Understanding the "Fast Track" Process on Hugging Face's Leaderboard

#121
by hunkim - opened

I've been observing the Hugging Face Leaderboard and noticed that certain projects, such as Llama2 and StabilityAI, seem to bypass the queue and make their way directly onto the leaderboard. I'm interested in understanding how this process operates.

Could anyone provide information about the criteria and process for this "Fast Track"? What are the required steps, and what conditions need to be fulfilled? Additionally, any advice on how one could enhance their chances of having their project fast-tracked would be greatly appreciated.

Thank you in advance for your guidance and assistance!

Open LLM Leaderboard org
edited Jul 27, 2023

Hi!
"Fast track" projects happen exceptionally, for companies/organizations 1) that are partners with HF, 2) that are releasing new SOTA models, to allow them to appear on the leaderboard on release day. They should actually be called "pre-release tracks", to be exact.

From a hardware standpoint, this does not affect the queue of the Open LLM Leadeboard, as they are evaluated independently (at the moment, @SaylorTwift is running the main leaderboard and I've been taking care of the fast releases).

clefourrier changed discussion status to closed

Sign up or log in to comment