growdark·9 ay önce·discussI'd love to see a benchmark that tests different LLMs for slop, not necessarily limited to code. That might be even more interesting than ARC-AGI.