Skip to content

About

Official website of New Ground AI, an AI research collective.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

New Ground AI website

GitHub Pages site for New Ground AI: Advanced · Analytical · Adaptive · Adventurous · Aspiring.

Shared look and feel lives in assets/css/site.css, assets/js/site.js, and _includes/. Visual tokens are documented in brand-spec.md.

Content model

Nav: Home · People · Publication

Projects:            on-going `_projects/*.md`  +  published `projects/<slug>/index.html`
People:              `_people/*.md`  →  Home / People cards  →  `/people/<slug>/`  →  optional `website`
  • On-going projects use the shared Markdown layout and URLs /projects/<slug>/.
  • Published projects are free HTML. YAML is only for directories (Projects, Publication, homepage ticker).
  • People intros are Markdown in _people/. Portraits always open that page. A website field is linked from the intro, not from the card.
  • Publication and the homepage ticker list pages with status: published and listed: true, newest date first.

Update each section

Most page introductions are a short lead on that HTML page. Lists under them are generated from YAML — edit the source project or member, not the listing page.

Home — index.html

What you see Where to edit
Eyebrow, New Ground AI title, five-A line .hero__copy
Overview paragraph .hero__overview p
Latest Publication ticker Automatic: newest 5 published pages. Cards show year, title, venue (no authors).
Research themes _data/research_themes.yml
People sentences .people-intro

To change ticker copy, edit the published project’s YAML (title, year, venue, date). To hide a paper from Home and Publication, set listed: false.

People — _people/

Everyone has a community intro at /people/<slug>/, generated from _people/<slug>.md. Home and People portraits always open that page. A personal homepage — GitHub Pages or a folder on this site — is only a Personal site link on the intro. Do not list papers on the intro; keep it to portrait, facts, links, and a short bio.

What you see Where to edit
People page title and lead people/index.html
Honeycomb Every _people/*.md with listed: true
Community intro _people/<slug>.md

Every member starts the same way. Create _people/<slug>.md (for example _people/ye-yuan.md). Put a photo at assets/images/people/<slug>.jpg when you have one. Then choose A or B below for the personal site.

Field Required What it does
title yes Name on cards and the intro
role yes Role line (Ph.D. Candidate, …)
institution yes School / community line
focus yes Short research focus on the card
listed yes true to show in the People honeycomb
portrait no Path such as /assets/images/people/ye-yuan.jpg
email no Mailto link on the intro
website no Personal site — see A and B
scholar / github / cv no Extra links on the intro
Markdown body no Short biography under the facts

A. They already have a GitHub (or other) homepage

Do not create a folder under people/. Only the Markdown file. Set website to the live URL.

title: Ye Yuan
role: Ph.D. Candidate
institution: McGill University · Mila
focus: Generative AI · BBO · Agents
email: ye.yuan3@mail.mcgill.ca
portrait: /assets/images/people/ye-yuan.jpg
website: https://stevenyuan666.github.io/
scholar: https://scholar.google.com/citations?user=lemEc74AAAAJ
github: https://github.com/StevenYuan666
cv: https://stevenyuan666.github.io/assets/pdf/CV_Ye_YUAN_2026.pdf
listed: true
Click portrait  →  /people/ye-yuan/                 (this community’s intro)
Personal site   →  https://stevenyuan666.github.io/ (their GitHub Pages)

Add scholar, github, or cv the same way if they have them. Omit any they do not have.

B. They will design a homepage in this repo

Still create _people/<slug>.md. Then add a separate HTML site under people/<slug>/site/ so it does not collide with the intro at /people/<slug>/.

  1. Copy people/member-template/ to people/<slug>/site/.
  2. In that folder’s index.html, set:
layout: null
nav: people
permalink: /people/<slug>/site/
  1. Design the page freely (own CSS, JS, layout). Shared community chrome is optional.
  2. In _people/<slug>.md, set website to that path, not an https:// URL:
title: Example Name
role: Ph.D. Student
institution: McGill University · Mila
focus: Agents
portrait: /assets/images/people/example-name.jpg
website: /people/example-name/site/
listed: true
homepage: false
Click portrait  →  /people/example-name/      (community intro)
Personal site   →  /people/example-name/site/ (their HTML in this repo)

Do not use permalink: /people/<slug>/ on the custom HTML. That address is reserved for the Markdown intro.

C. No personal homepage yet

Omit website. Portraits still open the community intro; that page has no Personal site link. Add website later using A or B.

To add a member: _people/<slug>.md + optional portrait + A, B, or C.

Research themes — homepage

The six homepage themes live in _data/research_themes.yml. Each item needs id, title, lead, and summary. They are not linked to separate pages.

Publication — publication/index.html

What you see Where to edit
Directory title and lead .page-hero
Paper cards Automatic from published YAML. Cards show year, title, venue, keywords (same format as Home).

Papers split into Conferences & Journals & Workshops and arXiv Preprint. A venue containing arXiv goes to preprints; set publication_group: venue or publication_group: preprint to override.

There is no separate publication database. Add or edit projects/<slug>/index.html instead.

Projects — projects/index.html

What you see Where to edit
Directory title and lead .page-hero
Filters and cards Automatic from _projects/ plus published HTML

Search matches title, summary, area, keywords, members, authors, and venue.

Shared chrome

What you see Where to edit
Header wordmark _includes/site-mark.html, logo at assets/images/brand/
GitHub / Hugging Face _includes/site-presence.html
Nav labels and order _includes/site-nav.html
Footer line _includes/site-footer.html
Browser tab icon _includes/head-icons.html and assets/images/brand/favicon.png (transparent PNG)

Site-wide five-A wording also appears in _config.yml (description) and the footer.

Add an on-going project

Copy _projects/ongoing-project-template.md to _projects/your-slug.md. Keep the Overview, Motivation, and Goals headings.

title: Project title
status: ongoing
research_area: BBO
research_area_slug: bbo
keywords:
  - Keyword
members:
  - Member name
summary: Short introduction used on cards.

The page is generated at /projects/your-slug/.

Add a published project

Create projects/your-slug/index.html. projects/project-template/ is a starting point (listed: false, so it stays out of catalogs).

Directories only read the YAML. The HTML body can be anything: custom CSS, Canvas, WebGL, video, or a layout that does not look like the rest of the site.

layout: null
nav: projects
status: published
listed: true
date: 2026-05-26
permalink: /projects/your-slug/
title: Paper or project title
year: 2026
summary: Short introduction used on project cards.
research_area: Social Deduction
research_area_slug: social-deduction
authors:
  - First Author
members:
  - Community member on the card
venue: EMNLP 2026 Main Conference
publication_group: venue
keywords:
  - Keyword
Field Used by
listed: true Home ticker, Publication, Projects catalog
date Sort order (newest first)
year Card year label
venue Ticker (venue only) and Publication cards
publication_group Optional. preprint or venue. If omitted, a venue containing arXiv is treated as a preprint
authors / members Publication cards and project cards
research_area_slug Projects area filter

Set listed: false for drafts and templates.

Local development

Ruby 3.4 and Bundler are required.

bundle install
bundle exec jekyll serve --port 4000

Open http://127.0.0.1:4000/.

On Windows the file watcher often misses CSS and includes. If a change does not appear, rebuild:

bundle exec jekyll build

Then refresh the browser (hard-refresh if the tab icon looks stale).

Restart jekyll serve after editing _config.yml. The watcher does not reload collections, so People can render empty until the server is restarted.

Search discovery and indexing

The tracked root sitemap.xml is a Liquid template. Jekyll renders it into _site/sitemap.xml at build time and deploys it at /sitemap.xml; _site/ stays untracked. jekyll-sitemap detects this template and does not replace it. The production origin is set in _config.yml; /robots.txt allows crawling and declares that sitemap. Formal pages use _includes/seo-canonical.html once inside their <head>. The homepage adds WebSite and Organization JSON-LD through _includes/home-identity.html. Keep its name, logo, canonical URL, and verified profiles consistent with the visible site. Its description comes from homepage front matter. Shared Home and wordmark links point directly to /. Reveal animations enhance content that is visible by default, including without JavaScript. The /projects/ redirect declares /publication/ as its canonical and uses sitemap: false. Google ownership verification stays published, unchanged, and is omitted from the sitemap.

The template keeps the plugin's URL inclusion rules but only emits <lastmod> when content supplies an explicit last_modified_at date. Update that field only after a significant change to content, structured data, or links. If a reliable modification date is unknown, omit it. Do not substitute a build time, file checkout time, or the paper's publication date: those can falsely suggest that unchanged pages were updated on every deployment.

A valid live sitemap does not guarantee that Google has fetched or processed it. If Search Console reports "Couldn't fetch", inspect the sitemap's error details and run its live URL test; a successful browser request alone does not resolve Google's report. Follow the official troubleshooting steps.

The sitemap lists HTML content pages, not resume PDFs. Rui Song's reserved Talk and Service pages remain accessible but use sitemap: false and noindex: true. When those sections contain real content, remove those two flags, link the pages from navigation, and update the explicit NOINDEX allowlist in the checker.

listed: false only hides a page from content lists. It does not prevent publication or indexing. The reusable project and member templates are excluded from the build by their exact source paths in _config.yml. Copy a template to a new slug to publish a real page; keep private drafts outside the published source or explicitly exclude them.

Published papers keep their formal title for lists and the visible subtitle. An optional seo_title supplies a concise browser/search title. Render authors with _includes/paper-authors.html: exact profile names link automatically; verified name variations belong in the member's author_names list. Authors without a matching profile remain plain text. equal_contributors preserves the paper's contribution markers.

After a production build, run:

bundle exec jekyll build
python scripts/check_seo.py _site

The checker validates homepage identity metadata, sitemap coverage, canonical targets, indexing directives, robots rules, content links, navigation from Home, and the verification file. It derives required member/project URLs from source files instead of fixing a member count. Pull requests run the same checks without deploying. Only a successful build from main can deploy.

Deployment

Push to main to run .github/workflows/pages.yml. The workflow builds the Jekyll source and deploys _site to GitHub Pages.

Do not commit _site/, .agents/, or skills-lock.json.

About

Official website of New Ground AI, an AI research collective.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages