SetMIR: Multi-Interest Retrieval as Set Prediction
2026-08-31 • Information Retrieval
Information Retrieval
AI summaryⓘ
The authors address a problem in recommender systems where using just one profile to represent a user doesn't capture all their interests well. They propose a new method called SetMIR that predicts multiple distinct interests from a user's behavior, making sure these interests don’t overlap using a special matching method during training. At serving time, their system smartly picks only the needed interests to speed up recommendations. Tested on Snap's ads data, SetMIR performed better than other methods while using fewer resources, and when deployed, it improved user engagement and conversion rates significantly.
embedding-based retrievalrecommender systemsmulti-interest retrievaltransformerHungarian matchingpresence scoreNon-Maximum Suppression (NMS)Approximate Nearest Neighbor (ANN)click-through rate (CTR)conversion rate (CVR)
Authors
Xiaodong Liu, Congfei Zhang, Hsiang-wei Chao, Siman Wang, Xiao Bai, Tong Zhao, Jingxiao Ma, Wen Zhang, Zhe Liu, Shantanu Aggarwal, Di Huang, William Leach, Yunzhi Zhou, Yajun Wang, Jinchao Li, Yu Zhang
Abstract
Embedding-based retrieval is at the core of industrial recommender systems, but a single user embedding is often too limited to capture a user's diverse interests. Multi-interest retrieval addresses this by using multiple user embeddings, yet existing methods still suffer from two issues: interest collapse, where different embeddings learn the same interest, and static dispatch, where serving uses a fixed retrieval budget even when some embeddings are unnecessary. We propose SetMIR, which treats multi-interest retrieval as a set prediction problem. SetMIR encodes a user's behavior history with a transformer and uses K learnable queries to decode a set of user interests, each producing a retrieval embedding and a presence score. During training, Hungarian matching assigns targets to queries one-to-one, so matched queries learn distinct interests and the presence head learns which queries are active. At serving time, SetMIR uses presence scores and query-level Non-Maximum Suppression (NMS) to issue only active, non-redundant ANN queries. On Snap's Dynamic Product Ads (DPA) data, SetMIR outperforms four learned multi-interest retrievers on every metric while issuing 33% fewer ANN queries per request. Deployed as a new retrieval source in the DPA production stack, SetMIR lifts overall CVR by 3.1%, while lifting CTR by 44% and CVR by 51% over the item-to-item retrieval source with the same item embeddings, ANN index, and retrieval quota.