.png){kind=icon}
.png){kind=icon}
Return
A multi agent collaborative framework for style guided knowledge reorganization of scientific source texts
.png){kind=icon}
OA
AI
DOI：10.1007/s40747-026-02259-7.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Popular science writing requires reorganizing dense scientific materials into accessible explanations, yet manual production is costly and difficult to scale. While large language models enable efficient rewriting, existing systems often lack explicit planning for information reorganization and are prone to single-agent bias, leading to fragmented explanations or weakened scientific rigor. To address these challenges, this paper proposes a multi-agent collaborative framework for style-guided, source-grounded knowledge reconstruction in popular science communication, termed STMAC. The framework restructures and refines scientific source materials through a role-divided workflow, where LLM-based domain expert agents provide domain grounding and verification, and rewriting agents realize reader-friendly expression under style and structure constraints. This collaboration improves clarity and readability while maintaining alignment with the original source materials. Experiments across multiple datasets with diverse content types show that STMAC achieves consistent preference gains under both LLM based evaluators and human assessment over representative recent style transfer baselines, indicating higher quality rewrites and improved scientific rigor. These results suggest a practical and controllable solution for applying LLMs to source grounded knowledge reorganization for popular science communication.
Keywords:
Large language model
Multi-agent collaboration
Source-grounded rewriting
Style transfer
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
C
IF:
4.6
Papers:
250
Citations:
0
View more organizations