About Royal Road

Royal Road is a living index of foundational machine-learning papers, each paired with an AI-generated review and an explorable citation graph.

What is this?

Every paper on this site has been read by a language model that drafts a structured review: the problem, the proposed method, prior work it builds on, the evaluation, and the strengths and weaknesses. Reviews are versioned in a vault of Markdown notes, and the citation graph between papers is rebuilt on every deploy.

How the reviews are produced

Reviews are drafted by an AI pipeline that reads the paper PDF, extracts a structured outline, and writes each section under a fixed template. Drafts are then revised by a human editor before being marked reviewed. Both states are visible on the site — the badge on each paper tells you which.

The pipeline is intentionally opinionated: it favours sentences over vague hedges, prefers concrete metrics over adjectives, and includes the figures it considered most informative when possible. It will be wrong sometimes.

The graph

Each paper carries a prior_papers list (work it builds on) and a further_papers list (work that builds on it). Those are the edges of the citation graph. The explorer lets you trace lineages forward and backward.

Contact

Royal Road is maintained by Taeung Jeong.