Robustifying Multi-hop QA through Pseudo-Evidentiality Training
This addresses the issue of unreliable reasoning in multi-hop QA for AI researchers, though it is incremental as it builds on existing supervision approaches.
The paper tackles the bias problem in multi-hop question answering models, where models answer correctly without proper reasoning, by proposing a method to learn evidentiality without expensive annotations, resulting in accurate and robust performance validated on HotpotQA datasets.
This paper studies the bias problem of multi-hop question answering models, of answering correctly without correct reasoning. One way to robustify these models is by supervising to not only answer right, but also with right reasoning chains. An existing direction is to annotate reasoning chains to train models, requiring expensive additional annotations. In contrast, we propose a new approach to learn evidentiality, deciding whether the answer prediction is supported by correct evidences, without such annotations. Instead, we compare counterfactual changes in answer confidence with and without evidence sentences, to generate "pseudo-evidentiality" annotations. We validate our proposed model on an original set and challenge set in HotpotQA, showing that our method is accurate and robust in multi-hop reasoning.