PROST: Physical Reasoning of Objects through Space and Time

TL;DR

PROST introduces a comprehensive dataset to evaluate physical reasoning in language models, revealing their significant limitations and the need for models with more human-like understanding of the physical world.

Contribution

The paper presents PROST, a new dataset for probing physical reasoning in language models, and provides extensive analysis showing current models' inadequacies in this domain.

Findings

State-of-the-art models perform poorly on physical reasoning tasks.

Models are affected by answer option order and superlative inversion.

Increasing data and parameters yields minimal improvements.

Abstract

We present a new probing dataset named PROST: Physical Reasoning about Objects Through Space and Time. This dataset contains 18,736 multiple-choice questions made from 14 manually curated templates, covering 10 physical reasoning concepts. All questions are designed to probe both causal and masked language models in a zero-shot setting. We conduct an extensive analysis which demonstrates that state-of-the-art pretrained models are inadequate at physical reasoning: they are influenced by the order in which answer options are presented to them, they struggle when the superlative in a question is inverted (e.g., most <-> least), and increasing the amount of pretraining data and parameters only yields minimal improvements. These results provide support for the hypothesis that current pretrained models' ability to reason about physical interactions is inherently limited by a lack of real world experience. By highlighting these limitations, we hope to motivate the development of models with a human-like understanding of the physical world.