MBD: A Model-Based Debiasing Framework Across User, Content, and Model Dimensions
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908887749230592 |
|---|---|
| author | Li, Yuantong Yuan, Lei Zheng, Zhihao Wu, Weimiao Liu, Songbin Lee, Jeong Min Aydin, Ali Selman Deng, Shaofeng Chen, Junbo Zhang, Xinyi Xia, Hongjing Fieldman, Sam Kosko, Matthew Fu, Wei Zhang, Du Yang, Peiyu Chung, Albert Jin Qiu, Xianlei Yu, Miao Teng, Zhongwei Chen, Hao Baek, Sunny Tang, Hui Lv, Yang Wang, Renze Wang, Qifan Li, Zhan Xu, Tiantian Wu, Peng Liu, Ji |
| author_facet | Li, Yuantong Yuan, Lei Zheng, Zhihao Wu, Weimiao Liu, Songbin Lee, Jeong Min Aydin, Ali Selman Deng, Shaofeng Chen, Junbo Zhang, Xinyi Xia, Hongjing Fieldman, Sam Kosko, Matthew Fu, Wei Zhang, Du Yang, Peiyu Chung, Albert Jin Qiu, Xianlei Yu, Miao Teng, Zhongwei Chen, Hao Baek, Sunny Tang, Hui Lv, Yang Wang, Renze Wang, Qifan Li, Zhan Xu, Tiantian Wu, Peng Liu, Ji |
| contents | Modern recommendation systems rank candidates by aggregating multiple behavioral signals through a value model. However, many commonly used signals are inherently affected by heterogeneous biases. For example, watch time naturally favors long-form content, loop rate favors short - form content, and comment probability favors videos over images. Such biases introduce two critical issues: (1) value model scores may be systematically misaligned with users' relative preferences - for instance, a seemingly low absolute like probability may represent exceptionally strong interest for a user who rarely engages; and (2) changes in value modeling rules can trigger abrupt and undesirable ecosystem shifts. In this work, we ask a fundamental question: can biased behavioral signals be systematically transformed into unbiased signals, under a user - defined notion of ``unbiasedness'', that are both personalized and adaptive? We propose a general, model-based debiasing (MBD) framework that addresses this challenge by augmenting it with distributional modeling. By conditioning on a flexible subset of features (partial feature set), we explicitly estimate the contextual mean and variance of the engagement distribution for arbitrary cohorts (e.g., specific video lengths or user regions) directly alongside the main prediction. This integration allows the framework to convert biased raw signals into unbiased representations, enabling the construction of higher-level, calibrated signals (such as percentiles or z - scores) suitable for the value model. Importantly, the definition of unbiasedness is flexible and controllable, allowing the system to adapt to different personalization objectives and modeling preferences. Crucially, this is implemented as a lightweight, built-in branch of the existing MTML ranking model, requiring no separate serving infrastructure. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_14422 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MBD: A Model-Based Debiasing Framework Across User, Content, and Model Dimensions Li, Yuantong Yuan, Lei Zheng, Zhihao Wu, Weimiao Liu, Songbin Lee, Jeong Min Aydin, Ali Selman Deng, Shaofeng Chen, Junbo Zhang, Xinyi Xia, Hongjing Fieldman, Sam Kosko, Matthew Fu, Wei Zhang, Du Yang, Peiyu Chung, Albert Jin Qiu, Xianlei Yu, Miao Teng, Zhongwei Chen, Hao Baek, Sunny Tang, Hui Lv, Yang Wang, Renze Wang, Qifan Li, Zhan Xu, Tiantian Wu, Peng Liu, Ji Machine Learning Artificial Intelligence Information Retrieval Modern recommendation systems rank candidates by aggregating multiple behavioral signals through a value model. However, many commonly used signals are inherently affected by heterogeneous biases. For example, watch time naturally favors long-form content, loop rate favors short - form content, and comment probability favors videos over images. Such biases introduce two critical issues: (1) value model scores may be systematically misaligned with users' relative preferences - for instance, a seemingly low absolute like probability may represent exceptionally strong interest for a user who rarely engages; and (2) changes in value modeling rules can trigger abrupt and undesirable ecosystem shifts. In this work, we ask a fundamental question: can biased behavioral signals be systematically transformed into unbiased signals, under a user - defined notion of ``unbiasedness'', that are both personalized and adaptive? We propose a general, model-based debiasing (MBD) framework that addresses this challenge by augmenting it with distributional modeling. By conditioning on a flexible subset of features (partial feature set), we explicitly estimate the contextual mean and variance of the engagement distribution for arbitrary cohorts (e.g., specific video lengths or user regions) directly alongside the main prediction. This integration allows the framework to convert biased raw signals into unbiased representations, enabling the construction of higher-level, calibrated signals (such as percentiles or z - scores) suitable for the value model. Importantly, the definition of unbiasedness is flexible and controllable, allowing the system to adapt to different personalization objectives and modeling preferences. Crucially, this is implemented as a lightweight, built-in branch of the existing MTML ranking model, requiring no separate serving infrastructure. |
| title | MBD: A Model-Based Debiasing Framework Across User, Content, and Model Dimensions |
| topic | Machine Learning Artificial Intelligence Information Retrieval |
| url | https://arxiv.org/abs/2603.14422 |