Philipp Hager
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Philipp Hager has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Maarten de Rijke
University of Amsterdam, Amsterdam, Netherlands
7 shared publications - Onno Zoeter
Booking.com, Amsterdam, Netherlands
6 shared publications - Harrie Oosterhuis
University of Amsterdam, Amsterdam, Netherlands
3 shared publications - Shashank Gupta
University of Amsterdam, Amsterdam, Netherlands
3 shared publications - Ali Vardasbi
Spotify, Amsterdam, Netherlands
2 shared publications - Jean-Michel Renders
Naver Labs Europe, Meylan, France
2 shared publications
Latest Publications
- Open Access
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 358-368
Diagnosing Identifiability in Two-Tower Models for Unbiased Learning to Rank
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 358-368
Open AccessDiagnosing Identifiability in Two-Tower Models for Unbiased Learning to Rank
- Open Access
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 3220-3227
CLAX: Fast and Flexible Neural Click Models in JAX
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 3220-3227
Open AccessCLAX: Fast and Flexible Neural Click Models in JAX
- Open Access
CLAX: Fast and Flexible Neural Click Models in JAX
CLAX is a JAX-based library that implements classic click models using modern gradient-based optimization. While neural click models have emerged over the past decade, complex click models based on probabilistic graphical models (PGMs) have not systematically adopted gradient-based optimization, preventing practitioners from leveraging modern deep learning frameworks while preserving the interpretability of classic models. CLAX addresses this gap by replacing EM-based optimization with direct gradient-based optimization in a numerically stable manner. The framework's modular design enables the integration of any component, from embeddings and deep networks to custom modules, into classic click models for end-to-end optimization. We demonstrate CLAX's efficiency by running experiments on the full Baidu-ULTR dataset comprising over a billion user sessions in $\approx$ 2 hours on a single GPU, orders of magnitude faster than traditional EM approaches. CLAX implements ten classic click models, serving both industry practitioners seeking to understand user behavior and improve ranking performance at scale and researchers developing new click models. CLAX is available at: https://github.com/philipphager/claxOpen AccessCLAX: Fast and Flexible Neural Click Models in JAX
CLAX is a JAX-based library that implements classic click models using modern gradient-based optimization. While neural click models have emerged over the past decade, complex click models based on probabilistic graphical models (PGMs) have not systematically adopted gradient-based optimization, preventing practitioners from leveraging modern deep learning frameworks while preserving the interpretability of classic models. CLAX addresses this gap by replacing EM-based optimization with direct gradient-based optimization in a numerically stable manner. The framework's modular design enables the integration of any component, from embeddings and deep networks to custom modules, into classic click models for end-to-end optimization. We demonstrate CLAX's efficiency by running experiments on the full Baidu-ULTR dataset comprising over a billion user sessions in $\approx$ 2 hours on a single GPU, orders of magnitude faster than traditional EM approaches. CLAX implements ten classic click models, serving both industry practitioners seeking to understand user behavior and improve ranking performance at scale and researchers developing new click models. CLAX is available at: https://github.com/philipphager/clax - Open Access
Unidentified and Confounded? Understanding Two-Tower Models for Unbiased Learning to Rank (Extended Abstract)
Additive two-tower models are popular learning-to-rank methods for handling biased user feedback in industry settings. Recent studies, however, report a concerning phenomenon: training two-tower models on clicks collected by well-performing production systems leads to decreased ranking performance. This paper investigates two recent explanations for this observation: confounding effects from logging policies and model identifiability issues. We theoretically analyze the identifiability conditions of two-tower models, showing that either document swaps across positions or overlapping feature distributions are required to recover model parameters from clicks. We also investigate the effect of logging policies on two-tower models, finding that they introduce no bias when models perfectly capture user behavior. However, logging policies can amplify biases when models imperfectly capture user behavior, particularly when prediction errors correlate with document placement across positions. We propose a sample weighting technique to mitigate these effects and provide actionable insights for researchers and practitioners using two-tower models.
Open AccessUnidentified and Confounded? Understanding Two-Tower Models for Unbiased Learning to Rank (Extended Abstract)
Additive two-tower models are popular learning-to-rank methods for handling biased user feedback in industry settings. Recent studies, however, report a concerning phenomenon: training two-tower models on clicks collected by well-performing production systems leads to decreased ranking performance. This paper investigates two recent explanations for this observation: confounding effects from logging policies and model identifiability issues. We theoretically analyze the identifiability conditions of two-tower models, showing that either document swaps across positions or overlapping feature distributions are required to recover model parameters from clicks. We also investigate the effect of logging policies on two-tower models, finding that they introduce no bias when models perfectly capture user behavior. However, logging policies can amplify biases when models imperfectly capture user behavior, particularly when prediction errors correlate with document placement across positions. We propose a sample weighting technique to mitigate these effects and provide actionable insights for researchers and practitioners using two-tower models.
- Open Access
- 18 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 347-357
Unidentified and Confounded? Understanding Two-Tower Models for Unbiased Learning to Rank
- 18 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 347-357
Open AccessUnidentified and Confounded? Understanding Two-Tower Models for Unbiased Learning to Rank
- Open Access
Understanding the Effects of the Baidu-ULTR Logging Policy on Two-Tower Models
Despite the popularity of the two-tower model for unbiased learning to rank (ULTR) tasks, recent work suggests that it suffers from a major limitation that could lead to its collapse in industry applications: the problem of logging policy confounding. Several potential solutions have even been proposed; however, the evaluation of these methods was mostly conducted using semi-synthetic simulation experiments. This paper bridges the gap between theory and practice by investigating the confounding problem on the largest real-world dataset, Baidu-ULTR. Our main contributions are threefold: 1) we show that the conditions for the confounding problem are given on Baidu-ULTR, 2) the confounding problem bears no significant effect on the two-tower model, and 3) we point to a potential mismatch between expert annotations, the golden standard in ULTR, and user click behavior.
Open AccessUnderstanding the Effects of the Baidu-ULTR Logging Policy on Two-Tower Models
Despite the popularity of the two-tower model for unbiased learning to rank (ULTR) tasks, recent work suggests that it suffers from a major limitation that could lead to its collapse in industry applications: the problem of logging policy confounding. Several potential solutions have even been proposed; however, the evaluation of these methods was mostly conducted using semi-synthetic simulation experiments. This paper bridges the gap between theory and practice by investigating the confounding problem on the largest real-world dataset, Baidu-ULTR. Our main contributions are threefold: 1) we show that the conditions for the confounding problem are given on Baidu-ULTR, 2) the confounding problem bears no significant effect on the two-tower model, and 3) we point to a potential mismatch between expert annotations, the golden standard in ULTR, and user click behavior.
- Open Access
- 10 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 1546-1556
Unbiased Learning to Rank Meets Reality: Lessons from Baidu's Large-Scale Search Dataset
Unbiased learning-to-rank (ULTR) is a well-established framework for learning from user clicks, which are often biased by the ranker collecting the data. While theoretically justified and extensively tested in simulation, ULTR techniques lack empirical validation, especially on modern search engines. The Baidu-ULTR dataset released for the WSDM Cup 2023, collected from Baidu's search engine, offers a rare opportunity to assess the real-world performance of prominent ULTR techniques. Despite multiple submissions during the WSDM Cup 2023 and the subsequent NTCIR ULTRE-2 task, it remains unclear whether the observed improvements stem from applying ULTR or other learning techniques. In this work, we revisit and extend the available experiments on the Baidu-ULTR dataset. We find that standard unbiased learning-to-rank techniques robustly improve click predictions but struggle to consistently improve ranking performance, especially considering the stark differences obtained by choice of ranking loss and query-document features. Our experiments reveal that gains in click prediction do not necessarily translate to enhanced ranking performance on expert relevance annotations, implying that conclusions strongly depend on how success is measured in this benchmark.
- 10 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 1546-1556
Open AccessUnbiased Learning to Rank Meets Reality: Lessons from Baidu's Large-Scale Search Dataset
Unbiased learning-to-rank (ULTR) is a well-established framework for learning from user clicks, which are often biased by the ranker collecting the data. While theoretically justified and extensively tested in simulation, ULTR techniques lack empirical validation, especially on modern search engines. The Baidu-ULTR dataset released for the WSDM Cup 2023, collected from Baidu's search engine, offers a rare opportunity to assess the real-world performance of prominent ULTR techniques. Despite multiple submissions during the WSDM Cup 2023 and the subsequent NTCIR ULTRE-2 task, it remains unclear whether the observed improvements stem from applying ULTR or other learning techniques. In this work, we revisit and extend the available experiments on the Baidu-ULTR dataset. We find that standard unbiased learning-to-rank techniques robustly improve click predictions but struggle to consistently improve ranking performance, especially considering the stark differences obtained by choice of ranking loss and query-document features. Our experiments reveal that gains in click prediction do not necessarily translate to enhanced ranking performance on expert relevance annotations, implying that conclusions strongly depend on how success is measured in this benchmark.
- Open Access
- 4 March 2024
- Published by Association for Computing Machinery (ACM)
- p. 1118-1121
Unbiased Learning to Rank: On Recent Advances and Practical Applications
Since its inception, the field of unbiased learning to rank (ULTR) has remained very active and has seen several impactful advancements in recent years. This tutorial provides both an introduction to the core concepts of the field and an overview of recent advancements in its foundations, along with several applications of its methods. The tutorial is divided into four parts: Firstly, we give an overview of the different forms of bias that can be addressed with ULTR methods. Secondly, we present a comprehensive discussion of the latest estimation techniques in the ULTR field. Thirdly, we survey published results of ULTR in real-world applications. Fourthly, we discuss the connection between ULTR and fairness in ranking. We end by briefly reflecting on the future of ULTR research and its applications. This tutorial is intended to benefit both researchers and industry practitioners interested in developing new ULTR solutions or utilizing them in real-world applications.
- 4 March 2024
- Published by Association for Computing Machinery (ACM)
- p. 1118-1121
Open AccessUnbiased Learning to Rank: On Recent Advances and Practical Applications
Since its inception, the field of unbiased learning to rank (ULTR) has remained very active and has seen several impactful advancements in recent years. This tutorial provides both an introduction to the core concepts of the field and an overview of recent advancements in its foundations, along with several applications of its methods. The tutorial is divided into four parts: Firstly, we give an overview of the different forms of bias that can be addressed with ULTR methods. Secondly, we present a comprehensive discussion of the latest estimation techniques in the ULTR field. Thirdly, we survey published results of ULTR in real-world applications. Fourthly, we discuss the connection between ULTR and fairness in ranking. We end by briefly reflecting on the future of ULTR research and its applications. This tutorial is intended to benefit both researchers and industry practitioners interested in developing new ULTR solutions or utilizing them in real-world applications.
- Open Access
- 15 December 2023
- Published by Association for Computing Machinery (ACM)
- p. 145-148
Recent Advancements in Unbiased Learning to Rank
- 15 December 2023
- Published by Association for Computing Machinery (ACM)
- p. 145-148
Open AccessRecent Advancements in Unbiased Learning to Rank
- Open Access
- 14 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 750-756
Collaborative filtering algorithms are prone to mainstream-taste bias
Collaborative filtering has been a dominant approach in the recommender systems community since the early 1990s. Collaborative filtering (and other) algorithms, however, have been predominantly evaluated by aggregating results across users or user groups. These performance averages hide large disparities: an algorithm may perform very well for some users (or groups) and poorly for others. We show that performance variation is large and systematic. In experiments on three large-scale datasets and using an array of collaborative filtering algorithms, we demonstrate large performance disparities across algorithms, datasets and metrics for different users. We then show that two key features that characterize users, their mean taste similarity and dispersion in taste similarity with other users, can systematically explain performance variation better than previously identified features. We use these two features to visualize algorithm performance for different users and we point out that this mapping can capture different categories of users that have been proposed before. Our results demonstrate an extensive mainstream-taste bias in collaborative filtering algorithms, which implies a fundamental fairness limitation that needs to be mitigated.
- 14 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 750-756
Open AccessCollaborative filtering algorithms are prone to mainstream-taste bias
Collaborative filtering has been a dominant approach in the recommender systems community since the early 1990s. Collaborative filtering (and other) algorithms, however, have been predominantly evaluated by aggregating results across users or user groups. These performance averages hide large disparities: an algorithm may perform very well for some users (or groups) and poorly for others. We show that performance variation is large and systematic. In experiments on three large-scale datasets and using an array of collaborative filtering algorithms, we demonstrate large performance disparities across algorithms, datasets and metrics for different users. We then show that two key features that characterize users, their mean taste similarity and dispersion in taste similarity with other users, can systematically explain performance variation better than previously identified features. We use these two features to visualize algorithm performance for different users and we point out that this mapping can capture different categories of users that have been proposed before. Our results demonstrate an extensive mainstream-taste bias in collaborative filtering algorithms, which implies a fundamental fairness limitation that needs to be mitigated.