Debiased Recommendation Beyond the Positive Propensity Assumption
sigir26-112 · Full research paper · Yanghao Xiao, Hao Wang, Xiang Li, Qian Zou, Cheng Bing, Wei Lin, Haoxuan Li, Zhouchen Lin
Popularity Bias and Debiasing in Rec · 50% CTR and Conversion Rate Prediction · 25% Popularity Bias and Long-Tail Debias · 25%
Abstract
Post-click conversion rate (CVR) prediction is a central task in recommender systems, yet selection bias creates a severe distributional gap between the clicked training samples and the entire inference space. To address selection bias, propensity-based methods such as inverse propensity scoring (IPS) and doubly robust (DR) have been adopted, which aim to estimate the unbiased learning objective from biased training samples. However, these approaches assume strictly positive propensities, implying every user-item pair has a nonzero probability of interaction. In practice, such positivity assumption maybe violated, for example, in food-delivery platforms, some restaurants located more than 10 kilometers away will be blocked for recommendation. In this study, we theoretically show that when such zero-propensity samples, termed extrapolation samples exist, both IPS and DR estimators become biased. To overcome this limitation, we propose ExtraDebias method, which enables debiased recommendation in both non-extrapolation and extrapolation samples. Specifically, we first train a propensity model to identify extrapolation samples with extremely small propensity estimates, then estimate their pseudo-label intervals, and derive an upper bound of the learning objective for extrapolation samples. By minimizing the derived upper bound, debiased learning on extrapolation samples is ensured, while unbiased learning on non-extrapolation samples is achieved by standard IPS. Experiments on four real-world offline datasets and one online A/B test show that ExtraDebias effectively minimizes prediction errors on extrapolation samples and achieves optimal performance.
Assigned reviewers
| # | Reviewer | Match | Why (closest evidence publications in the taxonomy) | |
|---|---|---|---|---|
| 1 | Yifan Wang 0017 | 69 | 83 Popularity-Aware Alignment and Contrast for Mitigating Popularity Bias (2024) Popularity Bias and Debiasing in Recommendation 83 Graph-Structured Driven Dual Adaptation for Mitigating Popularity Bias (2026) Popularity Bias and Debiasing in Recommendation 75 Improving Long-tail User CTR Prediction via Hierarchical Distribution Alignment (2025) Bias Mitigation Methods |
|
| 2 | Miaomiao Cai 0001 | 61 | 85 Mitigating Recommendation Biases via Group-Alignment and Global-Uniformity in Representation Learning (2024) Popularity Bias and Debiasing in Recommendation 85 Graph-Structured Driven Dual Adaptation for Mitigating Popularity Bias (2026) Popularity Bias and Debiasing in Recommendation 83 Popularity-Aware Alignment and Contrast for Mitigating Popularity Bias (2024) Popularity Bias and Debiasing in Recommendation |
|
| 3 | Zechun Niu | 56 | 70 Distributionally Robust Optimization for Unbiased Learning to Rank (2025) Unbiased Learning to Rank 56 Addressing Personalized Bias for Unbiased Learning to Rank (2025) Unbiased Learning to Rank 38 Investigating the Robustness of Counterfactual Learning to Rank Models: A Reproducibility Study (2025) Unbiased Learning to Rank |
Area chair
| Area chair | Match | |
|---|---|---|
| Xin Xin 0003 | 49 |
Area-chair candidates (4)
| Xin Xin 0003 assigned | 49 | 10 papers | |
| Yanan Niu | 48 | 15 papers | |
| Kun Gai COI: co-author | 46 | 5 papers | |
| Xiaoyi Zeng COI: co-author | 44 | 11 papers |
Candidates from the panel ranked by match score
| # | Reviewer | Match | Load | Why | |
|---|---|---|---|---|---|
| 1 | Yifan Wang 0017 assigned | 69 | 3 | 83 Popularity-Aware Alignment and Contrast for Mitigating Popularity Bias (2024) Popularity Bias and Debiasing in Recommendation 83 Graph-Structured Driven Dual Adaptation for Mitigating Popularity Bias (2026) Popularity Bias and Debiasing in Recommendation |
|
| 2 | Miaomiao Cai 0001 assigned | 61 | 3 | 85 Mitigating Recommendation Biases via Group-Alignment and Global-Uniformity in Representation Learning (2024) Popularity Bias and Debiasing in Recommendation 85 Graph-Structured Driven Dual Adaptation for Mitigating Popularity Bias (2026) Popularity Bias and Debiasing in Recommendation |
|
| 3 | Seongmin Park 0002 | 58 | 3 | 85 uCTRL: Unbiased Contrastive Representation Learning via Alignment and Uniformity for Collaborative Filtering (2023) Popularity Bias and Debiasing in Recommendation 84 Bilateral Self-unbiased Learning from Biased Implicit Feedback (2022) Popularity Bias and Debiasing in Recommendation |
|
| 4 | Zechun Niu assigned | 56 | 3 | 70 Distributionally Robust Optimization for Unbiased Learning to Rank (2025) Unbiased Learning to Rank 56 Addressing Personalized Bias for Unbiased Learning to Rank (2025) Unbiased Learning to Rank |
|
| 5 | Junwei Pan COI: co-author | 55 | 3 | 55 Ads Recommendation in a Collapsed and Entangled World (2024) CTR and Conversion Rate Prediction 47 Understanding the Ranking Loss for Recommendation with Sparse User Feedback (2024) CTR and Conversion Rate Prediction |
|
| 6 | Sunhao Dai | 53 | 3 | 76 Counteracting User Attention Bias in Music Streaming Recommendation via Reward Modification (2022) Popularity Bias and Debiasing in Recommendation 74 SAQRec: Aligning Recommender Systems to User Satisfaction via Questionnaire Feedback (2024) Popularity Bias and Debiasing in Recommendation |
|
| 7 | Yuta Saito | 53 | 3 | 73 Asymmetric Tri-training for Debiasing Missing-Not-At-Random Explicit Feedback (2020) Popularity Bias and Debiasing in Recommendation 58 Unbiased Recommender Learning from Missing-Not-At-Random Implicit Feedback (2020) Popularity Bias and Long-Tail Debiasing |
|
| 8 | Yong He 0009 | 52 | 3 | 60 Breaking the Length Barrier: LLM-Enhanced CTR Prediction in Long Textual User Behaviors (2024) Efficiency and Deployment 53 GFlow-FT: Pick a Child Network via Gradient Flow for Efficient Fine-Tuning in Recommendation Systems (2022) CTR and Conversion Rate Prediction |
|
| 9 | Ke Fei | 52 | 3 | 47 Inverse Contrastive Representation Learning for Post-Click Conversion Rate Estimation (2026) CTR and Conversion Rate Prediction 47 Entire-Space Variational Information Exploitation for Post-Click Conversion Rate Prediction (2025) CTR and Conversion Rate Prediction |
|
| 10 | Yi Zhang 0103 | 51 | 3 | 74 CAFU: Constrained Alignment and Filtered Uniformity for Denoising Recommendation (2026) Popularity Bias and Debiasing in Recommendation 47 CETN: Contrast-enhanced Through Network for Click-Through Rate Prediction (2025) CTR and Conversion Rate Prediction |
|
| 11 | Min Hou 0004 COI: co-author | 51 | 3 | 85 Mitigating Recommendation Biases via Group-Alignment and Global-Uniformity in Representation Learning (2024) Popularity Bias and Debiasing in Recommendation 70 InvDiff: Invariant Guidance for Bias Mitigation in Diffusion Models (2025) Popularity Bias and Debiasing in Recommendation |
|
| 12 | Jie Jiang 0015 COI: co-author | 51 | 3 | 51 Ads Recommendation in a Collapsed and Entangled World (2024) Advertising and Auction Mechanisms 49 Long-Sequence Recommendation Models Need Decoupled Embeddings (2025) CTR and Conversion Rate Prediction |
|
| 13 | Xin Zhang 0091 | 51 | 3 | 56 Deep Session Heterogeneity-Aware Network for Click Through Rate Prediction (2024) CTR and Conversion Rate Prediction 56 Deep Dynamic Interest Learning With Session Local and Global Consistency for Click-Through Rate Predictions (2022) CTR and Conversion Rate Prediction |
|
| 14 | Wei Guo 0006 COI: co-author | 50 | 3 | 47 Enhancing Explicit and Implicit Feature Interactions via Information Sharing for Parallel Deep CTR Models (2021) CTR and Conversion Rate Prediction 47 DLF: Enhancing Explicit-Implicit Interaction via Dynamic Low-Order-Aware Fusion for CTR Prediction (2025) CTR and Conversion Rate Prediction |
|
| 15 | Yimeng Bai | 50 | 3 | 52 Unconstrained Monotonic Calibration of Predictions in Deep Ranking Systems (2025) CTR and Conversion Rate Prediction 45 Leveraging Watch-time Feedback for Short-Video Recommendations: A Causal Labeling Framework (2023) Recommendation Systems and Evaluation |
|
| 16 | Yu Zhang 0206 | 50 | 3 | 53 Adversarial Filtering Modeling on Long-term User Behavior Sequences for Click-Through Rate Prediction (2022) CTR and Conversion Rate Prediction 53 LREA: Low-Rank Efficient Attention on Modeling Long-Term User Behaviors for CTR Prediction (2025) CTR and Conversion Rate Prediction |
|
| 17 | Chongming Gao | 50 | 3 | 83 Alleviating Matthew Effect of Offline Reinforcement Learning in Interactive Recommendation (2023) Popularity Bias and Debiasing in Recommendation 68 SPRec: Self-Play to Debias LLM-based Recommendation (2025) Preference Alignment and Optimization |
|
| 18 | Lichan Hong | 49 | 3 | 58 Empowering Long-tail Item Recommendation through Cross Decoupling Network (CDN) (2023) Popularity Bias and Long-Tail Debiasing 47 DCN V2: Improved Deep & Cross Network and Practical Lessons for Web-scale Learning to Rank Systems (2021) CTR and Conversion Rate Prediction |
|
| 19 | Jinxin Hu | 49 | 3 | 55 Counterfactual Multi-task Learning for Delayed Conversion Modeling in E-commerce Sales Pre-Promotion (2026) CTR and Conversion Rate Prediction 53 LREA: Low-Rank Efficient Attention on Modeling Long-Term User Behaviors for CTR Prediction (2025) CTR and Conversion Rate Prediction |
|
| 21 | Xuanhua Yang | 49 | 3 | 57 Hierarchical User Long-term Behavior Modeling for Click-Through Rate Prediction (2025) CTR and Conversion Rate Prediction 48 AdaSparse: Learning Adaptively Sparse Structures for Multi-Domain Click-Through Rate Prediction (2022) Advertising and Auction Mechanisms |
|
| 22 | Lingling Yi | 48 | 3 | 83 DVR: Micro-Video Recommendation Optimizing Watch-Time-Gain under Duration Bias (2022) Popularity Bias and Debiasing in Recommendation 83 Alleviating Video-length Effect for Micro-video Recommendation (2024) Popularity Bias and Debiasing in Recommendation |
|
| 24 | Jieming Zhu | 48 | 3 | 56 EASE: Learning Lightweight Semantic Feature Adapters from Large Language Models for CTR Prediction (2024) CTR and Conversion Rate Prediction 50 Data-free Knowledge Distillation for Reusing Recommendation Models (2023) CTR and Conversion Rate Prediction |
|
| 25 | Jizhi Zhang | 47 | 3 | 84 Popularity-aware Distributionally Robust Optimization for Recommendation System (2023) Popularity Bias and Debiasing in Recommendation 78 BSL: Understanding and Improving Softmax Loss for Recommendation (2024) Popularity Bias and Debiasing in Recommendation |
|
| 26 | James Caverlee | 46 | 3 | 68 Quantifying and Mitigating Popularity Bias in Conversational Recommender Systems (2022) Conversational Recommendation Systems 58 Fairness among New Items in Cold Start Recommender Systems (2021) Fairness Metrics and Evaluation |
|
| 27 | Hao Liu 0078 | 46 | 3 | 47 Mitigating Distribution Shifts in Sequential Recommendation: An Invariance Perspective (2025) Contrastive and Self-Supervised Sequential Recommendation 44 WeaveRec: An LLM-Based Cross-Domain Sequential Recommendation Framework with Model Merging (2026) Efficiency and Deployment |
|
| 29 | Qiang Luo 0004 | 46 | 3 | 47 One Model to Serve All: Star Topology Adaptive Recommender for Multi-Domain CTR Prediction (2021) CTR and Conversion Rate Prediction 44 MPFormer: Adaptive Framework for Industrial Multi-Task Personalized Sequential Retriever (2025) Multi-Task and Multi-Objective Optimization |
|
| 30 | Ching Law | 45 | 3 | 51 A Generative Contextual Comprehension Paradigm for Takeout Ranking Model (2026) Advertising and Auction Mechanisms 51 SMTIR: Scenario-Aware Multi-Trigger Induction Network for CTR Prediction (2025) Advertising and Auction Mechanisms |
|
| 31 | Bo Peng 0009 | 45 | 3 | 47 M2: Mixed Models With Preferences, Popularities and Transitions for Next-Basket Recommendation (2023) Efficient and Long-Sequence Sequential Recommendation 44 eCeLLM: Generalizing Large Language Models for E-commerce from Large-scale, High-quality Instruction Data (2024) Efficiency and Deployment |
|
| 32 | Jin Chen 0008 COI: co-author | 45 | 3 | 50 An Incremental Update Framework for Online Recommenders with Data-Driven Prior (2023) CTR and Conversion Rate Prediction 46 Cooperative Retriever and Ranker in Deep Recommenders (2023) Recommendation with Search and Ranking |
|
| 33 | Xialong Liu | 45 | 3 | 47 Adversarial Filtering Modeling on Long-term User Behavior Sequences for Click-Through Rate Prediction (2022) CTR and Conversion Rate Prediction 46 Tutorial: Data Denoising Metrics in Recommender Systems (2023) Robustness to Noise and Distribution Shift |