Researchers at EPFL have created MiCRo (Mixture of Cognitive Reasoners), a new large language model structured like the human brain with four specialized modules, presented at the International Conference on Learning Representations. Developed by the NLP Lab and the NeuroAI Lab within EPFL’s School of Computer and Communication Sciences and School of Life Sciences, MiCRo is divided into four experts: language, logic, social reasoning and world knowledge. Within each layer, a router can send each word of a sentence to the best-equipped expert, allowing a single sentence to be processed by several experts at once. For a prompt mixing arithmetic with social nuance, MiCRo routes numbers to the logic expert and social cues to the social reasoning expert, making it easier to see how the model reaches answers and letting users steer behavior by amplifying or suppressing an expert directly in the architecture. The team, led by PhD candidate Badr AlKhamissi, worked with neuroscientist Greta Tuckute of Harvard and MIT, applying methods used to map brain activity in humans to identify the model’s experts.