Context Sensitivity Improves Human-Machine Visual Alignment

TL;DR

This paper introduces a context-sensitive similarity method for neural embeddings, significantly improving human-aligned visual task performance by adapting to environmental context.

Contribution

It presents a novel approach for context-sensitive similarity computation that enhances alignment with human visual perception in neural network models.

Findings

Achieved up to 15% accuracy improvement in odd-one-out tasks.

Improvement is consistent across original and human-aligned models.

Modeling context enhances human-like visual understanding.

Abstract

Modern machine learning models typically represent inputs as fixed points in a high-dimensional embedding space. While this approach has been proven powerful for a wide range of downstream tasks, it fundamentally differs from the way humans process information. Because humans are constantly adapting to their environment, they represent objects and their relationships in a highly context-sensitive manner. To address this gap, we propose a method for context-sensitive similarity computation from neural network embeddings, applied to modeling a triplet odd-one-out task with an anchor image serving as simultaneous context. Modeling context enables us to achieve up to a 15% improvement in odd-one-out accuracy over a context-insensitive model. We find that this improvement is consistent across both original and "human-aligned" vision foundation models.