Meta-learning as a principle for human-like visual representations
For cognitive science and AI, this work proposes that meta-learning, rather than static pretraining, is key to aligning neural network representations with human visual cognition.
Meta-learned representations, trained across thousands of tasks without human data, better predict human similarity judgments, semantic rule learning, and high-level visual cortex activity compared to pretrained base encoders, suggesting that the flexibility of human visual representations arises from meta-learning.
The structure of human visual representations underpins our capacity for adaptive behaviour. While pretrained neural networks model human visual representations with unprecedented success, a large discrepancy remains. We propose one reason: these networks optimise a single fixed objective, whereas human representations must support open-ended tasks. We hypothesise this flexibility arises from meta-learning (learning to learn), a pressure shaping representations to acquire new tasks from few observations. To test this, we train a sequence model, without any supervision from human data, across thousands of semantically rich tasks mapping images to high-level concepts. Compared to their pretrained base encoders, meta-learned representations better predict human similarity judgements, semantic rule learning, and high-level visual cortex. Behavioural gains depend on disentangled, high-level task distributions, while brain alignment is driven primarily by the learning-to-learn pressure. Our results suggest the flexibility of human visual representations reflects the functional demand to learn new semantic relationships on the fly.