Autoregressive Reasoning over Chains of Facts with Transformers

TL;DR

This paper introduces an autoregressive inference algorithm for multi-hop explanation regeneration that effectively retrieves and combines relevant facts, outperforming previous methods in accuracy and efficiency.

Contribution

The paper presents a novel autoregressive fact selection method that improves multi-hop reasoning by conditioning on previously selected facts, enhancing performance and efficiency.

Findings

Outperforms previous state-of-the-art in precision

Reduces training time and inference complexity

Effectively handles longer fact chains in multi-hop reasoning

Abstract

This paper proposes an iterative inference algorithm for multi-hop explanation regeneration, that retrieves relevant factual evidence in the form of text snippets, given a natural language question and its answer. Combining multiple sources of evidence or facts for multi-hop reasoning becomes increasingly hard when the number of sources needed to make an inference grows. Our algorithm copes with this by decomposing the selection of facts from a corpus autoregressively, conditioning the next iteration on previously selected facts. This allows us to use a pairwise learning-to-rank loss. We validate our method on datasets of the TextGraphs 2019 and 2020 Shared Tasks for explanation regeneration. Existing work on this task either evaluates facts in isolation or artificially limits the possible chains of facts, thus limiting multi-hop inference. We demonstrate that our algorithm, when used with a pre-trained transformer model, outperforms the previous state-of-the-art in terms of precision, training time and inference efficiency.