This paper introduces Relation Networks (RNs) as a simple and effective module for relational reasoning in neural networks. RNs are designed to explicitly focus on relational reasoning, making them suitable for tasks that require understanding and inferring relationships between entities. The authors test RN-augmented networks on three diverse tasks: visual question answering using the CLEVR dataset, text-based question answering using the bAbI suite, and complex reasoning about dynamic physical systems. The results show that RNs achieve state-of-the-art performance on CLEVR, outperforming previous models by 27% and surpassing human performance. They also demonstrate the versatility of RNs by solving 18 out of 20 tasks in the bAbI suite and achieving high accuracy in the dynamic physical systems task. The paper highlights the strengths of RNs, including their ability to learn to infer relations, data efficiency, and order-invariance, and discusses future directions for improving the efficiency and applicability of RNs.This paper introduces Relation Networks (RNs) as a simple and effective module for relational reasoning in neural networks. RNs are designed to explicitly focus on relational reasoning, making them suitable for tasks that require understanding and inferring relationships between entities. The authors test RN-augmented networks on three diverse tasks: visual question answering using the CLEVR dataset, text-based question answering using the bAbI suite, and complex reasoning about dynamic physical systems. The results show that RNs achieve state-of-the-art performance on CLEVR, outperforming previous models by 27% and surpassing human performance. They also demonstrate the versatility of RNs by solving 18 out of 20 tasks in the bAbI suite and achieving high accuracy in the dynamic physical systems task. The paper highlights the strengths of RNs, including their ability to learn to infer relations, data efficiency, and order-invariance, and discusses future directions for improving the efficiency and applicability of RNs.