Human-like cognitive generalization for large models via mental representation-guided supervision

Recent advancements in deep neural networks (DNNs), particularly large-scale language models, have demonstrated remarkable capabilities in image and natural language understanding. Although scaling up model parameters with increasing volume of training data has progressively improved DNN capabilities, achieving complex cognitive abilities—such as understanding abstract concepts, reasoning, and adapting to novel scenarios, which are intrinsic to human cognition—remains a major challenge. In this study, we show that mental representation-guided supervised learning, utilizing a small set of brain signals, can effectively transfer human conceptual structures to DNNs, significantly enhancing their comprehension of abstract and even unseen concepts. Experimental results further indicate that the enhanced cognitive capabilities lead to substantial performance gains in challenging tasks, including few-shot/zero-shot learning and out-of-distribution recognition, while also yielding highly interpretable concept representations. These findings highlight that mental representation-guided supervision can effectively augment the complex cognitive abilities of large models, offering a promising pathway toward developing more human-like cognitive abilities in artificial systems. It remains unclear how best to equip large models with human-like conceptual understanding. Here, the authors demonstrate that guiding networks learning with human mental representations can effectively transfer conceptual structures to networks, leading to improvements on diverse tasks.