New study accuses LM Arena of gaming its popular AI benchmark - Ars Technica

The rapid proliferation of AI chatbots has made it difficult to know which models are actually improving and which are falling behind.