Towards Faithful Knowledge Graph Explanation Through Deep Alignment in Commonsense Question Answering

TL;DR

This paper proposes a new metric and alignment algorithm to improve the faithfulness of knowledge graph explanations in commonsense question answering, addressing misalignment issues between language models and KGs.

Contribution

It introduces the LM-KG Fidelity metric and the LKDA algorithm to enhance explanation faithfulness and model performance without requiring ground truth explanations.

Findings

LKDA significantly improves explanation fidelity.

LKDA enhances model accuracy on commonsense QA datasets.

Addressing distributional misalignment is crucial for reliable explanations.

Abstract

The fusion of language models (LMs) and knowledge graphs (KGs) is widely used in commonsense question answering, but generating faithful explanations remains challenging. Current methods often overlook path decoding faithfulness, leading to divergence between graph encoder outputs and model predictions. We identify confounding effects and LM-KG misalignment as key factors causing spurious explanations. To address this, we introduce the LM-KG Fidelity metric to assess KG representation reliability and propose the LM-KG Distribution-aware Alignment (\textit{LKDA}) algorithm to improve explanation faithfulness. Without ground truth, we evaluate KG explanations using the proposed Fidelity-Sparsity Trade-off Curve. Experiments on CommonsenseQA and OpenBookQA show that LKDA significantly enhances explanation fidelity and model performance, highlighting the need to address distributional misalignment for reliable commonsense reasoning.