InteractRank: Personalized Web-Scale Search Pre-Ranking with Cross Interaction Features
Not on the map. Why?
Abstract (arXiv)
Modern search systems use a multi-stage architecture to deliver personalized results efficiently. Key stages include retrieval, pre-ranking, full ranking, and blending, which refine billions of items to top selections. The pre-ranking stage, vital for scoring and filtering hundreds of thousands of items down to a few thousand, typically relies on two tower models due to their computational efficiency, despite often lacking in capturing complex interactions. While query-item cross interaction features are paramount for full ranking, integrating them into pre-ranking models presents efficiency-related challenges. In this paper, we introduce InteractRank, a novel two tower pre-ranking model with robust cross interaction features used at Pinterest. By incorporating historical user engagement-based query-item interactions in the scoring function along with the two tower dot product, InteractRank significantly boosts pre-ranking performance with minimal latency and computation costs. In real-world A/B experiments at Pinterest, InteractRank improves the online engagement metric by 6.5% over a BM25 baseline and by 3.7% over a vanilla two tower baseline. We also highlight other components of InteractRank, like real-time user-sequence modeling, and analyze their contributions through offline ablation studies. The code for InteractRank is available at https://github.com/pinterest/atg-research/tree/main/InteractRank.
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 30 of its 51 references that are in the map. First posted on arXiv 9 Apr 2025.
Builds on
30 of 51 references.
References · 30
- 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
- 2018Self-Attentive Sequential Recommendationcited by 1,184
- 2020ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERTcited by 1,178
- 2018Graph Convolutional Neural Networks for Web-Scale Recommender Systemscited by 1,138
- 2019BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformercited by 964
- 2016Wide & Deep Learning for Recommender Systemscited by 747
- 2015Session-based Recommendations with Recurrent Neural Networkscited by 695
- 2017Deep Interest Network for Click-Through Rate Predictioncited by 544
- 2018Personalized Top-N Sequential Recommendation via Convolutional Sequence Embeddingcited by 476
- 2021ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interactioncited by 476
- 2018Deep Interest Evolution Network for Click-Through Rate Predictioncited by 338
- 2020DCN V2: Improved Deep & Cross Network and Practical Lessons for Web-scale Learning to Rank Systemscited by 327
18 more
- 2018AutoInt: Automatic Feature Interaction Learning via Self-Attentive Neural Networkscited by 320
- 2021SPLADE: Sparse Lexical and Expansion Model for First Stage Rankingcited by 286
- 2018xDeepFM: Combining Explicit and Implicit Feature Interactions for Recommender Systemscited by 278
- 2020Embedding-based Retrieval in Facebook Searchcited by 156
- 2019Behavior Sequence Transformer for E-commerce Recommendation in Alibabacited by 130
- 2022From Distillation to Hard Negative Sampling: Making Sparse Neural IR Models More Effectivecited by 98
- 2019FiBiNET: Combining Feature Importance and Bilinear feature Interaction for Click-Through Rate Predictioncited by 83
- 2022PinnerFormer: Sequence Modeling for User Representation at Pinterestcited by 57
- 2021MaskNet: Introducing Feature-Wise Multiplication to CTR Ranking Models by Instance-Guided Maskcited by 48
- 2023TransAct: Transformer-based Realtime User Action Model for Recommendation at Pinterestcited by 35
- 2020COLD: Towards the Next Generation of Pre-Ranking Systemcited by 31
- 2019Learning a Unified Embedding for Visual Search at Pinterestcited by 25
- 2021Billion-Scale Pretraining with Vision Transformers for Multi-Task Visual Representationscited by 24
- 2022IntTower: the Next Generation of Two-Tower Model for Pre-Ranking Systemcited by 18
- 2021Towards a Better Tradeoff between Effectiveness and Efficiency in Pre-Ranking: A Learnable Feature Selection based Approachcited by 15
- 2023Beyond Two-Tower Matching: Learning Sparse Retrievable Cross-Interactions for Recommendationcited by 13
- 2024OmniSearchSage: Multi-Task Multi-Entity Embeddings for Pinterest Searchcited by 13
- 2023LightSAGE: Graph Neural Networks for Large Scale Item Retrieval in Shopee's Advertisement Recommendationcited by 3
Built on by
Citing papers per year: 2026: 2.
2 papers.