LMArena is a public, web-based platform that evaluates large language models through anonymous, crowd-sourced pairwise comparisons. From Wikipedia
A new study accuses LM Arena of granting major AI labs preferential testing access, prompting denials and plans to revise sampling methods.
