The reasoning systems are based on a technology called large language models, or L.L.M.s. To build reasoning systems, companies put L.L.M.s through an additional process called reinforcement learning.