Spaces:

open-llm-leaderboard

/

open_llm_leaderboard

like 10.2k

Running on CPU Upgrade

Right now we have filters for different benchmarks, but those filters only show/hide columns; the avarage score stays the same (taking into consideration also what's hidden).

It would make a lot more sense to recompute avarage score based on shown column.

For example, I might look for a model to further finetune and might only be interested in the avarage of ARC and GSM8K, but there's no possibility to sort based on that at this moment.