# Time-Aware Prior Fitted Networks for Zero-Shot Forecasting with Exogenous Variables

Source: [Time-Aware Prior Fitted Networks for Zero-Shot Forecasting with Exogenous Variables](<https://www.sota2.com/research/paper/69ba31730636e5bf4afc0687>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-07-22T05:00:05.107Z

Authors: Andres Potapczynski, Ravi Kiran Selvam, Tatiana Konstantinova, Malcolm Wolff, Kin G. Olivares, Ruijun Ma, Michael W. Mahoney, Andrew Gordon Wilson, Boris N. Oreshkin, Dmitry Efimov

Published: 2026-03-16T18:33:23.000Z

## Abstract

In many time series forecasting settings, the target time series is accompanied by exogenous covariates, such as promotions and prices in retail demand; temperature in energy load; calendar and holiday indicators for traffic or sales; and grid load or fuel costs in electricity pricing. Ignoring these exogenous signals can substantially degrade forecasting accuracy, particularly when they drive spikes, discontinuities, or regime and phase changes in the target series. Most current time series foundation models (e.g., Chronos, Sundial, TimesFM, TimeMoE, TimeLLM, and LagLlama) ignore exogenous covariates and make forecasts solely from the numerical time series history, thereby limiting their performance. In this paper, we develop ApolloPFN, a prior-data fitted network (PFN) that is time-aware (unlike prior PFNs) and that natively incorporates exogenous covariates (unlike prior univariate forecasters). Our design introduces two major advances: (i) a synthetic data generation framework that injects realistic temporal patterns, structural changes, and exogenous dependencies into the PFN prior; and (ii) time-aware architectural modifications that embed inductive biases needed to exploit temporal context. We demonstrate that ApolloPFN outperforms existing baselines across several forecasting benchmarks with exogenous information, including M5, electric price forecasting, UCI Air Quality, and Solar Energy datasets.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Electricity Price Forecasting on NP horizon 24 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-on-np-horizon-24-test.md>): Electricity Price Forecasting; Electricity Price Forecasting NP horizon 24 (test); sCRPS: 0.038.

- [Electricity Price Forecasting on FR horizon 24 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-on-fr-horizon-24-test.md>): Electricity Price Forecasting; Electricity Price Forecasting (FR) - horizon 24 (test); sCRPS: 0.04.

- [Electricity Price Forecasting BE horizon 24 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-be-horizon-24-test.md>): Electricity Price Forecasting; Electricity Price Forecasting BE horizon 24 (test); sCRPS: 0.042.

- [Electricity Price Forecasting on PJM horizon 24 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-on-pjm-horizon-24-test.md>): Electricity Price Forecasting; Electricity Price Forecasting PJM horizon 24 (test); sCRPS: 0.04.

- [Electricity Price Forecasting on DE horizon 48 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-on-de-horizon-48-test.md>): Electricity Price Forecasting; Electricity Price Forecasting (DE) horizon 48 (test); sCRPS: 0.056.

- [Electricity Price Forecasting on NP horizon 48 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-on-np-horizon-48-test.md>): Electricity Price Forecasting; Electricity Price Forecasting (NP) horizon 48 (test); sCRPS: 0.053.

- [Electricity Price Forecasting BE horizon 48 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-be-horizon-48-test.md>): Electricity Price Forecasting; Electricity Price Forecasting BE horizon 48 (test); sCRPS: 0.058.

- [Electricity Price Forecasting on PJM horizon 48 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-on-pjm-horizon-48-test.md>): Electricity Price Forecasting; PJM horizon 48 (test); sCRPS: 0.057.

- [Time Series Forecasting on M5 Competition State Level](<https://www.sota2.com/research/sota/time-series-forecasting-on-m5-competition-state-level.md>): Time Series Forecasting; M5 Competition State Level; M5 Daily (Basic) RMSSE: 0.58.

- [Univariate Time Series Forecasting on M3 M-competition](<https://www.sota2.com/research/sota/univariate-time-series-forecasting-on-m3-m-competition.md>): Univariate Time Series Forecasting; M3 M-competition; sCRPS: 0.034.

- [Univariate Time Series Forecasting on M4 Daily M-competition](<https://www.sota2.com/research/sota/univariate-time-series-forecasting-on-m4-daily-m-competition.md>): Univariate Time Series Forecasting; M4 Daily M-competition; sCRPS: 0.023.

- [Univariate Time Series Forecasting on M4 Monthly M-competition](<https://www.sota2.com/research/sota/univariate-time-series-forecasting-on-m4-monthly-m-competition.md>): Univariate Time Series Forecasting; M4 Monthly M-competition; sCRPS: 0.092.

- [Univariate Time Series Forecasting on Tourism Monthly](<https://www.sota2.com/research/sota/univariate-time-series-forecasting-on-tourism-monthly.md>): Univariate Time Series Forecasting; Tourism Monthly; sCRPS: 0.084.

- [Electricity Price Forecasting on DE horizon 24 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-on-de-horizon-24-test.md>): Electricity Price Forecasting; Electricity Price Forecasting (DE) horizon 24 (test); sCRPS: 0.04.

- [Electricity Price Forecasting FR horizon 48 (test)](<https://www.sota2.com/research/sota/electricity-price-forecasting-fr-horizon-48-test.md>): Electricity Price Forecasting; Electricity Price Forecasting FR horizon 48 (test); sCRPS: 0.069.

- [Time Series Forecasting on M5 Competition Store Level](<https://www.sota2.com/research/sota/time-series-forecasting-on-m5-competition-store-level.md>): Time Series Forecasting; M5 Competition Store Level; M5 Daily Base RMSSE: 0.675.

- [Univariate Time Series Forecasting on M1 Monthly](<https://www.sota2.com/research/sota/univariate-time-series-forecasting-on-m1-monthly.md>): Univariate Time Series Forecasting; M1 Monthly; sCRPS: 0.152.

- [Univariate Time Series Forecasting on M3 Monthly M-competition](<https://www.sota2.com/research/sota/univariate-time-series-forecasting-on-m3-monthly-m-competition.md>): Univariate Time Series Forecasting; M3 Monthly M-competition; sCRPS: 0.094.

- [Univariate Time Series Forecasting on M4 Yearly M-competition](<https://www.sota2.com/research/sota/univariate-time-series-forecasting-on-m4-yearly-m-competition.md>): Univariate Time Series Forecasting; M4 Yearly M-competition; sCRPS: 0.113.

- [Univariate Time Series Forecasting on Tourism Yearly](<https://www.sota2.com/research/sota/univariate-time-series-forecasting-on-tourism-yearly.md>): Univariate Time Series Forecasting; Tourism Yearly; sCRPS: 0.137.

- [Univariate Time Series Forecasting on M1 Yearly M-competition](<https://www.sota2.com/research/sota/univariate-time-series-forecasting-on-m1-yearly-m-competition.md>): Univariate Time Series Forecasting; M1 Yearly M-competition; sCRPS: 0.142.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.15802>)

[PDF](<https://arxiv.org/pdf/2603.15802.pdf>)
