Human-like Object Grouping in Self-supervised Vision Transformers
This work addresses the gap in understanding human-like object grouping in vision models for researchers in computer vision and cognitive science, though it is incremental in building on existing psychophysics and model analysis methods.
The paper tackled the problem of understanding how well self-supervised vision models align with human object perception by introducing a behavioral benchmark with over 1000 trials, finding that transformer-based models trained with DINO showed the strongest performance in predicting human reaction times. They further demonstrated that stronger object-centric structure in model representations predicts human segmentation behavior more accurately and that matching Gram matrices through distillation improves alignment.
Vision foundation models trained with self-supervised objectives achieve strong performance across diverse tasks and exhibit emergent object segmentation properties. However, their alignment with human object perception remains poorly understood. Here, we introduce a behavioral benchmark in which participants make same/different object judgments for dot pairs on naturalistic scenes, scaling up a classical psychophysics paradigm to over 1000 trials. We test a diverse set of vision models using a simple readout from their representations to predict subjects' reaction times. We observe a steady improvement across model generations, with both architecture and training objective contributing to alignment, and transformer-based models trained with the DINO self-supervised objective showing the strongest performance. To investigate the source of this improvement, we propose a novel metric to quantify the object-centric component of representations by measuring patch similarity within and between objects. Across models, stronger object-centric structure predicts human segmentation behavior more accurately. We further show that matching the Gram matrix of supervised transformer models, capturing similarity structure across image patches, with that of a self-supervised model through distillation improves their alignment with human behavior, converging with the prior finding that Gram anchoring improves DINOv3's feature quality. Together, these results demonstrate that self-supervised vision models capture object structure in a behaviorally human-like manner, and that Gram matrix structure plays a role in driving perceptual alignment.