ARC solvers can induce rules by tuning under 0.01% of parameters
A two-step test-time training method helps ARC-like models separate learning a task’s rule from executing it, while tuning less than 0.01% of parameters can already solve a non-trivial fraction of tasks. The method first adapts a task embedding, a compact representation of the transformation rule, then freezes it while adapting the model backbone. Across ARC-AGI-1, ConceptARC, and controlled datasets, the embeddings align better with known rules and improve retrieval and linear probing, though they interpolate rules without extrapolating them.