# MBD: A Model-Based Debiasing Framework Across User, Content, and Model Dimensions

arXiv: https://arxiv.org/abs/2603.14422 · HTML: https://mlsanity.com/paper/2603.14422/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 26 references that are in the map.
- First posted on arXiv 15 Mar 2026.

## Builds on (most cited first, 20 of 20)

- [Deep & Cross Network for Ad Click Predictions](https://mlsanity.com/paper/1708.05123/) (2017)
- [DCN V2: Improved Deep & Cross Network and Practical Lessons for Web-scale Learning to Rank Systems](https://mlsanity.com/paper/2008.13535/) (2020)
- [Recommendations as Treatments: Debiasing Learning and Evaluation](https://mlsanity.com/paper/1602.05352/) (2016)
- [Bias and Debias in Recommender System: A Survey and Future Directions](https://mlsanity.com/paper/2010.03240/) (2020)
- [Prediction-Powered Inference](https://mlsanity.com/paper/2301.09633/) (2023)
- [Model-Agnostic Counterfactual Reasoning for Eliminating Popularity Bias in Recommender System](https://mlsanity.com/paper/2010.15363/) (2020)
- [Large language model validity via enhanced conformal prediction methods](https://mlsanity.com/paper/2406.09714/) (2024)
- [Deconfounded Recommendation for Alleviating Bias Amplification](https://mlsanity.com/paper/2105.10648/) (2021)
- [Causal Inference in Recommender Systems: A Survey and Future Directions](https://mlsanity.com/paper/2208.12397/) (2022)
- [Deconfounding Duration Bias in Watch-time Prediction for Video Recommendation](https://mlsanity.com/paper/2206.06003/) (2022)
- [DVR: Micro-Video Recommendation Optimizing Watch-Time-Gain under Duration Bias](https://mlsanity.com/paper/2208.05190/) (2022)
- [Exploration in Online Advertising Systems with Deep Uncertainty-Aware Learning](https://mlsanity.com/paper/2012.02298/) (2020)
- [Unbiased Knowledge Distillation for Recommendation](https://mlsanity.com/paper/2211.14729/) (2022)
- [Alleviating Video-Length Effect for Micro-video Recommendation](https://mlsanity.com/paper/2308.14276/) (2023)
- [Conditional Quantile Estimation for Uncertain Watch Time in Short-Video Recommendation](https://mlsanity.com/paper/2407.12223/) (2024)
- [Debiasing Neural Retrieval via In-batch Balancing Regularization](https://mlsanity.com/paper/2205.09240/) (2022)
- [SWaT: Statistical Modeling of Video Watch Time through User Behavior Analysis](https://mlsanity.com/paper/2408.07759/) (2024)
- [Deep Distributional Learning with Non-crossing Quantile Network](https://mlsanity.com/paper/2504.08215/) (2025)
- [AlignPxtr: Aligning Predicted Behavior Distributions for Bias-Free Video Recommendations](https://mlsanity.com/paper/2503.06920/) (2025)
- [Epinet for Content Cold Start](https://mlsanity.com/paper/2412.04484/) (2024)

## Abstract (arXiv)

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.