A model benchmark joke: count Fields Medals or Abel Prizes instead

beffjezos · x · 2026-07-23

A tongue-in-cheek benchmark proposal: judge models by how many Fields Medals or Abel Prizes they win.

It’s a joke, but it riffs on the obsession with benchmarks and implies the next real bar for reasoning models might be mathematical excellence rather than synthetic evals.

Original post →

More from Fun

Fun channel →