mlsanityData to 14 Oct 2026

Off-Policy Evaluation of Ranking Policies via Embedding-Space User Behavior Modeling

arXiv 2506.00446 ↗First posted 31 May 2025stat.ML cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Off-policy evaluation (OPE) in ranking settings with large ranking action spaces, which stems from an increase in both the number of unique actions and length of the ranking, is essential for assessing new recommender policies using only logged bandit data from previous versions. To address the high variance issues associated with existing estimators, we introduce two new assumptions: no direct effect on rankings and user behavior model on ranking embedding spaces. We then propose the generalized marginalized inverse propensity score (GMIPS) estimator with statistically desirable properties compared to existing ones. Finally, we demonstrate that the GMIPS achieves the lowest MSE. Notably, among GMIPS variants, the marginalized reward interaction IPS (MRIPS) incorporates a doubly marginalized importance weight based on a cascade behavior assumption on ranking embeddings. MRIPS effectively balances the trade-off between bias and variance, even as the ranking action spaces increase and the above assumptions may not hold, as evidenced by our experiments.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 21 of its 35 references that are in the map. First posted on arXiv 31 May 2025.

Builds on

21 of 35 references.

References · 21

  1. 2015Doubly Robust Policy Evaluation and Optimizationcited by 158
  2. 2016Off-policy evaluation for slate recommendationcited by 103
  3. 2018Offline A/B testing for Recommender Systemscited by 96
  4. 2019Doubly robust off-policy evaluation with shrinkagecited by 87
  5. 2020Open Bandit Dataset and Pipeline: Towards Realistic and Reproducible Off-Policy Evaluationcited by 77
  6. 2019Policy Learning for Fairness in Rankingcited by 74
  7. 2018Policy Evaluation and Optimization with Continuous Treatmentscited by 58
  8. 2022Off-Policy Evaluation for Large Action Spaces via Embeddingscited by 54
  9. 2020Adaptive Estimator Selection for Off-Policy Evaluationcited by 38
  10. 2021Computationally Efficient Optimization of Plackett-Luce Ranking Models for Relevance and Fairnesscited by 36
  11. 2020Counterfactual Evaluation of Slate Recommendations with Sequential Reward Interactionscited by 35
  12. 2023Off-Policy Evaluation for Large Action Spaces via Conjunct Effect Modelingcited by 34
9 more
  1. 2018Offline Evaluation of Ranking Policies with Click Modelscited by 32
  2. 2022Doubly Robust Off-Policy Evaluation for Ranking Policies under the Cascade Behavior Modelcited by 30
  3. 2022Policy-Adaptive Estimator Selection for Off-Policy Evaluationcited by 24
  4. 2024Off-Policy Evaluation of Slate Bandit Policies via Optimizing Abstractioncited by 17
  5. 2023Off-Policy Evaluation of Ranking Policies under Diverse User Behaviorcited by 16
  6. 2023Off-Policy Evaluation for Large Action Spaces via Policy Convolutioncited by 14
  7. 2024POTEC: Off-Policy Learning for Large Action Spaces via Two-Stage Policy Decompositioncited by 13
  8. 2024Long-term Off-Policy Evaluation and Learningcited by 9
  9. 2024Effective Off-Policy Evaluation and Learning in Contextual Combinatorial Banditscited by 8

Built on by

Not cited yet.