Continual-ARCby unify

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

Program library

Every answer is a program, kept once it reproduces the demonstrations and run first when the task returns.

Share learned
+0.62 ±0.15
Stream, budget used
8.9 solved 97%
Epilogue, budget used
18.6 solved 99%
Model spend
$1.70 782 calls
Instances
300 9.9 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. The share learned reads the stream's budget against the learner itself reset, the same system wiped before every instance of the same stream: the share of the budget left to learn that the learner saves, 0 for a learner that keeps nothing and 1 for perfect memory.

Variant
Effort
Order

Loading the run