Large Language ModelsLAVOIR: Gaining Defending Information to Artificial Intelligence ModelsA new approach teaches when and what information should be requested to decide models working on one pass.