QASC: A Dataset for Question Answering via Sentence Composition

TL;DR

The paper introduces QASC, a multi-hop question answering dataset that emphasizes complex fact retrieval and composition, with annotated supporting facts and a two-step reasoning approach that improves model performance but still lags behind humans.

Contribution

QASC is the first dataset with annotated facts and compositions for multi-hop reasoning, enabling better training and evaluation of reasoning models.

Findings

Proposed a two-step retrieval and reasoning approach.

Achieved 11% improvement over state-of-the-art models.

Model performance is still 20% below human accuracy.

Abstract

Composing knowledge from multiple pieces of texts is a key challenge in multi-hop question answering. We present a multi-hop reasoning dataset, Question Answering via Sentence Composition(QASC), that requires retrieving facts from a large corpus and composing them to answer a multiple-choice question. QASC is the first dataset to offer two desirable properties: (a) the facts to be composed are annotated in a large corpus, and (b) the decomposition into these facts is not evident from the question itself. The latter makes retrieval challenging as the system must introduce new concepts or relations in order to discover potential decompositions. Further, the reasoning model must then learn to identify valid compositions of these retrieved facts using common-sense reasoning. To help address these challenges, we provide annotation for supporting facts as well as their composition. Guided by these annotations, we present a two-step approach to mitigate the retrieval challenges. We use other multiple-choice datasets as additional training data to strengthen the reasoning model. Our proposed approach improves over current state-of-the-art language models by 11% (absolute). The reasoning and retrieval problems, however, remain unsolved as this model still lags by 20% behind human performance.