A more open process
As with the previous edition of the Blog Post track, we forgo the requirement for total anonymity. The blog posts must be anonymized for the review process, but users will submit their anonymized blog posts via a pull request to the blog track’s repository (in addition to a submission on OpenReview). The pull request will trigger an automated pipeline that will build and deploy your post onto a website dedicated to the reviewing process. Reviewers will be able to access the posts directly through a public URL (generated by the Github action), and will submit their reviews on OpenReview. Reviewers should refrain from looking at the git history for the post, which may reveal information about the authors.
This still largely follows the Double-Blind reviewing principle; it is no less double-blind than when reviewers are asked to score papers that have previously been released to arXiv, an overwhelmingly common practice in the ML community. This approach was chosen to lower the burden on both the organizers and the authors; in 2022, many submissions had to be reworked once deployed due to a variety of reasons. By allowing the authors to render their websites to Github Pages prior to the review process, we hope to avoid this issue entirely.
However, we understand the desire for total anonymity. Authors that wish to have a fully double-blind process might consider creating new GitHub accounts without identifying information which they will only be use for this track. For an example of a submission in the past which used an anonymous account in this manner, you can check out the World Models blog post (Ha and Schmidhuber, 2018) and the accompanying repository.
Template
The workflow you will use to participate in this track should be relatively familiar to you if have used Github Pages. Specifically, our website uses the Al-Folio template. This template uses Github Pages as part of its process, but it also utilizes a separate build step using Github Actions and intermediary Docker Images.