Releases·Model Release
Google Research announces TimesFM-3, a zero-shot foundation model for multivariate forecasting
Google Research announced TimesFM-3, a zero-shot foundation model for multivariate time-series forecasting. The announcement post appeared via research.google, shared on Reddit. Practitioners in forecasting may find a pretrained zero-shot alternative to task-specific model training.
95 engagement1 sourceTue, Sep 1, 2026, 12:56 AM
Entities
Related
Social·Community
TimesFM-3: Bringing LLM-Style Foundation Models to Time-Series Forecasting
Medium article examines TimesFM-3's LLM-style approach to time-series forecasting
Medium article examines TimesFM-3's LLM-style approach to time-series forecasting
0 engagement1 sourcerssWed, Sep 2, 2026, 07:35 AM
Products·launched
Ephemeris: API for ensemble forecasts from multiple open-source time series foundation models
Ephemeris is an inference provider exposing several open-source time series foundation models (Chronos2, Flowstate, TimesFM, and others) through one API. Users can select multiple models to produce ensemble forecasts, and can hand the API to an LLM agent to use if unfamiliar with TSFMs. It lowers the barrier to running these models for forecasting without downloading them locally.
Ephemeris is an inference provider exposing several open-source time series foundation models (Chronos2, Flowstate, TimesFM, and others) through one API. Users can select multiple models to produce ensemble forecasts, and can hand the API to an LLM agent to use if unfamiliar with TSFMs. It lowers the barrier to running these models for forecasting without downloading them locally.
1 engagement1 sourceredditFri, Sep 25, 2026, 10:04 AM
Releases·Model Release
$t_0$: Open-Weights Time-Series Foundation Models for Forecasting with Covariate Context
A new arXiv report introduces $t_0$, a family of open-weights time-series foundation models that condition forecasts on past covariates and known-future covariates without task-specific retraining. The first two releases, t0-alpha (102M) and t0-beta (256M), rank third on GIFT-Eval among zero-shot TSFMs and perform strongly on fev-bench and an independent Macrocosm evaluation.
A new arXiv report introduces $t_0$, a family of open-weights time-series foundation models that condition forecasts on past covariates and known-future covariates without task-specific retraining. The first two releases, t0-alpha (102M) and t0-beta (256M), rank third on GIFT-Eval among zero-shot TSFMs and perform strongly on fev-bench and an independent Macrocosm evaluation.
0 engagement1 sourcearxivMon, Sep 21, 2026, 01:22 PM
Developments·Paper
Latent Inference-Time Guidance of Time Series Foundation Models
An arXiv paper (2026-09-29) proposes Latent Inference-Time Guidance for Time Series Foundation Models, adaptively combining a pool of TSFM forecasts through a time-dependent latent space with independent components. The method carries identifiability and reconstruction guarantees, keeps foundation models off-the-shelf, and performs competitively with traditional ensembling across multi-domain, multi-frequency datasets.
An arXiv paper (2026-09-29) proposes Latent Inference-Time Guidance for Time Series Foundation Models, adaptively combining a pool of TSFM forecasts through a time-dependent latent space with independent components. The method carries identifiability and reconstruction guarantees, keeps foundation models off-the-shelf, and performs competitively with traditional ensembling across multi-domain, multi-frequency datasets.
0 engagement1 sourcearxivTue, Sep 29, 2026, 05:25 PM
Developments·Paper
Research paper explores zero-shot foundation models for multivariate time series anomaly detection
A new arXiv paper investigates whether univariate time series forecasting foundation models can be applied zero-shot to multivariate time series anomaly detection (MTSAD), a task traditionally requiring costly application-specific models. The study tests a univariate forecasting FM on MTSAD benchmarks, aiming to assess generalization without task-specific training. This could reduce the cost and complexity of anomaly detection in industrial and financial domains.
A new arXiv paper investigates whether univariate time series forecasting foundation models can be applied zero-shot to multivariate time series anomaly detection (MTSAD), a task traditionally requiring costly application-specific models. The study tests a univariate forecasting FM on MTSAD benchmarks, aiming to assess generalization without task-specific training. This could reduce the cost and complexity of anomaly detection in industrial and financial domains.
0 engagement1 sourcearxivTue, Jul 14, 2026, 07:32 AM