HellaSwag: Can a Machine Really Finish Your Sentence?

TL;DR

HellaSwag is a challenging new dataset for commonsense inference that exposes the limitations of current models, even those reaching near-human performance levels, by using adversarial filtering to create difficult examples.

Contribution

The paper introduces HellaSwag, a new adversarially filtered dataset that significantly increases the difficulty of commonsense inference tasks for state-of-the-art models.

Findings

Models perform below 48% accuracy on HellaSwag.

Humans achieve over 95% accuracy on the dataset.

Adversarial filtering creates robust, challenging examples for NLP models.

Abstract

Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as "A woman sits at a piano," a machine must select the most likely followup: "She sets her fingers on the keys." With the introduction of BERT, near human-level performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even state-of-the-art models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), state-of-the-art models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machine-generated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical 'Goldilocks' zone wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges.