This is the source code of the KDD'25 accepted paper Multi-task Offline Reinforcement Learning for Online Advertising in Recommender Systems.

Here, we provide codes for the proposed MTORL and baseline models. Moreover, we provide two CSV data files named Kuairand_squence and Criteo_squence, both demonstrating a small part of the processed datasets. We do not provide raw datasets here due to the space limit, and we have attached the links to the dataset below. Furthermore, we will publicize the private commercial dataset to foster research on this critical topic.

The link to KuaiRand: https://kuairand.com

The link to Criteo: http://ailab.criteo.com/criteo-attribution-modeling-bidding-dataset