GRASP: Graph-Reasoning Aided Survey Planning for High-Fidelity Related Work Generation

2026-07-04

AI Summary

This work addresses the challenge of accurately modeling complex inter-paper relationships—such as inheritance, contradiction, and alternative perspectives—in literature reviews. To this end, it proposes GRASP, a novel framework that synergistically integrates large language models with graph-based reasoning. GRASP constructs a two-layer graph structure comprising a Thought graph and an argument–counterargument network, and employs Steiner tree-based topological pruning to distill core citation relationships. Experimental results demonstrate that literature review sections generated by GRASP exhibit strong alignment with human-written counterparts in terms of citation discourse roles, intent recognition, and thematic grouping of references, thereby significantly enhancing the logical coherence and fidelity of scholarly syntheses.

Abstract

Writing a literature review requires a deep understanding of the relationships among cited papers: how they build on, challenge, or offer alternative perspectives to one another. We present Graph-Reasoning Aided Survey Planning (GRASP), a framework combining LLM planning for related work generation with graph algorithms to extract key relationships among cited papers. Our two-layer graph structure consists of a Graph of Thoughts and an Argument-Counterargument Planning Network, representing the cited papers at different levels of granularity, and we apply topology-aware pruning via a Steiner tree to identify the core inter-paper relationships captured in our graph. Our citation analysis-based evaluation shows that GRASP generates related work sections (RWS) that closely match human-written targets in terms of the discourse roles, intents, and grouping of citations.

Problem

• literature review
• citation relationships
• related work generation
• graph reasoning
• survey planning

Innovation

• Graph Reasoning
• Survey Generation
• Steiner Tree Pruning
• Argument-Counterargument Network
• LLM Planning