Researchers from the Chinese Academy of Sciences have developed a breakthrough neural network framework called CATS Net, which successfully replicates the human capacity for forming, understanding, and communicating concepts. Unlike current artificial intelligence systems that struggle with abstract reasoning, this new model mimics the unique human ability to extract high-level concepts from sensory experiences, think within a conceptual space, and reconstruct those concepts into perceptions. The framework, introduced by a team from the Institute of Automation at the Chinese Academy of Sciences, consists of two primary components: a Concept Abstraction (CA) module and a Task Solving (TS) module. In visual tasks, the CA module automatically compresses high-dimensional visual inputs into compact, low-dimensional "concept vectors." These vectors function as dynamic keys that, through a hierarchical gating mechanism, generate switch signals to regulate the neural activity within the TS module. This process allows the system to efficiently and flexibly guide the execution of specific visual perception tasks. A defining feature of CATS Net is its ability to autonomously generate a vast array of new concepts based on environmental interactions, thereby constructing its own conceptual space. Furthermore, when concept spaces generated by different networks are aligned, the system can transfer knowledge directly through these concept vectors. These capabilities effectively simulate the human cognitive processes of concept generation, comprehension, and communication. To validate the model's biological fidelity, the research team compared the spontaneously formed concept representations of CATS Net with human conceptual spaces and neural activity data. Functional magnetic resonance imaging (fMRI) analyses revealed that the concept space generated by CATS Net aligns with psychological semantic models of human cognition. Moreover, its representation patterns correlate with activity in the ventral occipitotemporal cortex, a region of the human brain responsible for visual semantic representation. Additionally, the dynamic gating mechanism of the CA module matches the activity patterns of the semantic control network involved in concept extraction and manipulation within the human brain. These findings indicate that CATS Net not only simulates human concept cognition at a functional level but also elucidates the computational principles underlying human concept formation and understanding. This work provides a foundational model for comprehending human conceptual cognition and paves the way for the development of next-generation intelligent systems capable of human-like conceptual intelligence and application. The research, supported by the National Natural Science Foundation of China and various projects from the Chinese Academy of Sciences, has been published online in Nature Computational Science. This significant advancement marks a crucial step toward bridging the gap between artificial intelligence and human-like reasoning, offering a robust computational model for the future of AI.