Seeing the Twitter rumors about how great GPT4.5 was at drawing unicorns made me think that there should be a way to rank models based on their SVG drawing skills!
This started (and still mostly is) a joke, but looking at the outputs and the leaderboard, I think this is actually very decent eval (much better than most lmsys categories!)
Seeing the Twitter rumors about how great GPT4.5 was at drawing unicorns made me think that there should be a way to rank models based on their SVG drawing skills!
This started (and still mostly is) a joke, but looking at the outputs and the leaderboard, I think this is actually very decent eval (much better than most lmsys categories!)
And looking at previous battles is just fun :)