Mitigating Lost-in-Retrieval Problems in Retrieval Augmented Multi-Hop Question Answering

Rongzhi Zhu, Xiangyu Liu, Zequn Sun, Yiwei Wang, Wei Hu · 2025

Read Mitigating Lost-in-Retrieval Problems in Retrieval Augmented Multi-Hop Question Answering by Rongzhi Zhu, Xiangyu Liu, Zequn Sun, Yiwei Wang, Wei Hu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-14245.txt.

Description

Rongzhi Zhu and co-authors fix a retrieval flaw in multi step question answering in a 2025 paper. Systems that answer hard questions split them into sub questions first. Key names can vanish during that split, breaking the search chain. Answers then come out wrong even when the facts exist nearby.

Their method repairs each sub question before searching for it. Missing names are filled back in from what earlier steps found. A sentence graph supplies the passages for each step in turn. Tests on three data sets beat rival systems in accuracy and speed.

Catalog record

CREATORRongzhi Zhu, Xiangyu Liu, Zequn Sun, Yiwei Wang, Wei Hu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE47 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-14245

Browse the full library · About the Tower of Babel