Follow the Path: Reasoning over Knowledge Graph Paths to Improve Large Language Model Factuality

TL;DR

This paper presents fs1, a method that enhances LLM factuality by grounding reasoning traces in knowledge graph paths, leading to improved performance on complex QA tasks.

Contribution

Introducing fs1, a technique that fine-tunes LLMs with knowledge graph-grounded reasoning traces, significantly boosting factual reasoning accuracy.

Findings

fs1-tuned models outperform instruction-tuned models by 6-14 points on pass@16.

Significant improvements on questions requiring 3+ hops and numerical answers.

Smaller LLMs benefit most from the fs1 approach in single-pass inference.

Abstract

We introduce fs1, a simple yet effective method that improves the factuality of reasoning traces by collecting them from large reasoning models and grounding them in knowledge graph (KG) paths. We fine-tune eight instruction-tuned Large Language Models (LLMs) on 3.9K factually grounded reasoning traces and rigorously evaluate them on six complex open-domain question-answering (QA) benchmarks encompassing 23.9K questions. Our results demonstrate that our fs1-tuned model consistently outperforms instruction-tuned counterparts with parallel sampling by 6-14 absolute points (pass@16). Our detailed analysis shows that fs1 considerably improves model performance over more complex questions (requiring 3 or more hops on KG paths) and numerical answer types compared to the baselines. Furthermore, in single-pass inference, we notice that smaller LLMs show the most improvements. While prior works demonstrate the effectiveness of reasoning traces primarily in the STEM domains, our work shows strong evidence that anchoring reasoning to factual KG paths is a critical step in transforming LLMs for reliable knowledge-intensive tasks.