Enhancing Analogical Reasoning in the Abstraction and Reasoning Corpus via Model-Based RL

TL;DR

This paper shows that model-based reinforcement learning, exemplified by DreamerV3, improves analogical reasoning in the ARC by creating internal models, outperforming model-free methods in efficiency and generalization.

Contribution

It demonstrates that model-based RL enhances analogical reasoning capabilities in the ARC, offering a more effective approach than traditional model-free methods.

Findings

Model-based RL outperforms model-free RL in ARC tasks.

DreamerV3 demonstrates better generalization from single tasks.

Model-based RL shows significant advantages in reasoning across similar tasks.

Abstract

This paper demonstrates that model-based reinforcement learning (model-based RL) is a suitable approach for the task of analogical reasoning. We hypothesize that model-based RL can solve analogical reasoning tasks more efficiently through the creation of internal models. To test this, we compared DreamerV3, a model-based RL method, with Proximal Policy Optimization, a model-free RL method, on the Abstraction and Reasoning Corpus (ARC) tasks. Our results indicate that model-based RL not only outperforms model-free RL in learning and generalizing from single tasks but also shows significant advantages in reasoning across similar tasks.