Generate-then-Ground in Retrieval-Augmented Generation for Multi-hop Question Answering

TL;DR

The paper introduces GenGround, a novel generate-then-ground framework that enhances multi-hop question answering by iteratively generating and grounding answers in retrieved documents, outperforming existing methods.

Contribution

It proposes a new generate-then-ground approach for multi-hop QA, combining parametric and external knowledge, with an instructional grounding distillation for smaller models.

Findings

GenGround achieves superior performance on four datasets.

The iterative generate-then-ground process improves answer accuracy.

Instructional grounding distillation enables smaller models to perform effectively.

Abstract

Multi-Hop Question Answering (MHQA) tasks present a significant challenge for large language models (LLMs) due to the intensive knowledge required. Current solutions, like Retrieval-Augmented Generation, typically retrieve potential documents from an external corpus to read an answer. However, the performance of this retrieve-then-read paradigm is constrained by the retriever and the inevitable noise in the retrieved documents. To mitigate these challenges, we introduce a novel generate-then-ground (GenGround) framework, synergizing the parametric knowledge of LLMs and external documents to solve a multi-hop question. GenGround empowers LLMs to alternate two phases until the final answer is derived: (1) formulate a simpler, single-hop question and directly generate the answer; (2) ground the question-answer pair in retrieved documents, amending any wrong predictions in the answer. We also propose an instructional grounding distillation method to generalize our method into smaller models. Extensive experiments conducted on four datasets illustrate the superiority of our method.