Continual-ARCby unify

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

Demo-cache floor

The model alone, shown every demonstration it bought for the task again for free.

Held-out tasks, budget used
37.6 solved 80%
Model spend
$0.95 588 calls
Instances
100 4.9 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