How to use this guide¶
The Living Guide is built around five commitments, taken directly from the transition paper that introduced this resource:
- No claim of completeness. Sections do not attempt to list every relevant paper. This is not a limitation — it is the point.
- Curation through community interest. Sections exist because someone with expertise wrote them. Active subfields naturally attract more frequent updates.
- Annotation is required. Every recommended paper or cluster of papers comes with a short explanation of what it establishes and how it relates to adjacent work. A list of links without context is a search result, not a guide.
- Complement INSPIRE-HEP and arXiv, do not compete with them. Comprehensive bibliographic search already exists and works well. Finding papers is your job and the search engines' job. The Guide's job is telling you which few are worth reading first, and why. It provides what those tools do not and provides a structured context and explicit guidance on where to start.
- Sustainability by design. Sections are written as named, timestamped one-time contributions and remain stable until a new contribution updates them. A stale bibliography is merely incomplete, while a stale guide is misleading, so every section displays when it was last reviewed and automatically carries a notice once that is more than 12 months ago. You should never have to guess the vintage of a recommendation.
Scope¶
The Guide covers machine learning for particle physics: collider physics and phenomenology, formal and theoretical particle physics, lattice field theory, and neutrino physics.
Nuclear and heavy-ion physics, astroparticle physics, cosmology, astronomical data science, accelerator ML, and generic ML methodology with no particle-physics content are deliberately out of scope — see About for the reasoning, and Related resources for where to go instead.
What you'll find in each section¶
Each topical section follows a consistent structure:
| Element | Purpose |
|---|---|
| Overview | One to two paragraphs on the problem, why it matters in HEP, and how it connects to adjacent areas. |
| Recommended starting points | Three to six reviews, tutorials, or lecture notes, each with a one-sentence annotation. |
| Curated paper list | Foundational and representative papers, grouped thematically, each with a brief annotation. Not exhaustive by design. |
| Benchmarks, datasets, and software | Key references for reproducibility and comparison, where they exist. |
| Contributor & vintage | Who wrote the section, and when. |
What you will not find¶
- A complete bibliography, or any attempt at one. For that, use INSPIRE-HEP, arXiv, or the archived Living Review. Those tools are good at finding papers and this resource is not trying to be.
- Peer review. Curation here is editorial, not adjudicative — inclusion is not an endorsement and exclusion is not a judgment of quality.
- Real-time freshness. Sections are timestamped. If a section is older than you'd like, the right response is to contribute an update.
A note on cross-cutting work¶
Many important papers sit at the intersection of an application (e.g. fast calorimeter simulation) and a method (e.g. diffusion models). Where this is the case, the paper is annotated in the section where it is most useful as an entry point and cross-linked from the other. If you think a cross-link is missing, please open an issue or a pull request.