Relative Drawing Identification Complexity is Invariant to Modality in Vision-Language Models

TL;DR

This paper investigates whether the complexity of teaching vision-language models is consistent across different modalities, finding that concept simplicity appears invariant to whether data is represented as images or coordinate traces.

Contribution

It introduces a method to compare teaching complexity across modalities and demonstrates that concept simplicity is an inherent property, not dependent on data representation.

Findings

Image-based representations require fewer segments for teaching.

Higher accuracy is achieved with image-based data.

Concept complexity is similar across modalities, indicating modality invariance.

Abstract

Large language models have become multimodal, and many of them are said to integrate their modalities using common representations. If this were true, a drawing of a car as an image, for instance, should map to a similar area in the latent space as a textual description of the strokes that form the drawing. To explore this in a black-box access regime to these models, we propose the use of machine teaching, a theory that studies the minimal set of examples a teacher needs to choose so that the learner captures the concept. In this paper, we evaluate the complexity of teaching vision-language models a subset of objects in the Quick, Draw! dataset using two presentations: raw images as bitmaps and trace coordinates in TikZ format. The results indicate that image-based representations generally require fewer segments and achieve higher accuracy than coordinate-based representations. But, surprisingly, the teaching size usually ranks concepts similarly across both modalities, even when controlling for (a human proxy of) concept priors, suggesting that the simplicity of concepts may be an inherent property that transcends modality representations.