Scientific documents often present dense and intricate information, making the generation of concise and coherent summaries a challenging task. This study proposes Gated Graph Neural Attention Networks (GGNAT) to address this issue. GGNAT integrates graph-based neural networks with attention mechanisms to effectively model hierarchical relationships and contextual dependencies within scientific texts. By leveraging this novel approach, the proposed method enhances the model’s ability to understand the structural and semantic nuances of scientific papers. The results show the improvements in the accuracy and the contextual relevance of abstractive summarization, delivering precise and meaningful summaries that faithfully represent the original content.