This paper introduces Gated Graph Sequence Neural Networks (GGS-NNs), a novel class of neural network models designed for learning features from graph-structured inputs and producing sequence outputs. The authors build upon previous work on Graph Neural Networks (GNNs) by incorporating gated recurrent units and modern optimization techniques, extending them to handle sequences. GGS-NNs are particularly useful for problems with graph inputs that require sequence outputs, such as path finding, node selection, and graph-level classifications. The model is evaluated on various tasks, including bAbI reasoning tasks and graph algorithm learning, demonstrating its effectiveness. A key application is program verification, where GGS-NNs can learn to map graph representations of memory states to logical formulas describing data structures. The paper also discusses related work and future directions, highlighting the potential of GGS-NNs in combining structured representations with deep learning algorithms.This paper introduces Gated Graph Sequence Neural Networks (GGS-NNs), a novel class of neural network models designed for learning features from graph-structured inputs and producing sequence outputs. The authors build upon previous work on Graph Neural Networks (GNNs) by incorporating gated recurrent units and modern optimization techniques, extending them to handle sequences. GGS-NNs are particularly useful for problems with graph inputs that require sequence outputs, such as path finding, node selection, and graph-level classifications. The model is evaluated on various tasks, including bAbI reasoning tasks and graph algorithm learning, demonstrating its effectiveness. A key application is program verification, where GGS-NNs can learn to map graph representations of memory states to logical formulas describing data structures. The paper also discusses related work and future directions, highlighting the potential of GGS-NNs in combining structured representations with deep learning algorithms.