Model Leaderboard

Comprehensive comparison using both Weighted and Unweighted ranking methods.

Model Params Weighted Rank Weighted Score
[95% CI] 		Unweighted Rank Unweighted Score
[95% CI]
* Click headers to sort. "Weighted" refers to our judge-aware ranking method; "Unweighted" is the raw aggregate score.
* Scores are shown with 95% confidence intervals in brackets.