SCALE PRIVATE LEADERBOARD

Summary from the TLDR AI newsletter

Scale has made a private leaderboard for evaluating language models. The ordering isn't too surprising, but one interesting finding is the Llama 3 70B is extremely good at following instructions, often better than Claude Opus.