Continual-ARCby unify

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

Program library, nothing kept

The program library with its programs deleted before every instance.

Held-out tasks, budget used
24.2 solved 92%
Model spend
$0.71 529 calls
Instances
100 3.6 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.

Variant
Effort
Order

Loading the run