Slople – Can you tell real ML papers from AI-generated ones?

I was inspired by https://slop.zackg.me/, which does the same thing for PL papers but with entire fully generated AI papers (i.e. LaTeX-rendered PDFs and all) - check it out! I thought it was fun to play, and wanted a version for ML, so made this website.

It's set up with a GitHub Actions cron job so that every day there are 5 real arXiv papers and 5 Claude-generated ones, and you try to guess which are real or fake from just the title and abstract.

Ironically, most of this was vibe coded with Claude Code. The exception is the UI, which uses Tufte CSS.