# Exploring Graph-structured Passage Representation for Multi-hop Reading Comprehension with Graph Neural Networks

Source: [Exploring Graph-structured Passage Representation for Multi-hop Reading Comprehension with Graph Neural Networks](<https://www.sota2.com/research/paper/68fa64c91e209131cea06a40>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-04T03:02:01.381Z

Authors: Linfeng Song, Zhiguo Wang, Mo Yu, Yue Zhang, Radu Florian, Daniel Gildea

Published: 2018-09-06T15:18:14.000Z

## Abstract

Multi-hop reading comprehension focuses on one type of factoid question, where a system needs to properly integrate multiple pieces of evidence to correctly answer a question. Previous work approximates global evidence with local coreference information, encoding coreference chains with DAG-styled GRU layers within a gated-attention reader. However, coreference is limited in providing information for rich inference. We introduce a new method for better connecting global evidence, which forms more complex graphs compared to DAGs. To perform evidence integration on our graphs, we investigate two recent graph neural networks, namely graph convolutional network (GCN) and graph recurrent network (GRN). Experiments on two standard datasets show that richer global information leads to better answers. Our method performs better than all published results on these datasets.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Multi-hop Question Answering on ComplexWebQuestions v1.1 (dev)](<https://www.sota2.com/research/sota/multi-hop-question-answering-on-complexwebquestions-v1-1-dev.md>): Multi-hop Question Answering; ComplexWebQuestions v1.1 (dev); Accuracy: 33.2.

- [Multi-hop Question Answering on ComplexWebQuestions 1.1 (test)](<https://www.sota2.com/research/sota/multi-hop-question-answering-on-complexwebquestions-1-1-test.md>): Multi-hop Question Answering; ComplexWebQuestions 1.1 (test); Accuracy: 30.1.

- [Multi-hop Reading Comprehension on WikiHop unmasked (dev)](<https://www.sota2.com/research/sota/multi-hop-reading-comprehension-on-wikihop-unmasked-dev.md>): Multi-hop Reading Comprehension; WikiHop unmasked (dev); Accuracy: 62.8.

- [Multi-hop Reading Comprehension on WikiHop unmasked (test)](<https://www.sota2.com/research/sota/multi-hop-reading-comprehension-on-wikihop-unmasked-test.md>): Multi-hop Reading Comprehension; WikiHop unmasked (test); Accuracy: 65.4.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/1809.02040>)

[PDF](<https://arxiv.org/pdf/1809.02040.pdf>)