Cursor Upgrade Tab Model to enhance code completion quality

The AI coding platform Cursor made a major upgrade to the system-wide system-wide coding of the Tab model, focusing on reducing low-quality recommendations, improving accuracy and reducing the number of new model proposals by 21% and increasing the acceptance rate by 28%. In order to address the problem of the old model, the first consideration was to predict the acceptance of the recommendation and eventually to use the enhanced learning strategy gradient method to obtain data on the "online strategy" through the deployment of new checkpoints to users and the rapid re-training model. At present, the Tab model handles over 400 million requests per day, with positive industry reactions, and in June this year Cursor parent company financed, launched and upgraded its high-end subscription scheme。