Skip to main content
Search

# Time-Series Foundation Model

7 related articles

Subscribe to updates
  1. arXiv

    Power-system dynamic simulation gets a fast stand-in: fine-tuned Chronos-2 beats a traditional AI model on error and runs 563 times faster

    The pretrained time-series foundation model Chronos-2 was tested as a surrogate for power-system dynamic simulation. On 1,800 load-disturbance samples from an IEEE 9-bus system, retrieval-augmented in-context learning beat random reference selection, and fine-tuned Chronos-2 had lower errors than a traditional spatiotemporal graph neural network baseline on the 178 held-out samples, with power-balance regularization improving every reported metric except voltage-angle MAE. Its 35.5 ms inference was 563 times faster than the 20 s ANDES simulation.

  2. arXiv

    RWKV fades on long horizons: alternating in attention blocks lets a 36M model match a 1B forecaster on GIFT long-horizon MASE

    HART alternates Transformer and RWKV blocks so prediction positions can retrieve history directly, beating pure RWKV at medium and long horizons on GIFT-Eval and TIME-Benchmark and matching Toto-2.0-1B's long-horizon MASE on GIFT with 36M parameters.

  3. Mistral Blog

    Mistral's 209M time-series model tops three public forecasting benchmarks, and it runs a one-day forecast on a CPU in 140 ms

    Mistral released Timestral, its first family of time-series foundation models, whose largest version at 209M parameters ranks first on the public GIFT-Eval, TIME and fev-bench forecasting benchmarks while using two-thirds the parameters of the previous state of the art. It produces probabilistic multivariate forecasts with past-only and known-future covariates, spans four sizes from 5M to 209M parameters, and needs no GPU: on an 8-thread server CPU, a one-day-ahead forecast from three weeks of 15-minute history takes 10 ms at 5M and 140 ms at 209M.

  4. arXiv

    Calendar encoding is not a detail: swapping it shifts holiday error by up to 83% in one model

    Time series foundation models now accept covariates, yet how calendar features are encoded has been treated as an implementation detail. Across four models and nine datasets shaped by weekends and holidays, seven calendar encodings were compared in a zero-shot rolling evaluation: encoding choice changed accuracy in every model, with within-model MASE gaps up to 9.9% on non-holidays and 83% on holidays, and the best encoding differed by model. SHAP attribution locates the gains around holidays, where signals cannot be inferred from target history.

  5. arXiv

    Time-series foundation models can skip whole Transformer blocks: A100 latency drops 11% to 13% at moderate sparsity, with dataset-level variation

    Two pretrained time-series foundation models skip whole Transformer blocks at inference, leaving mean forecasting quality nearly unchanged across 1,088 held-out windows while cutting median A100 batch-1 latency by 11.2% and 13.4% at moderate sparsity.

  6. arXiv

    A unified scaling law predicts held-out Toto 2.0 capacity curves at 1.09% and 1.50% MAPE, and activation interventions show frozen time series foundation models reuse historical rules across queries

    Using 18,768 experimental cells from 21 checkpoints on 23 dataset-frequency tasks across six domains, the work fits a five-parameter unified scaling law linking capacity, input length, and forecast horizon, and predicts the fully held-out Toto 2.0 series (about 4M-2.5B parameters) with MAPEs of 1.09% at input length 2048 and 1.50% at 4096 on horizon-averaged capacity curves; it also proposes a unified theory of time series learning in which Gaussian regression, matched-history comparisons, parameter exchanges, and activation interventions support the account that full-shot models accumulate rule information in weights while frozen time series foundation models extract and reuse historical rules through activations.

  7. bioRxiv

    Time-series foundation modeling enables accurate lake ecosystem forecasting

    Using up to 30 years of monthly monitoring records from two ecologically contrasting Japanese lakes, the deep-stratified Lake Biwa and the shallow nutrient-rich Lake Kasumigaura, this study benchmarked two time-series foundation models, the Transformer-based Chronos-T5 and the probabilistic Lag-Llama, against 15 statistical (AR, ARIMA, SARIMA, Prophet), machine learning (Random Forest, XGBoost, KNN, SVR), and deep learning (LSTM, CNN, TCN, and SSA-hybrid) approaches, finding that fine-tuned Chronos-T5 ranked first across all six monitoring sites with R2 > 0.80-1.00 for dissolved oxygen, water temperature, pH, and nutrients, while chlorophyll-a remained below R2 0.

You’re all caught up