Break, Perturb, Build: Automatic Perturbation of Reasoning Paths Through Question Decomposition

TL;DR

This paper introduces the BPB framework for automatically generating perturbed question-answer pairs by decomposing and modifying reasoning steps, enabling robust evaluation and training of reading comprehension models.

Contribution

The novel BPB method automates reasoning path perturbation, creating high-quality evaluation and training data without human effort, and facilitates detailed model analysis.

Findings

Models perform significantly worse on perturbed examples.

BPB-generated data improves model robustness.

Perturbation analysis reveals model strengths and weaknesses.

Abstract

Recent efforts to create challenge benchmarks that test the abilities of natural language understanding models have largely depended on human annotations. In this work, we introduce the "Break, Perturb, Build" (BPB) framework for automatic reasoning-oriented perturbation of question-answer pairs. BPB represents a question by decomposing it into the reasoning steps that are required to answer it, symbolically perturbs the decomposition, and then generates new question-answer pairs. We demonstrate the effectiveness of BPB by creating evaluation sets for three reading comprehension (RC) benchmarks, generating thousands of high-quality examples without human intervention. We evaluate a range of RC models on our evaluation sets, which reveals large performance gaps on generated examples compared to the original data. Moreover, symbolic perturbations enable fine-grained analysis of the strengths and limitations of models. Last, augmenting the training data with examples generated by BPB helps close the performance gaps, without any drop on the original data distribution.