Method

Two-tower model

Separate user and item encoders trained for retrieval, incl. DSSM.

60 cases from 23 companies apply Two-tower model. 30 were checked in an online A/B test; 18 report results in production. Includes variants: DSSM.

Typical tasks: Retrieve and rank from millions of items, Rank search results, Recommend to new users and new items, Learn user and item embeddings and 9 more. Companies: Yandex, Pinterest, LinkedIn, Meta, Avito, eBay and 17 more. Often combined with: DSSM, Contrastive learning, Multi-task learning, In-batch negatives, Approximate nearest neighbors and 105 more.

Meta Andromeda: Powering the Next Generation of Ads Retrieval Through Model-System Co-Designadtech2026-10
GenRecs Tokenization: Transforming Ads Recommendation at Meta with Stable Model Vocabularyadtech2026-10
Beyond Two Towers: Launching the 3-Tower Engagement Co-Train Model (Part 2)adtech2026-09
Semantic Search for AI Agents at Scale: Retrieval and Ranking for LinkedIn’s Hiring Assistantsocial2026-06
Building a unified consumer memory for personalization at scalefood-delivery2026-06
Introducing CLEPR, our model for semantic understandingadtech2026-06
SilverTorch: Index as Model — A New Retrieval Paradigm for Recommendation Systemssocial2026-05
Shaping product understanding with contrastive reinforcement learningmarketplace2026-05
AI Blender in Search: How We Stopped Worshipping Clicks and Moved Block Placement to LLMssearch-engine2026-05
How Ad Ranking Works at Avitoclassifieds2026-05
X-split ARGUS: An Autoregressive Encoder-Decoder for Ad Rankingadtech2026-05
Enhancing Ad Relevance: Integrating Real-Time Context into Sequential Recommender Modelsadtech2026-05
From Clicks to Conversions: Architecting Shopping Conversion Candidate Generation at Pinterestadtech2026-04
Scaling Recommendation Systems with Request-Level Deduplicationsocial2026-04
GPU-Serving Two-Tower Models for Lightweight Ads Engagement Predictionadtech2026-02
Leveraging Commerce Data for Outcome-Based Relevancy in Agentic Recommendation Systemsadtech2026-02
Ads Candidate Generation using Behavioral Sequence Modelingsocial2026-01
How Etsy Uses LLMs to Improve Search Relevancemarketplace2026-01
Dense Retrievalmarketplace2025-11
Reducing Irrelevant Impressions of the Thematic Block in Geo Search for Navigation Scenariossearch-engine2025-11
DoorDash: an LLM-assisted personalization framework balancing familiarity, affordability and noveltyretail2025-10
Modernizing Home Feed Pre-Ranking Stagesocial2025-05
A Neural Network User Interest Model: How We Improved Diversity and Relevance of Recommendations on the Avito Home Pageclassifieds2025-05
Scaling Recommender Systemsmedia-streaming2025-05
Experience with Transformer-Based Personalization at Avitoclassifieds2025-05
Runtime Transformer in Personal Recommendations at Yandex Lavkafood-delivery2025-05
How Large Language Models Help Users Formulate Queries in Yandex Mapssearch-engine2025-05
How We Built a Smart Assistant in Lavka on Top of YaGPTfood-delivery2025-05
Do We Need Graph Neural Networks for Encoding Items in Recommendations?e-commerce2025-05
JUDE: LLM-based representation learning for LinkedIn job recommendationsmarketplace2025-05
Embedding-based retrieval for Airbnb searchtravel2025-03
Advancements in Embedding-Based Retrieval at Pinterest Homefeedsocial2025-02
Establishing a Large Scale Learned Retrieval System at Pinterestsocial2025-01
Candidate Generation in a Large Scale Graph Recommendation System: People You May Knowsocial2024-06
Transforming Recommendationssocial2024-05
How we improved the 'Unfamiliar' setting recommendation algorithm in Yandex Musicmedia-streaming2024-05
Neural Filter with YaGPT: Helping Users Sort Out Their Mailboxgeneral2024-05
RetrievalRecsys: An Architecture Modification to Account for Text in Personal Recommender Systemsmarketplace2024-05
Matching LinkedIn members with the right Premium productssocial2024-05
Learning Embeddings for Lodging Travel Conceptstravel2024-01
Candidate Generation Using a Two Tower Approach With Expedia Group Traveler Datatravel2023-12
Adopting the Vespa search engine for serving personalized second-hand fashion recommendations at Vintedmarketplace2023-10
How LinkedIn Is Using Embeddings to Up Its Match Game for Job Seekerssocial2023-10
Evolving Recommendations: A Personalized User-Based Ranking Modele-commerce2023-10
Beyond Words: How Multimodal Embeddings Elevate eBay's Product Recommendationsmarketplace2023-09
From Pre-trained to Fine-tuned: Nextdoor’s Path to Effective Embedding Applicationssocial2023-09
Scaling the Instagram Explore recommendations systemsocial2023-08
Embedding-based Retrieval with Two-Tower Models in Spotlightsocial2023-06
Deploying Transformers in Yandex Musicmedia-streaming2023-05
Creating the Conditions to Launch a Recommender System with Manual Labelingsocial2023-05