Towards Efficient Neurally-Guided Program Induction for ARC-AGI

TL;DR

This paper investigates different neurally-guided program induction methods for ARC-AGI, focusing on their efficiency and generalization, and proposes a new approach based on learning the transform space.

Contribution

It compares three paradigms of program induction, identifies their strengths and weaknesses, and introduces a novel transform space approach for improved generalization.

Findings

Learning the grid space is efficient but limited in generalization.

Learning the program space offers better generalization but is less efficient.

The transform space approach shows promise in balancing efficiency and generalization.

Abstract

ARC-AGI is an open-world problem domain in which the ability to generalize out-of-distribution is a crucial quality. Under the program induction paradigm, we present a series of experiments that reveal the efficiency and generalization characteristics of various neurally-guided program induction approaches. The three paradigms we consider are Learning the grid space, Learning the program space, and Learning the transform space. We implement and experiment thoroughly on the first two, and retain the second one for ARC-AGI submission. After identifying the strengths and weaknesses of both of these approaches, we suggest the third as a potential solution, and run preliminary experiments.