Can Language Models Understand Physical Concepts?

TL;DR

This paper investigates whether language models can understand physical concepts, introducing a benchmark VEC, analyzing their performance, and proposing a method to transfer embodied knowledge from vision-language models to improve understanding.

Contribution

The paper introduces the VEC benchmark for physical concept understanding, analyzes LM performance, and proposes a distillation method to transfer embodied knowledge from vision-language models.

Findings

Scaling up LMs improves understanding of some visual concepts.

Vision-augmented LMs like CLIP and BLIP achieve human-level understanding of embodied concepts.

A distillation method transfers embodied knowledge, boosting LM performance significantly.

Abstract

Language models~(LMs) gradually become general-purpose interfaces in the interactive and embodied world, where the understanding of physical concepts is an essential prerequisite. However, it is not yet clear whether LMs can understand physical concepts in the human world. To investigate this, we design a benchmark VEC that covers the tasks of (i) Visual concepts, such as the shape and material of objects, and (ii) Embodied Concepts, learned from the interaction with the world such as the temperature of objects. Our zero (few)-shot prompting results show that the understanding of certain visual concepts emerges as scaling up LMs, but there are still basic concepts to which the scaling law does not apply. For example, OPT-175B performs close to humans with a zero-shot accuracy of 85\% on the material concept, yet behaves like random guessing on the mass concept. Instead, vision-augmented LMs such as CLIP and BLIP achieve a human-level understanding of embodied concepts. Analysis indicates that the rich semantics in visual representation can serve as a valuable source of embodied knowledge. Inspired by this, we propose a distillation method to transfer embodied knowledge from VLMs to LMs, achieving performance gain comparable with that by scaling up the parameters of LMs 134x. Our dataset is available at \url{https://github.com/TobiasLee/VEC}