Abstract: An online concierge system compensates pickers who fulfill orders including one or more items based in part on weights of the items included in an order. Because the online concierge system does not physically possess the items that are obtained, the online concierge system cannot directly weigh the items and weights specified for items in a catalog from a retailer may be inaccurate. To more accurately determine weights of items, the online concierge system trains a weight prediction model to estimate an item's weight from attributes of the item and uses the output of the weight prediction model to determine compensation to a picker. The weight prediction model may output a predicted weight of an item or a classification of the item as heavy or light. Where discrepancies are found between a predicted weight and the catalog weight of an item, additional information about the item is obtained.

Abstract: An online system validates item updates using a machine-learning model to identify item updates that need independent review. The online system maintains an item database that has item entries for items on the online system. The online system receives item updates from an item update system and applies an error prediction model to the item updates to generate an error likelihood score for each item update. The online system samples a subset of the item updates based on the error likelihood scores and passes these sampled item updates to a human reviewer system. The human reviewer system labels each of the sampled item updates with an error label indicating whether the corresponding item update is actually erroneous. The online system determines whether to update the item database with the full set of received item updates based on the error labels.