Time Series Forecastingcited by 31

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 61; 2026: 107.

168 papers.

Sort:

By area:

  1. 2025Chronos-2: From Univariate to Universal Forecastingcited by 158
  2. 2025TiRex: Zero-Shot Forecasting Across Long and Short Horizons with Enhanced In-Context Learningcited by 107
  3. 2025Moirai 2.0: When Less Is More for Time Series Forecastingcited by 59
  4. 2025fev-bench: A Realistic Benchmark for Time Series Forecastingcited by 52
  5. 2025TS-RAG: Retrieval-Augmented Generation based Time Series Foundation Models are Stronger Zero-Shot Forecastercited by 36
  6. 2025Aurora: Towards Universal Generative Multimodal Time Series Forecastingcited by 28
  7. 2025Kronos: A Foundation Model for the Language of Financial Marketscited by 25
  8. 2026It's TIME: Towards the Next Generation of Time Series Forecasting Benchmarkscited by 25
  9. 2026Timer-S1: A Billion-Scale Time Series Foundation Model with Serial Scalingcited by 24
  10. 2025MIRA: Medical Time Series Foundation Model for Real-World Health Datacited by 21
  11. 2025Context parroting: A simple but tough-to-beat baseline for foundation models in scientific machine learningcited by 14
  12. 2025TimeMosaic: Temporal Heterogeneity Guided Time Series Forecasting via Adaptive Granularity Patch and Segment-wise Decodingcited by 12
38 more
  1. 2025AlphaCast: A Human Wisdom-LLM Intelligence Co-Reasoning Framework for Interactive Time Series Forecastingcited by 11
  2. 2026Reverso: Efficient Time Series Foundation Models for Zero-shot Forecastingcited by 11
  3. 2025TS-Reasoner: Aligning Time Series Foundation Models with LLM Reasoningcited by 10
  4. 2025Rethinking Evaluation in the Era of Time Series Foundation Models: (Un)known Information Leakage Challengescited by 10
  5. 2026Revisiting the Generic Transformer: Deconstructing a Strong Baseline for Time Series Foundation Modelscited by 10
  6. 2026TiRex-2: Generalizing TiRex to Multivariate Data and Streamingcited by 10
  7. 2025Towards Foundation Models for Zero-Shot Time Series Anomaly Detection: Leveraging Synthetic Data and Relative Context Discrepancycited by 9
  8. 2025Pre-trained Forecasting Models: Strong Zero-Shot Feature Extractors for Time Series Classificationcited by 9
  9. 2026The Forecast After the Forecast: A Post-Processing Shift in Time Seriescited by 9
  10. 2025TimeCopilotcited by 8
  11. 2025Kairos: Toward Adaptive and Parameter-Efficient Time Series Foundation Modelscited by 8
  12. 2026Position: Beyond Model-Centric Prediction -- Agentic Time Series Forecastingcited by 8
  13. 2025FlowState: Sampling-Rate-Equivariant Time-Series Forecastingcited by 7
  14. 2025Xihe: Scalable Zero-Shot Time Series Learner Via Hierarchical Interleaved Block Attentioncited by 7
  15. 2025Synapse: Adaptive Arbitration of Complementary Expertise in Time Series Foundational Modelscited by 7
  16. 2026Falcon-X: A Time Series Foundation Model for Heterogeneous Multivariate Modelingcited by 7
  17. 2025Foundation models for time series forecasting: Application in conformal predictioncited by 6
  18. 2025STAR: Boosting Time Series Foundation Models for Anomaly Detection through State-aware Adaptercited by 6
  19. 2025SEMPO: Lightweight Foundation Models for Time Series Forecastingcited by 6
  20. 2025Predicting the Future by Retrieving the Pastcited by 6
  21. 2025Byte Pair Encoding for Efficient Time Series Forecastingcited by 5
  22. 2025Fidel-TS: A High-Fidelity Multimodal Benchmark for Time Series Forecastingcited by 5
  23. 2025CoRA: Covariate-Aware Adaptation of Time Series Foundation Modelscited by 5
  24. 2025The Seismic Wavefield Common Task Frameworkcited by 5
  25. 2026TimeSage-MT: A Multi-Turn Benchmark for Evaluating Agentic Time Series Reasoningcited by 5
  26. 2026Learning the Context of Errors: Black-Box Online Adaptation of Time Series Foundation Modelscited by 5
  27. 2025GTM: A General Time-series Model for Enhanced Representation Learning of Time-Series Datacited by 4
  28. 2025Exploring Accuracy Law for Deep Time Series Forecasters: An Empirical Studycited by 4
  29. 2026Unlocking the Value of Text: Event-Driven Reasoning and Multi-Level Alignment for Time Series Forecastingcited by 4
  30. 2026Baguan-TS: A Sequence-Native In-Context Learning Model for Time Series Forecasting with Covariatescited by 4
  31. 2026TempusBench: An Evaluation Framework for Time-Series Forecastingcited by 4
  32. 2026CTF4Nuclear: Common Task Framework for Nuclear Fission and Fusion Modelscited by 4
  33. 2025SeisMoLLM: Advancing Seismic Monitoring via Cross-modal Transfer with Pre-trained Large Language Modelcited by 3
  34. 2025One-Embedding-Fits-All: Efficient Zero-Shot Time Series Forecasting by a Model Zoocited by 3
  35. 2025CarbonX: An Open-Source Tool for Computational Decarbonization Using Time Series Foundation Modelscited by 3
  36. 2025The Few Govern the Many:Unveiling Few-Layer Dominance for Time Series Modelscited by 3
  37. 2025FLAME: Flow Enhanced Legendre Memory Models for General Time Series Forecastingcited by 3
  38. 2025TS-Arena -- A Live Forecast Pre-Registration Platformcited by 3

50 most cited of 168 All 168 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Pythia: Toward Foundation World Models for Multimodal Time Series
  2. 2025WaveStitch: Flexible and Fast Conditional Time Series Generation with Diffusion Models
  3. 2026ReDiTT: Retrieval Augmented Conditional Diffusion Transformers for Asynchronous Time Series
  4. 2026Towards a Unified Generative Model for Scarce Time Series with Domain Experts
  5. 2025Conversational Time Series Foundation Models: Towards Explainable and Effective Forecasting
  6. 2025Bridging the Last Mile of Prediction: Enhancing Time Series Forecasting with Conditional Guided Flow Matching
  7. 2025Time Series Generation Under Data Scarcity: A Unified Generative Modeling Approach
  8. 2024TimeDiT: General-purpose Diffusion Transformers for Time Series Foundation Model