Frontier AI Evals Hackathon at Cornell

​Why do we have so many benchmarks, and yet agents still fail at so many real-world tasks?

And how can this be when we have models discovering math proofs every week?

The answer: We haven't measured everything.

The Frontier AI Evals Hackathon brings Cornell's most ambitious, of varied backgrounds, to push the frontier of evals in AI. Whether it is to redefine a long-standing domain or to define "what good looks like?" in a brand new area.

We want to make evals where the best models today still get things wrong.

Registration opening soon.

Sponsors: to be announced.