Spaces:
Running
on
CPU Upgrade
Update model size filter intervals
Hello!
We have cloned this leaderboard to run a Open Korean LLM Leaderboard. We have over 100 models in the past couple of weeks so far!
As discussed with
@clefourrier
, I'm proposing this change to make the model size filter behavior more clear.
Currently the notation like ~13B
means around 13B
and shows models between 12B and 15B. Although individuals may interpret the notation differently, we thought this could be improved and be more intuitive. Moreover, with ~3B
and ~7B
shows 1.5B~5B
and 6B~11B
respectively, there is no way to filter for models with size between 5B and 6B, such as 5.8B. We actually had a request from one of participants for size 5.8B. (See discussion)
So the proposal is to change the notation to 1.5~3B
and filter for models with size between 1.5B and 3B.
I hope this makes sense!
LGTM, thank you!