ScholarSum: Student-Teacher Abstractive Summarization via Knowledge Graph Reasoning and Reflective Refinement

ScholarSum introduces a hierarchical reflective graph-based framework for abstractive summarization, designed to enhance both linguistic fluency and factual accuracy in scientific literature. The model organizes documents into a knowledge graph, facilitating a student-teacher dynamic where an initial draft is refined through evidence retrieval and iterative review to ensure content fidelity. Extensive experiments show that ScholarSum surpasses existing baselines in completeness and faithfulness, providing a valuable tool for practitioners focused on generating high-quality summaries from complex texts.