Papers With Code was a free, community-driven resource launched in July 2018 by Robert Stojnic and Ross Taylor. Its core idea was simple but influential: connect each machine-learning paper to its open-source implementation, and organize results into public leaderboards by task and dataset so anyone could see, at a glance, the current state of the art and how it had moved over time. By late 2019 it indexed over 18,000 papers with code, more than 1,000 tasks, and over 1,500 leaderboards, with all data released under an open license.
In December 2019 the founders announced the project was joining Facebook AI (now Meta) to accelerate its growth, with a public commitment that it would remain a neutral, open, and free resource. The site became a default tool for researchers and practitioners checking what method led on a given benchmark and whether code was available to reproduce or build on it - reinforcing a norm that progress claims should come with runnable code and a place on a shared scoreboard.
Papers With Code matters as a piece of the field’s measurement culture: it made “state of the art” a tracked, public, comparable quantity rather than a claim buried in each paper’s abstract. That transparency speeds up adoption of good ideas and exposes when reported gains do not hold up. For a general reader, it is a clear example of community infrastructure quietly shaping how a scientific field judges itself.