A Stepwise Questioning Expert-Editor Multi-Agent Framework for Long-Document Summarization

The paper proposes a multi-agent framework for long-document summarization in which an expert agent and an editor agent ask stepwise questions about different aspects of a document, then provide targeted clues to another agent that revises the summary. Experiments were conducted on two scientific long-document datasets and evaluated with recognized automatic metrics. The abstract reports that the method was effective, but does not specify the dataset names, model configurations, baseline systems, metric values, computational cost, or statistical significance.