Inverse Scaling Prize: Identify tasks where LLMs performance worsens with size (github.com)

TL;DR: Win up to $100,000 for finding an important task where larger language models do worse.