Visual Graph Scaffolds for Structural Reasoning in Large Language Models
Quick Take
This study explores the use of visual graph scaffolds to enhance reasoning in large language models (LLMs) during multi-hop question answering. Unlike traditional methods that rely on external knowledge, the research shows that internal graph structures significantly improve reasoning efficiency and answer quality, even without direct answer hints. The findings suggest that graphs should be integrated as internal reasoning aids rather than just external knowledge sources.
Key Points
- Visual graphs improve reasoning efficiency in LLMs during multi-hop question answering tasks.
- Flattening graph structures into text limits their effectiveness when direct answer hints are absent.
- Graph mind maps serve as effective internal reasoning aids, enhancing answer quality.
- Supervised fine-tuning and KL-based distillation maintain the advantages of visual graph guidance.
- The study advocates for integrating graphs as scaffolds for organizing reasoning in LLMs.
Article Content
From source RSS / original summaryarXiv:2606. 02673v1 Announce Type: new Abstract: Graphs have been used to enhance large language models (LLMs) for structured reasoning, mostly as external knowledge sources are provided to models at test time. In this paper, we take a different view: the value of graphs for LLMs lie not only in supplying information, but also in organizing reasoning.
Inspired by how humans use graph-structured mind maps to organize branching and converging thoughts, we ask whether graphs can serve as an internal form of reasoning assistance. We study this question on multi-hop question answering tasks, where teacher-provided reasoning traces are rewritten as graph mind maps and used to guide a student model. Our experiments reveal a clear modality gap. When graph structures are flattened into text, their benefits become limited once direct answer hints are removed.
Under this abstract guidance setting, both reasoning efficiency and answer quality degrade substantially. In contrast, visual graph guidance remains effective without direct answer clues, and its advantage persists after supervised fine-tuning and KL-based distillation. The above findings support the claim that graphs should be studied not only as external knowledge structures for LLMs, but also as visual scaffolds for organizing reasoning.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →AUDITFLOW: Executable Symbolic Environments for Structured Financial Reporting Verification
AuditFlow introduces a multi-agent framework for structured financial reporting verification, achieving 82.09% accuracy with GPT-5.5, outperforming the baseline by 14.93 points. It utilizes a symbolic environment for effective audit processes, demonstrating the necessity of deterministic checks for reliable verification.