SGR: A Stepwise Reasoning Framework for LLMs with External Subgraph Generation

TL;DR

SGR is a framework that improves large language model reasoning by generating and utilizing external subgraphs from knowledge bases for structured, multi-step inference.

Contribution

It introduces a novel stepwise reasoning framework that grounds LLMs in external structured knowledge to enhance accuracy and factual consistency.

Findings

SGR achieves consistent improvements on benchmark datasets.

Grounding reasoning in external subgraphs enhances factual reliability.

The method supports multi-step inference with structured external knowledge.

Abstract

Large Language Models (LLMs) have demonstrated strong capabilities across diverse NLP applications, such as translation, text generation, and question answering. Nevertheless, they remain limited in complex settings that demand deep reasoning and logical inference. Since these models are trained on large-scale text corpora, their generation process may still introduce irrelevant, noisy, or factually inconsistent content. To mitigate this problem, we introduce SGR, a stepwise framework that enhances LLM reasoning through external subgraph generation. SGR builds query-specific subgraphs from external knowledge bases and uses their semantic structure to support multi-step inference. By grounding intermediate reasoning steps in structured external knowledge, the framework helps the model concentrate on relevant entities, relations, and supporting evidence. In particular, SGR first constructs a subgraph tailored to the input question. It then guides the model to reason progressively over the generated structure and combines multiple reasoning trajectories to obtain the final prediction. Experimental results across several benchmark datasets show that SGR achieves consistent improvements over competitive baselines, highlighting its value for improving both reasoning accuracy and factual reliability.