VaDar research website

Website for VaDar: Implicit Goal Conditioning via Value Disaggregation.

Live site: https://simchowitzlabpublic.github.io/vadar-site/

Editing

• dist/index.html: text, figures, videos, section order, and citation.
• dist/style.css: layout and typography.
• dist/replay.js: shared-critic animation.
• dist/weighting.js: next-future weighting animation.
• dist/media.js: video controls and expandable panels.
• dist/assets/: paper, figures, and rollout videos.

Edit files directly on GitHub or clone this repository. Changes pushed to main automatically publish through GitHub Pages. No build step is required.

For local preview:

python3 -m http.server 8765 --directory dist

Open http://localhost:8765.

The success videos encode 4× timing in their files and play at normal browser speed. Main-results videos use trained policies; best-of-N videos use frozen-policy critic steering. The main-results video panel is collapsed by default.

KaTeX is vendored with its license in dist/vendor/katex/LICENSE.