When to Trust: A Causality-Aware Calibration Framework for Accurate Knowledge Graph Retrieval-Augmented Generation

TL;DR

This paper introduces Ca2KG, a causality-aware calibration framework for KG-RAG that improves the reliability and calibration of knowledge graph-based language models, especially in high-stakes scenarios.

Contribution

It presents a novel calibration method combining counterfactual prompting and panel-based re-scoring to address overconfidence in KG-RAG models.

Findings

Ca2KG improves calibration across datasets.

Maintains or enhances predictive accuracy.

Reduces overconfidence in uncertain retrievals.

Abstract

Knowledge Graph Retrieval-Augmented Generation (KG-RAG) extends the RAG paradigm by incorporating structured knowledge from knowledge graphs, enabling Large Language Models (LLMs) to perform more precise and explainable reasoning. While KG-RAG improves factual accuracy in complex tasks, existing KG-RAG models are often severely overconfident, producing high-confidence predictions even when retrieved sub-graphs are incomplete or unreliable, which raises concerns for deployment in high-stakes domains. To address this issue, we propose Ca2KG, a Causality-aware Calibration framework for KG-RAG. Ca2KG integrates counterfactual prompting, which exposes retrieval-dependent uncertainties in knowledge quality and reasoning reliability, with a panel-based re-scoring mechanism that stabilises predictions across interventions. Extensive experiments on two complex QA datasets demonstrate that Ca2KG consistently improves calibration while maintaining or even enhancing predictive accuracy.