Fig. 1: The workflow of this work. | Nature Communications

Fig. 1: The workflow of this work.

From: RSGPT: a generative transformer model for retrosynthesis planning pre-trained on ten billion datapoints

Fig. 1: The workflow of this work.The alternative text for this image may have been generated using AI.

In this work, we generated 10 billion reactions using RDChiral38, followed by pre-training based on Llama233 architecture to enhance the acquisition of chemical reaction knowledge and used reinforcement learning from artificial intelligence feedback (RLAIF) to elucidate the relationships among products, reactants, and templates, rendering RSGPT as a template-free model.

Back to article page