This page organizes papers related to artificial intelligence published around the world. This page is summarized using Google Gemini and is operated on a non-profit basis. The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.
MTRec: Learning to Align with User Preferences via Mental Reward Models
Created by
Haebom
Author
Mengchen Zhao, Yifan Gao, Yaqing Hou, Xiangyang Li, Pengjie Gu, Zhenhua Dong, Ruiming Tang, Yi Cai
Outline
This paper proposes MTRec, a novel sequential recommendation framework that overcomes the limitations of implicit feedback and improves the performance of recommendation systems by identifying users' true preferences. MTRec introduces a mental reward model that quantifies user satisfaction and uses distributed inverse reinforcement learning to train it. The learned mental reward model guides the recommendation model to better match users' true preferences. Experimental results and application on a real-world short video platform demonstrate the effectiveness of MTRec.
Takeaways, Limitations
•
Takeaways:
◦
We present a novel framework to improve the performance of recommendation systems by addressing the problems of implicit feedback.
◦
Introduction of a mental reward model that quantifies user satisfaction and utilizes distributed inverse reinforcement learning.
◦
Proving practicality through application in actual industrial environments.
◦
It shows consistent performance improvement when applied to various recommendation models.
•
Limitations:
◦
Dependence on the accuracy of the mental reward model.
◦
Computational cost required to train and apply the model.
◦
Further research is needed to determine the generalizability of the experimental results to specific platforms (short videos).
