MOCHA: A Dataset for Training and Evaluating Generative Reading Comprehension Metrics

TL;DR

This paper introduces MOCHA, a large dataset of human judgments for evaluating generative reading comprehension models, and proposes LERC, a learned metric that outperforms baselines in correlating with human scores and robustness testing.

Contribution

The paper presents MOCHA, a new benchmark dataset with human annotations, and develops LERC, a learned evaluation metric that better aligns with human judgments for generative reading comprehension.

Findings

LERC outperforms baseline metrics by 10-36 Pearson points.

LERC achieves 80% accuracy on minimal pair robustness tests.

MOCHA provides a challenging benchmark for future metric development.

Abstract

Posing reading comprehension as a generation problem provides a great deal of flexibility, allowing for open-ended questions with few restrictions on possible answers. However, progress is impeded by existing generation metrics, which rely on token overlap and are agnostic to the nuances of reading comprehension. To address this, we introduce a benchmark for training and evaluating generative reading comprehension metrics: MOdeling Correctness with Human Annotations. MOCHA contains 40K human judgement scores on model outputs from 6 diverse question answering datasets and an additional set of minimal pairs for evaluation. Using MOCHA, we train a Learned Evaluation metric for Reading Comprehension, LERC, to mimic human judgement scores. LERC outperforms baseline metrics by 10 to 36 absolute Pearson points on held-out annotations. When we evaluate robustness on minimal pairs, LERC achieves 80% accuracy, outperforming baselines by 14 to 26 absolute percentage points while leaving significant room for improvement. MOCHA presents a challenging problem for developing accurate and robust generative reading comprehension metrics.