{"id":"arxiv/2602.10696","url":"https://arxiv.org/abs/2602.10696","tags":["stat.ML","cs.LG","math.OC","math.ST"],"score":0,"title":"Robust Assortment Optimization from Observational Data","domain":"stat.ML","source":"arxiv","authors":["Miao Lu","Yuxuan Han","Han Zhong","Zhengyuan Zhou","Jose Blanchet"],"reviews":0,"subject":"stat.ML","summary":"Assortment optimization is a fundamental challenge in modern retail and recommendation systems, where the goal is to select a subset of products that maximizes expected revenue under complex customer choice behaviors. While recent advances in data-driven methods have leveraged historical data to learn and optimize assortments, these approaches typically rely on strong assumptions -- namely, the stability of customer preferences and the correctness of the underlying choice models. However, such assumptions frequently break in real-world scenarios due to preference shifts and model…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-02-11","reactions":{}}