What it does
Jev sees only titles, not the labels used to check its answers. The experiment also asks which posts seem worth reading, but has no reference answers for that opinion.
Benchmarks & research
my-jev-playground
Classify GitHub changelog titles with Jev, then compare the answers with labels GitHub already supplies.
Bookmark: my-jev-playground
Keep this in your collection.
Leave a noteWhat would you try with this?
: my-jev-playground
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
Borrow the idea of hiding known answers before testing an AI. A developer can send article titles without their labels, then compare the AI’s answers with those labels and look at its mistakes. The source also asks which articles seem worth reading, but has no known answers for checking that opinion.