Qian-Rui Lee — Academic Homepage

Academic website: https://toelul.github.io/

Source repository for my academic homepage. The site presents my research in theoretical and computational physics, quantum many-body systems, machine learning for physics, geometric and mathematical physics, and scientific computing.

Academic profiles: Google Scholar · ORCID · GitHub

Research focus

My work centers on using physical and mathematical structure—symmetry, geometry, locality, constraints, and scaling—to make difficult problems more interpretable and computationally tractable. Current themes include:

• quantum many-body physics, critical phenomena, and quantum geometry;
• generative modeling and Flow Matching for physical systems;
• equivariant learning and lattice gauge structure;
• geometric formulations of field theory and gravity;
• GPU-accelerated simulation and scientific-computing tools.

Site contents

Section	Purpose
Research	Research directions with links to related papers, notes, and software.
Publications	Peer-reviewed articles, preprints, and technical notes with stable landing pages.
Notes	Technical, pedagogical, and working lecture notes kept distinct from journal publications.
Software	Selected research software organized by scientific purpose.
CV	Web-based academic CV generated from the same structured content.

Scholarly publishing and indexing

Publication pages are designed as stable scholarly landing pages rather than simple file listings. Each record can expose:

• explicit publication status (journal, preprint, or note);
• DOI, arXiv, Zenodo, and source-code links where applicable;
• Highwire-style citation_* metadata for scholarly crawlers;
• ScholarlyArticle structured data for general search engines;
• a directly visible abstract without client-side rendering;
• same-site full-text PDF URLs for selected works when appropriate.

For the ADM-to-Ashtekar technical note, the canonical PDF remains in ToelUl/adm-to-ashtekar-notes. The Pages workflow stages that file into the deployed scholarly landing-page directory, avoiding a second manually maintained source copy while preserving a same-site PDF URL for indexing.

Architecture

The site is deliberately lightweight and static:

• Jekyll for templating and content collections;
• GitHub Pages for hosting;
• GitHub Actions for validation, build, PDF staging, and deployment;
• YAML front matter as the publication metadata source of truth;
• static HTML navigation with no JavaScript dependency for discovering scholarly content.

The main content model is organized around _publications/, _data/, and reusable layouts/includes. Research pages, publication indexes, notes, the web CV, citation metadata, and the sitemap are generated from the same structured records.

Maintaining publication records

Each scholarly work is represented by one file in _publications/. Publication status is explicit rather than inferred:

• journal — peer-reviewed journal article;
• preprint — public preprint not represented as a journal publication;
• note — technical or pedagogical note.

Use citation_date (YYYY/MM/DD when known) for chronological ordering. An arXiv record should provide both arxiv_id and arxiv_url; code_url and release_url add separate reproducibility links. Set selected: true to show a record on the home page and reserve featured: true for at most one current work, which appears in the home-page preprint callout.

Before merging publication changes, run:

python scripts/validate_publications.py

The same check runs automatically in GitHub Actions. The validator checks required bibliographic fields, publication status, Scholar-facing metadata, and the configuration of deployment-managed PDFs.

Repository structure

_data/          Academic profile and software metadata
_includes/      Reusable site components
_layouts/       Page and publication layouts
_publications/  Structured scholarly records
assets/         Site styles and static assets
scripts/        Metadata validation utilities
.github/        Validation and GitHub Pages workflows

The implementation favors long-term maintainability, explicit scholarly metadata, and crawler-friendly static pages over dynamic widgets or presentation-heavy dependencies.