mlsanityData to 7 Oct 2026

Counterfactual Evaluation of Ads Ranking Models through Domain Adaptation

arXiv 2409.19824 ↗First posted 29 Sep 2024cs.IR cs.AIcited by 0

On the map: LLM alignment / RLHF / preference optimisation → Preference alignment for recommenders and RAG

Abstract (arXiv)

We propose a domain-adapted reward model that works alongside an Offline A/B testing system for evaluating ranking models. This approach effectively measures reward for ranking model changes in large-scale Ads recommender systems, where model-free methods like IPS are not feasible. Our experiments demonstrate that the proposed technique outperforms both the vanilla IPS method and approaches using non-generalized reward models.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Preference alignment for recommenders and RAG, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 6 of its 13 references that are in the map. First posted on arXiv 29 Sep 2024.

Builds on

6 of 13 references.

References · 6

  1. 2021The Simpson's Paradox in the Offline Evaluation of Recommendation Systemscited by 16
  2. 2019Seeker: Real-Time Interactive Searchcited by 9
  3. 2022Adaptive Experimental Design and Counterfactual Inferencecited by 5
  4. 2020Causal Transfer Random Forest: Combining Logged Data and Randomized Experiments for Robust Predictioncited by 4
  5. 2023Neural Insights for Digital Marketing Content Designcited by 4
  6. 2018Genie: An Open Box Counterfactual Policy Estimator for Optimizing Sponsored Search Marketplacecited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025RewardRank: Optimizing True Learning-to-Rank Utility
  2. 2026Adaptive Doubly Robust Off-Policy Evaluation for Ranking Policies under Diverse User Behavior
  3. 2025Multi-task Offline Reinforcement Learning for Online Advertising in Recommender Systems
  4. 2025Reward Balancing Revisited: Enhancing Offline Reinforcement Learning for Recommender Systems
  5. 2025Identifying Offline Metrics that Predict Online Impact: A Pragmatic Strategy for Real-World Recommender Systems
  6. 2026Robust Post-Training for Generative Recommenders: Why Exponential Reward-Weighted SFT Outperforms RLHF
  7. 2025MTRec: Learning to Align with User Preferences via Mental Reward Models
  8. 2023Towards Validating Long-Term User Feedbacks in Interactive Recommendation Systems