Profile data synchronization

Run the following command from the repository root:

python3 scripts/update_profile_data.py

The script updates _data/profile.json from:

• Haodong Duan’s Google Scholar profile: all paginated publications, citation counts, and citation rank.
• Every Scholar publication detail page: full author list, publication date, journal/conference/book, volume, issue, pages, publisher, institution, report/patent metadata, abstract/description, and source paper URL when Scholar provides them. Unknown bibliographic fields are preserved instead of discarded.
• GitHub: current Star, Fork, watcher, language, and repository metadata for the configured projects and publication repositories.
• _publications/*.md: existing paper/code/project links and legacy contribution markers.
• scripts/profile_overrides.json: curated topics, exact paper-verified author contributions, one-line summaries, repository mappings, thumbnails, project list, and Selected Works policy.
• resume/resume.tex: explicit Corresponding Author and Project Lead markers for Haodong, synchronized from the maintained Overleaf CV.

On the first run, the script opens each Scholar publication detail page to obtain complete authors and bibliographic metadata. Later runs reuse those details and normally update the full Scholar publication index, citation counts, and GitHub statistics. Use --refresh-details to re-crawl all detail pages when their metadata has changed:

python3 scripts/update_profile_data.py --refresh-details

The generated sync object records how many profile rows and unique Scholar IDs were seen, how many detail pages were fetched or failed, and the author/metadata coverage. The Publications page surfaces those coverage counts so an incomplete refresh is visible instead of silently passing.

GitHub’s unauthenticated API is rate-limited. The script automatically falls back to public repository HTML for Star/Fork counts. Supplying a token gives richer metadata:

GITHUB_TOKEN=... python3 scripts/update_profile_data.py

Google Scholar does not provide a reliable machine-readable source for equal-contribution, corresponding-author, or project-lead roles. Keep paper-verified annotations in scripts/profile_overrides.json; for an overridden paper the role map is treated as the exact curated source and is displayed with colored dots instead of legacy *, †, or ‡ symbols. The synchronizer also imports Haodong’s explicit \\dag/\\ddag markers from resume/resume.tex. When no author has an explicit corresponding-author mark, the project convention assigns that role to the final author and records the fallback in corresponding_fallback_author. Strictly alphabetical technical-report lists and institution-authored reports are automatically left entirely unannotated; use "author_order": "alphabetical" or "suppress_author_roles": true for ambiguous cases.

Per-paper metadata can be updated ahead of Scholar with venue, venue_display, venue_short, and year fields in scripts/profile_overrides.json. Prefer an official proceedings page, conference accepted-paper list, or publisher record when applying these overrides.

Selected Works is generated using three configurable rules:

1. First/equal/corresponding-author or project-lead papers with at least lead_min_citations citations.
2. Papers in the top top_citation_rank by citations with no more than top_cited_max_authors authors.
3. Per-paper force_include and force_exclude overrides.

The shared homepage/All News timeline lives in _data/news.json. Add a dated item there, mark only the most important items with "featured": true, and the homepage will remain a concise subset while /news/ shows the complete timeline.

Writing a blog post

Add a Markdown (.md) or HTML (.html) file under _blog/. Both formats use the same article layout, and automatically appear in the homepage’s Blog section (latest three), the Blog navigation, and the /blog/ archive, newest first. No separate index entry or build plugin is needed. The original template posts in _posts/ are excluded from this Blog index.

For example, _blog/my-first-note.md becomes /blog/my-first-note/:

---
title: "My first note"
date: 2026-11-03
lang: zh-CN
description: "A short search and social preview description."
cover: /images/blog/hello-world.svg
cover_alt: "Describe the cover image."
---

## A section heading

Write the article in **Markdown**.

Only title and date are required; date uses YYYY-MM-DD. Cover and language are optional. Cards display only the cover and title. Article pages display the title, optional cover, and the Markdown or HTML body. An empty body creates a cover-only post, as in Hello World. Use description for a search/social summary and permalink to override the URL. Use published: false to withhold an article from the normal build and index.

For an HTML article, use the same metadata header in a .html file, followed by an HTML body fragment:

---
title: "An HTML note"
date: 2026-11-03
description: "An article with custom HTML content."
---

<h2>A section heading</h2>
<p>Write the article using <strong>HTML</strong>.</p>

Do not add a second <html>, <head>, or <body> wrapper; the shared layout provides those. Markdown can also contain inline HTML. The article stylesheet covers headings, code, quotations, lists, images, and tables at desktop and mobile widths.

_data/blogs.json remains available for optional external articles. Each entry accepts date, title, url, and an optional cover (other metadata is retained but not displayed on cards). External links and local articles share the same date-sorted index. An empty index hides Blog from the homepage and navigation.