Fecha de publicación: -- Fuente: WIPO "tomato" An online system detects and corrects errors in a catalog of items. The online system receives an indication to autonomously review an item database and retrieves exposure data for a set of items, wherein the exposure data includes data about user interactions with the items in the set. The online system inputs the exposure data to a machine learning model, which is trained to output an indication of whether an attribute associated with an item in the set is erroneous. The online system creates a prompt that includes information from the item database, images of each of the set of items associated with an erroneous attribute, and a request to suggest a correction for each erroneous attribute. The online system inputs the prompt to a generative model. The online system receives a set of suggested corrections from the generative model and updates the item database according to the suggested corrections.