Do Multi-hop Readers Dream of Reasoning Chains?
This addresses the problem of multi-hop reasoning in QA systems, showing incremental progress by highlighting the need for improved models.
The paper systematically analyzes whether providing full reasoning chains improves multi-hop question answering models, finding limited improvements (up to 5.8% error reduction) with existing methods, but a co-matching-based method achieves 13.1% error reduction, indicating potential for better reasoning.
General Question Answering (QA) systems over texts require the multi-hop reasoning capability, i.e. the ability to reason with information collected from multiple passages to derive the answer. In this paper we conduct a systematic analysis to assess such an ability of various existing models proposed for multi-hop QA tasks. Specifically, our analysis investigates that whether providing the full reasoning chain of multiple passages, instead of just one final passage where the answer appears, could improve the performance of the existing QA models. Surprisingly, when using the additional evidence passages, the improvements of all the existing multi-hop reading approaches are rather limited, with the highest error reduction of 5.8% on F1 (corresponding to 1.3% absolute improvement) from the BERT model. To better understand whether the reasoning chains could indeed help find correct answers, we further develop a co-matching-based method that leads to 13.1% error reduction with passage chains when applied to two of our base readers (including BERT). Our results demonstrate the existence of the potential improvement using explicit multi-hop reasoning and the necessity to develop models with better reasoning abilities.