Dear huggingface leaderboard

#820
by MattMills - opened

Dear hugginface leaderboard, you measure so many things, but why are you so biased to $now?

Dear people-that-actually-develop-the-huggingface-leaderboard,

Can you please add something that shows the changing performance of the individual models, the aggregate leader, and different fancy statistical averages or whatever you smart people would use to see the rates of improvement over time?

Can we determine anything of value from that rate information, or make any predicions?

Open LLM Leaderboard org

Hi @MattMills ,

Thanks for the issue! As all our data is public, you're very welcome to develop a tool to extract the information you want from our data :)
You'll find all the paths to all the info in our doc (linked at the top of the leaderboard), both for this version and the previous one.

Looking forward to see what you'll build!

clefourrier changed discussion status to closed

OK I'll do that, I figured you might rather have the suggestion since I don't know your platform and anything I developed for my own consumption would be outside of it in something like grafana.

Thanks,
Matt.

Ah, I see that existed on the old leaderboard but not the new one for whatever reason, I'll just look at the old one for now that works fine.

Sign up or log in to comment