PROST: Physical Reasoning about Objects through Space and Time

By Stephane T Aroca-Ouellette, Cory Paik, Alessandro Roncone, Katharina Kann. 2021

57 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Natural Language Processing, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

In Evaluation & Benchmarks

Citations in the corpus

GPT-NeoX-20B: An Open-Source Autoregressive Language Model (2022, neutral): “Due to space constraints a representative subset of the results are shown here, with the rest in Appendix E. Natural Language Tasks We evaluate our model on a diverse collection of standard language model evaluation datasets: ANLI (Nie et al., 2020), ARC (Clark et al., 2018), HeadQA (English) (Vilar”
The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Only (2023, neutral): “…(Liu et al., 2021) Multiple-choice question answering 25.0 ✓ HeadQA (Vilares & G´omez-Rodr´ıguez, 2019) Multiple-choice question answering 20.0 ✓ MathQA (Amini et al., 2019) Multiple-choice question answering 20.0 ✓ PROST (Aroca-Ouellette et al., 2021) Paraphrase identification 50.0 ✓ PubMedQA…”
Natural Language Processing with Commonsense Knowledge: A Survey (2021, background): “• PROST (Physical Reasoning about Objects through Space and Time) [54] is a probing dataset designed to evaluate the capability of language models to reason about physical interactions in the real world.”
UnifiedQA-v2: Stronger Generalization via Broader Cross-Format Training (2022, background): “Small 79.4 80.7 67.6 66.4 77.1 78.0 45.5 53.7 55.9 55.2 50.4 50.0 42.9 47.1 35.9 37.0 80.0 72.8 59.4 60.1 0.7 60 Base 88.2 86.8 78.1 74.7 82.5 82.4 49.8 60.4 70.3 69.0 59.0 59.4 53.0 58.5 42.4 45.7 86.9 86.3 67.8 69.2 1.5 50 Large 91.1 90.8 85.9 86.3 86.1 85.8 51.6 63.2 78.0 78.8 68.2 71.8 65.9 70.5”
A Comprehensive Overview of Large Language Models (2023, unknown): “Physical Knowledge and World Understanding PIQA [275], TriviaQA [276], ARC [277], ARC-Easy [277], ARC-Challenge [277], PROST [278], OpenBookQA [279], WebNLG [280], DogWhistle Insider & Outsider [281] Contextual Language Understanding RACE [282], RACE-Middle [282], RACE-High [282], QuAC [283], Strate”