![[2510.06200] StarEmbed: Benchmarking Time Series Foundation Models on Astronomical Observations of Variable Stars](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2510.06200] StarEmbed: Benchmarking Time Series Foundation Models on Astronomical Observations of Variable Stars
Summary
The paper introduces StarEmbed, a benchmark for evaluating time series foundation models on astronomical observations of variable stars, demonstrating their potential in outperforming traditional methods.
Why It Matters
This research is significant as it addresses the gap in applying advanced machine learning models to astronomical data, which traditionally lacks representation in training datasets. By establishing a benchmark, it encourages the adoption of foundation models in time-domain astronomy, potentially transforming data analysis practices in the field.
Key Takeaways
- StarEmbed is the first benchmark for time series foundation models in astronomy.
- The study evaluates TSFMs on tasks like clustering, classification, and source detection.
- Results show that TSFMs can outperform traditional astrophysics-specific methods.
- Chronos models excel in generalizing to new astronomical data.
- The research advocates for a shift towards generic models for analyzing large astronomical datasets.
Astrophysics > Solar and Stellar Astrophysics arXiv:2510.06200 (astro-ph) [Submitted on 7 Oct 2025 (v1), last revised 18 Feb 2026 (this version, v3)] Title:StarEmbed: Benchmarking Time Series Foundation Models on Astronomical Observations of Variable Stars Authors:Weijian Li, Hong-Yu Chen, Nabeel Rehemtulla, Ved G. Shah, Dennis Wu, Dongho Kim, Qinjie Lin, Adam A. Miller, Han Liu View a PDF of the paper titled StarEmbed: Benchmarking Time Series Foundation Models on Astronomical Observations of Variable Stars, by Weijian Li and 8 other authors View PDF HTML (experimental) Abstract:Time series foundation models (TSFMs) are increasingly being adopted as highly-capable general-purpose time series representation learners. Although their training corpora are vast, they exclude astronomical time series data. Observations of stars produce peta-scale time series with unique challenges including irregular sampling and heteroskedasticity. We introduce StarEmbed, the first public benchmark for rigorous and standardized evaluation of state-of-the-art TSFMs on stellar time series observations (``light curves''). We benchmark on three scientifically-motivated downstream tasks: unsupervised clustering, supervised classification, and out-of-distribution source detection. StarEmbed integrates a catalog of expert-vetted labels with multi-variate light curves from the Zwicky Transient Facility, yielding ~40k hand-labeled light curves spread across seven astrophysical classes. We evaluate the ...
