3 ms·Inverse Scaling Prize: Identify tasks where LLMs performance worsens with size2 points by ravmachre 3y agoravmachre 3y agoTL;DR: Win up to $100,000 for finding an important task where larger language models do worse.