Abstract
The accurate prediction of chemical reaction outcomes is a major challenge incomputational chemistry. Current models rely heavily on either highly specificreaction templates or template-free methods, both of which present limitations.To address these limitations, this work proposes the Broad Reaction Set (BRS),a dataset featuring 20 generic reaction templates that allow for the efficientexploration of the chemical space. Additionally, ProPreT5 is introduced, a T5model tailored to chemistry that achieves a balance between rigid templates andtemplate-free methods. ProPreT5 demonstrates its capability to generateaccurate, valid, and realistic reaction products, making it a promisingsolution that goes beyond the current state-of-the-art on the complex reactionproduct prediction task.