Continual-ARCby unify

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

Hermes Agent, as shipped

Every learning mechanism the harness ships switched on.

Held-out tasks, budget used
54.1 solved 61%
Model spend
$7.02 16,398 calls
Instances
100 9.7 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