Continual-ARCby unify

ReplayARC-AGI-1 · Fresh · GPT-6 Luna, high

Stateless floor

The model alone, with nothing kept between instances.

Held-out tasks, budget used
29.7 solved 88%
Model spend
$1.27 511 calls
Instances
100 8.2 hours

Budget used is the share of the cap of 16 that an instance was charged, averaged over the instances: a solved instance costs 1 for each wrong answer and each demonstration, an unsolved one all 16, so lower is better.

Loading the run