When to Trust: A Causality-Aware Calibration Framework for Accurate Knowledge Graph Retrieval-Augmented Generation
This addresses calibration issues in KG-RAG for high-stakes domains, representing an incremental improvement with a novel method for a known bottleneck.
The paper tackles the problem of overconfidence in Knowledge Graph Retrieval-Augmented Generation (KG-RAG) models, which produce high-confidence predictions even with unreliable retrieved knowledge, by proposing Ca2KG, a causality-aware calibration framework that improves calibration while maintaining or enhancing predictive accuracy in complex QA tasks.
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.