Untrained neural networks can demonstrate memorization-independent abstract reasoning

TL;DR

This paper demonstrates that untrained neural networks can perform abstract reasoning tasks without prior training or memorization, challenging the belief that extensive training is necessary for such capabilities.

Contribution

It introduces a method where neural network weights are optimized during problem solving, showing that untrained networks can achieve abstract reasoning without memorization.

Findings

Untrained neural networks can solve visual reasoning problems.

Performance does not depend on memorizing similar problems.

Problem solving involves dynamic weight optimization in the network.

Abstract

The nature of abstract reasoning is a matter of debate. Modern artificial neural network (ANN) models, like large language models, demonstrate impressive success when tested on abstract reasoning problems. However, it has been argued that their success reflects some form of memorization of similar problems (data contamination) rather than a general-purpose abstract reasoning capability. This concern is supported by evidence of brittleness, and the requirement of extensive training. In our study, we explored whether abstract reasoning can be achieved using the toolbox of ANNs, without prior training. Specifically, we studied an ANN model in which the weights of a naive network are optimized during the solution of the problem, using the problem data itself, rather than any prior knowledge. We tested this modeling approach on visual reasoning problems and found that it performs relatively well. Crucially, this success does not rely on memorization of similar problems. We further suggest an explanation of how it works. Finally, as problem solving is performed by changing the ANN weights, we explored the connection between problem solving and the accumulation of knowledge in the ANNs.