mlsanityData to 7 Oct 2026

TiRex-2: Generalizing TiRex to Multivariate Data and Streaming

arXiv 2607.01204 ↗First posted 1 Jul 2026cs.LGcited by 10

Not on the map. Why?

Abstract (arXiv)

We introduce TiRex-2, a recurrent xLSTM-based time series foundation model that generalizes the univariate TiRex to multivariate forecasting with both past and future covariates. Real-world forecasting is inherently sequential: observations arrive continuously, variables evolve jointly, and a subset of covariates is known ahead of time. Existing Transformer-based time series foundation models capture cross-variate dependencies but incur quadratic complexity in context length and require full-history recomputation as new observations arrive. TiRex-2 addresses these limitations through a memory-centric recurrent design that operates at constant per-patch cost under streaming. The model combines a bidirectional time mixer with an asymmetric grouped-attention variate mixer, enabling the integration of future-known covariates while preserving strict causality over target variables. To our knowledge, this is the first time series foundation model that achieves this combination of properties. To support scalable multivariate pretraining, we propose a synthetic coupling pipeline that composes diverse multivariate samples on the fly from large univariate corpora. Empirically, TiRex-2 achieves state-of-the-art zero-shot performance on GIFT-Eval and fev-bench, remains stable when streamed to arbitrary context lengths, and maintains constant inference cost per patch. The model uses 38.4M active parameters in univariate mode, with an additional 44.1M parameters activated for multivariate forecasting.

Key facts. Cited by 10 papers in the map (data through 7 Oct 2026). Builds on 38 of its 65 references that are in the map. First posted on arXiv 1 Jul 2026.

Builds on

38 of 65 references.

References · 38

  1. 2022A Time Series is Worth 64 Words: Long-term Forecasting with Transformerscited by 1,981
  2. 2023iTransformer: Inverted Transformers Are Effective for Time Series Forecastingcited by 906
  3. 2024Chronos: Learning the Language of Time Seriescited by 712
  4. 2023A decoder-only foundation model for time-series forecastingcited by 650
  5. 2024Unified Training of Universal Time Series Forecasting Transformerscited by 588
  6. 2023Large Language Models Are Zero-Shot Time Series Forecasterscited by 491
  7. 2024xLSTM: Extended Long Short-Term Memorycited by 320
  8. 2024TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variablescited by 241
  9. 2023TSMixer: An All-MLP Architecture for Time Series Forecastingcited by 232
  10. 2024GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluationcited by 188
  11. 2022PromptCast: A New Prompt-based Learning Paradigm for Time Series Forecastingcited by 170
  12. 2025Sundial: A Family of Highly Capable Time Series Foundation Modelscited by 168
26 more
  1. 2024Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Seriescited by 120
  2. 2025TiRex: Zero-Shot Forecasting Across Long and Short Horizons with Enhanced In-Context Learningcited by 107
  3. 2024UniTS: A Unified Multi-Task Time Series Modelcited by 93
  4. 2024Moirai-MoE: Empowering Time Series Foundation Models with Sparse Mixture of Expertscited by 88
  5. 2021Chaos as an interpretable benchmark for forecasting and data-driven modellingcited by 87
  6. 2024Timer-XL: Long-Context Transformers for Unified Time Series Forecastingcited by 83
  7. 2025Moirai 2.0: When Less Is More for Time Series Forecastingcited by 59
  8. 2025This Time is Different: An Observability Perspective on Time Series Foundation Modelscited by 57
  9. 2025fev-bench: A Realistic Benchmark for Time Series Forecastingcited by 52
  10. 2024Toto: Time Series Optimized Transformer for Observabilitycited by 44
  11. 2024In-Context Fine-Tuning for Time-Series Foundation Modelscited by 31
  12. 2026Timer-S1: A Billion-Scale Time Series Foundation Model with Serial Scalingcited by 24
  13. 2023A Comprehensive Survey of Deep Transfer Learning for Anomaly Detection in Industrial Time Series: Methods, Applications, and Directionscited by 22
  14. 2025ChronosX: Adapting Pretrained Time Series Models with Exogenous Variablescited by 21
  15. 2025Output Scaling: YingLong-Delayed Chain of Thought in a Large Pretrained Time Series Forecasting Modelcited by 21
  16. 2025TempoPFN: Synthetic Pre-training of Linear RNNs for Zero-shot Time Series Forecastingcited by 16
  17. 2025AdaPTS: Adapting Univariate Foundation Models to Probabilistic Multivariate Time Series Forecastingcited by 15
  18. 2024Bio-xLSTM: Generative modeling, representation and in-context learning of biological and chemical sequencescited by 13
  19. 2025Zero-Shot Time Series Forecasting with Covariates via In-Context Learningcited by 12
  20. 2026Reverso: Efficient Time Series Foundation Models for Zero-shot Forecastingcited by 11
  21. 2023How Does It Function? Characterizing Long-term Trends in Production Serverless Workloadscited by 10
  22. 2026Revisiting the Generic Transformer: Deconstructing a Strong Baseline for Time Series Foundation Modelscited by 10
  23. 2026CoRA: Boosting Time Series Foundation Models for Multivariate Forecasting through Correlation-aware Adaptercited by 6
  24. 2025CITRAS: Covariate-Informed Transformer for Time Series Forecastingcited by 5
  25. 2026Zero-shot Forecasting by Simulation Alonecited by 4
  26. 2026WaveMoE: A Wavelet-Enhanced Mixture-of-Experts Foundation Model for Time Series Forecastingcited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 10.

10 papers.

  1. 2026Into the ORBIT for Time Series: Training Regimes for Foundation Modelscited by 3
  2. 2026SANE: State Anomaly Neutralization for Stable Extreme-Context Delta-Rule Modelscited by 0
  3. 2026Distillation of Synthetic Data for Time Series Foundation Modelscited by 0
  4. 2026FlowTSFM: Turning Encoder Depth into Quantile Transportcited by 0
  5. 2026t0: A Time-Series Foundation Model for Forecasting with Contextcited by 0
  6. 2026Aurora-X: Built for Extreme Time Series Forecastingcited by 0
  7. 2026Self-Evolving Time-Series Forecasting Agents with Episodic Memory and Online Policy Learningcited by 0
  8. 2026OpenTSLM TeeMoE: A Unified Time-Series Language Model for Forecasting, Contextual Prediction, and Reasoningcited by 0
  9. 2026Pooling Helps, Learned Weighting Hurts In-Context: Decomposing Group Attentioncited by 0
  10. 2026Timer-M1: A Multivariate Time Series Foundation Model via Learning Primitivescited by 0