Explainable Multi-hop Verbal Reasoning Through Internal Monologue

Abstract

Many state-of-the-art (SOTA) language models have achieved high accuracy on several multi-hop reasoning problems.However, these approaches tend to not be interpretable because they do not make the intermediate reasoning steps explicit. Moreover, models trained on simpler tasks tend to fail when directly tested on more complex problems. We propose the Explainable multi-hop Verbal Reasoner (EVR) to solve these limitations by (a) decomposing multi-hop reasoning problems into several simple ones, and (b) using natural language to guide the intermediate reasoning hops. We implement EVR by extending the classic reasoning paradigm General Problem Solver (GPS) with a SOTA generative language model to generate subgoals and perform inference in natural language at each reasoning step. Evaluation of EVR on Clark et al. (2020)'s synthetic question answering (QA) dataset shows that EVR achieves SOTA performance while being able to generate all reasoning steps in natural language. Furthermore, EVR generalizes better than other strong methods when trained on simpler tasks or less training data (up to 35.7% and 7.7% absolute improvement respectively). 1

Interpretable Proof Generation via Iterative Backward Reasoning

“…There has been also an increasing interest in solving proof generation iteratively. EVR (Liang et al, 2021) splits the question into sub-questions, using generated intermediate texts to guide proof generation step by step. ProofWriter (Tafjord et al, 2021) shares a similar idea but uses intermediate textual conclusions instead and a more powerful T5-11B model (Raffel et al, 2020) for generation, which makes it hard to reproduce.…”

“…prediction modules once at first to predict the answer A and the strategy of the proof (refer to §3.1, where the latter one will result in different proof generation procedures. In order to improve the reasoning efficiency as well as accuracy, instead of using generated intermediate texts (Liang et al, 2021;Tafjord et al, 2021), all possible nodes (rules and facts) are represented by node embeddings in IBR. The initial state of the proof is only the representation of the question h Q , then the rest of the reasoning path will be constructed based on it.…”

“…Samples with Fail-Proof strategy differs from ones with Proof, because their proofs are usually short without sub-branches, and only consist of rules due to lacking essential supporting facts. To take the advantage of such a property distinction and extend the applicability compared to former models (Liang et al, 2021) that cannot generate proofs for Fail-Proof samples, we apply different actions in modules ( 2) and (3) depending on the output from strategy prediction.…”

“…But they lack interpretability on tracking the detailed reason for selecting each step. To make proof generation more interpretable, Proofwriter (Tafjord et al, 2021) and EVR (Liang et al, 2021) decompose complex reasoning over the question into multiple simple procedures, resulting in iterative and interpretable processes with the help of intermediate texts. Nevertheless, both of them suffer from efficiency and external errors issues.…”

“…In addition, to make IBR applicable for samples with incomplete proof paths, which are abandoned in the former backward iterative model EVR (Liang et al, 2021), we employ a proof strategy predictor to output a proof type. This prediction is then integrated into the later proof generation actions, making the process more controllable under different conditions.…”