Research in practice
Open-source models, research implementations, and datasets.
Foundation model · 2026
A ~105M-parameter foundation model for zero-shot probabilistic time-series forecasting, released under the MIT license. Competitive with leading forecasting models at a fraction of the size of several larger entries.
*3rd in CRPS and 4th in MASE among 36 models in the GIFT-Eval pretrained category. Benchmark details ↗
Research code for Goal-Oriented Time-Series Forecasting: Foundation Framework Design, AAAI 2026.
View repository ↗This repository houses the code used for experiments in our paper accepted to AAAI 2024.
View repository ↗Simulator of experiments presented in the paper accepted to ACM SIGMETRICS 2023.
View repository ↗Simulator of experiments presented in the paper accepted to ACM SIGMETRICS 2022.
View repository ↗