Researchers at the Swiss Federal Institute of Technology (EPFL) have unveiled MiCRo, a novel large language model architected to mirror human cognitive processing, offering a transparent alternative to traditional opaque AI systems. Developed jointly by the university’s NLP and NeuroAI Labs, the model was presented at the International Conference on Learning Representations (ICLR) 2026 and is available on arXiv. Unlike conventional language models that process text through monolithic parameter stacks, MiCRo partitions each computational layer into four distinct modules modeled after specialized brain regions: language, logic, social reasoning, and world knowledge. A dynamic routing mechanism directs individual words or concepts to the most relevant expert at every stage. For example, a prompt requiring financial division alongside social nuance automatically channels numerical data to the logic module while contextual cues activate the social reasoning module. This modular design replaces generic prompting techniques with direct architectural intervention, allowing developers to amplify or suppress specific cognitive pathways without altering training data or prompt structure. The project synthesizes artificial intelligence with cognitive neuroscience. Lead researcher Badr AlKhamissi collaborated with neuroscientist Greta Tuckute from Harvard and MIT to map human neural responses to varied cognitive tasks. By exposing the model to complex problem sets, the team calibrated each module to mirror the activation patterns observed in human subjects, enabling the architecture to self-identify which experts handle specific computational demands. MiCRo fundamentally shifts how artificial reasoning is observed and controlled. The explicit routing of information provides real-time visibility into the model’s decision-making process, effectively dismantling the traditional AI black box. Furthermore, the research establishes a reciprocal relationship between disciplines: while neuroscience inspired the model’s architecture, MiCRo now serves as a computational proxy to analyze how different brain regions contribute to human cognition. This bidirectional framework promises more interpretable AI systems and new methodologies for cognitive research, marking a significant step toward transparent, cognitively aligned artificial intelligence.