{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/unbiased-lambdamart-an-unbiased-pairwise","title":"Unbiased LambdaMART: An Unbiased Pairwise Learning-to-Rank Algorithm","arxiv_id":"1809.05818","date":"2018-09-16","proceeding":null,"authors":["Ziniu Hu","Yang Wang","Qu Peng","Hang Li"],"abstract":"Although click data is widely used in search systems in practice, so far the\ninherent bias, most notably position bias, has prevented it from being used in\ntraining of a ranker for search, i.e., learning-to-rank. Recently, a number of\nauthors have proposed new techniques referred to as 'unbiased\nlearning-to-rank', which can reduce position bias and train a relatively\nhigh-performance ranker using click data. Most of the algorithms, based on the\ninverse propensity weighting (IPW) principle, first estimate the click bias at\neach position, and then train an unbiased ranker with the estimated biases\nusing a learning-to-rank algorithm. However, there has not been a method for\npairwise learning-to-rank that can jointly conduct debiasing of click data and\ntraining of a ranker using a pairwise loss function. In this paper, we propose\na novel algorithm, which can jointly estimate the biases at click positions and\nthe biases at unclick positions, and learn an unbiased ranker. Experiments on\nbenchmark data show that our algorithm can significantly outperform existing\nalgorithms. In addition, an online A/B Testing at a commercial search engine\nshows that our algorithm can effectively conduct debiasing of click data and\nenhance relevance ranking.","url_abs":"http://arxiv.org/abs/1809.05818v2","url_pdf":"http://arxiv.org/pdf/1809.05818v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"unbiased-lambdamart-an-unbiased-pairwise","repo_url":"https://github.com/acbull/Unbiased_LambdaMart","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"learning-to-rank","task_name":"Learning-To-Rank"},{"task_slug":null,"task_name":"Position"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}