Mask-to-Correct$^+$: Leveraging Retriever Diversity for Masking-guided Faithful Fact Correction

TL;DR

This paper introduces Mask-to-Correct (M2C), a retrieval-augmented framework for fact correction that leverages diverse retrievers and ensemble methods to improve robustness and faithfulness without supervised data.

Contribution

The paper proposes a training-free, retrieval-based fact correction method and an ensemble extension to enhance robustness and reduce retrieval bias.

Findings

Achieves up to 14% improvement in SARI scores over baselines.

Effectively leverages multiple retrievers to mitigate retrieval bias.

Does not require gold evidence or supervised claim-evidence pairs.

Abstract

The rapid spread of misinformation on social media highlights the need for robust, automated fact correction frameworks. However, existing works rely on supervised learning from manually annotated claim-evidence pairs, which are scarce and prone to biases, limiting their generalization across domains. Moreover, these methods overlook semantic faithfulness in their correction process. To address these challenges, we propose Mask-to-Correct (M$_2$C), a training-free, inference-only Retrieval Augmented Generation (RAG) based framework that leverages diversity-aware masking to identify erroneous spans of claims and evaluate the faithfulness of corrections using retrieved evidence. However, the effectiveness of RAG heavily depends on the choice of retriever, which may vary across queries. To mitigate this, we further introduce M$_2$C$^+$, an ensemble-based framework that combines corrections across multiple rankers to reduce retrieval bias and improve robustness. Extensive experiments on the benchmark datasets demonstrate that our proposed frameworks consistently outperform all baselines, achieving up to 14% improvement in SARI scores, without using gold evidence.