benchmarks safetyNew
ARC Prize
ARC-AGI benchmark designed to resist memorization and test true generalization.
weight 0.0FreeLaunched 2026-06-07
💸 No earnings reported yet
What it is
The consortium behind ARC-AGI, a benchmark of novel tasks absent from training corpora that forces genuine generalization; ARC-AGI-3 has broken every agent tested against it.
How AI plugs in
Defines and scores AI on ARC-AGI, a benchmark of novel reasoning puzzles designed to resist memorization and force genuine generalization beyond what training data can supply.
★ Reviews
No reviews yet — be the first.Your rating