Retriever-generator-verification: A novel approach to enhancing factual coherence in open-domain question answering

Shiqi Sun, Kun Zhang, Jingyuan Li, Min Yu, Kun Hou, Yuanzhuo Wang, Xueqi Cheng

Abstract: Highlights•This paper proposes the novel Retriever-Generator-Verification (RGV) framework to address factual coherence in ODQA systems.•This study tackles two main challenges in ODQA: introduction noise during retrieval and granularity error during generation.•This study introduces a comprehensive approach to ensure factual accuracy.•This study achieves significant improvements over baseline methods on widely-used datasets.