Xin-YuXiao/Awesome-AI-for-Astronomy

What it covers

This repository indexes machine learning research in astronomy, astrophysics, cosmology, heliophysics and planetary science. It organizes 402 papers and 66 software packages by topic and modality, with an emphasis on foundation models, language models, agents and benchmarks from 2021 onward, alongside established deep-learning methods. Each entry carries a date, venue, code or model links when public, and a one-sentence summary. Coverage spans surveys, images, spectra, time series, multimodal work, vision-language models, retrieval agents, benchmarks, datasets, and domain subfields from galaxies to gravitational waves. Readers are researchers and students who need to find prior work or reusable code, not a tutorial series.

Getting started

There is no installation step. The most direct entry point is the searchable web page, which filters all papers and packages by section, year, venue type, public code, model weights or data, and data modality. To browse without a browser filter, open the README contents table and jump to a section such as Foundation Models and Representation Learning. Papers are ordered newest first within each subsection, and hovering a title in the web view reveals its summary. For literature-specific use, the AstroAI Today page is a separate daily-facing view linked from the README header.

Constraints

• The list is an index, not a runnable package: it does not provide training code, weights or installation instructions of its own.
• Links point to third-party papers, repositories and datasets, each under its own license and maintenance status.
• Papers are labeled by arXiv month or publication year when no arXiv version exists; venue labels vary in how they were confirmed, so treat them as catalog metadata rather than formal citation records.
• Contributions must go through data/entries.yml and the build script rather than direct README edits; the entry format and checks are described in CONTRIBUTING.md.
• The list is released under CC0-1.0.

HotRepo's take

The repository was recorded only once, on the New list for Python on 2026-11-13, at 108 stars and 6 forks, with 7 stars gained during a roughly 90-minute window that day; no all-language daily rank is available. That single snapshot is enough to confirm early attention but not a trend. What stands out technically is the split between a YAML-backed data file and a generated README plus searchable site, which makes the 402-paper catalog filterable rather than a flat link dump; the site also reports code, weights and modality per entry. For a survey of AI in astronomy, that structure is more useful than the star count.