Deep Neural Networks Can Learn Generalizable Same-Different Visual Relations
This addresses a key limitation in AI for visual reasoning, showing that with appropriate methods, networks can overcome prior failures in abstract relation learning, though it is incremental as it builds on existing architectures and pretraining.
The study tackled the problem of whether deep neural networks can learn and generalize same-different visual relations, finding that certain pretrained transformers achieve near perfect accuracy on out-of-distribution stimuli, particularly when fine-tuned on abstract shapes without texture or color.
Although deep neural networks can achieve human-level performance on many object recognition benchmarks, prior work suggests that these same models fail to learn simple abstract relations, such as determining whether two objects are the same or different. Much of this prior work focuses on training convolutional neural networks to classify images of two same or two different abstract shapes, testing generalization on within-distribution stimuli. In this article, we comprehensively study whether deep neural networks can acquire and generalize same-different relations both within and out-of-distribution using a variety of architectures, forms of pretraining, and fine-tuning datasets. We find that certain pretrained transformers can learn a same-different relation that generalizes with near perfect accuracy to out-of-distribution stimuli. Furthermore, we find that fine-tuning on abstract shapes that lack texture or color provides the strongest out-of-distribution generalization. Our results suggest that, with the right approach, deep neural networks can learn generalizable same-different visual relations.