I Spy With My Model’s Eye: Visual Search as a Behavioural Test for MLLMs

John Burden, Jonathan Prunty, Ben Slater, Matthieu Tehenan, Greg Davis, Lucy G Cheke

Keywords: Multimodal Language Model, Visual Search, Cognitively Inspired Evaluation, Pop out

TL;DR: We apply visual search tasks from cognitive science to frontier models --finding that existing models exhibit human-like behaviour for a number of feature types.

Abstract: Multimodal large language models (MLLMs) achieve strong performance on vision-language tasks, yet their visual processing is opaque. Most black-box evaluations measure task accuracy, but reveal little about underlying mechanisms. Drawing on cognitive psychology, we adapt classic visual search paradigms---originally developed to study human perception---to test whether MLLMs exhibit the ``pop-out'' effect, where salient visual features are detected independently of distractor set size. Using controlled experiments targeting colour, size and lighting features, we find that advanced MLLMs exhibit human-like pop-out effects in colour or size-based disjunctive (single feature) search, as well as capacity limits for conjunctive (multiple feature) search. We also find evidence to suggest that MLLMs, like humans, incorporate natural scene priors such as lighting direction into object representations. We reinforce our findings using targeted fine-tuning and mechanistic interpretability analyses. Our work shows how visual search can serve as a cognitively grounded diagnostic tool for evaluating perceptual capabilities in MLLMs.

Primary Area: other topics in machine learning (i.e., none of the above)

Submission Number: 9406