Method

Contrastive learning

Representation learning with positive/negative pairs (InfoNCE, NT-Xent).

48 cases from 28 companies apply Contrastive learning. 18 were checked in an online A/B test; 14 report results in production. Includes variants: InfoNCE.

Typical tasks: Retrieve and rank from millions of items, Learn user and item embeddings, Rank search results, Recommend to new users and new items and 7 more. Companies: Yandex, Pinterest, LinkedIn, Criteo, VK, Skoltech and 22 more. Often combined with: Two-tower model, InfoNCE, Embedding fine-tuning, Multimodal embeddings, Self-supervised learning and 86 more.

Safe and Efficient Multi-Epoch Embedding Trainingsocial2026-11
GenRecs Tokenization: Transforming Ads Recommendation at Meta with Stable Model Vocabularyadtech2026-10
Rebuilding LinkedIn's Follows Recommendations with LLM-Based Semantic Retrieval and Rankingsocial2026-08
Semantic Search for AI Agents at Scale: Retrieval and Ranking for LinkedIn’s Hiring Assistantsocial2026-06
Introducing CLEPR, our model for semantic understandingadtech2026-06
Semantic IDs: Product Understanding at Scalee-commerce2026-06
A Hybrid Generative-Ranking Model Based on Semantic IDs in Yandex Music Recommendationsmedia-streaming2026-05
Learning from Unstructured Sequences in 2026fintech2026-05
Scaling Recommendation Systems with Request-Level Deduplicationsocial2026-04
DashCLIP: Leveraging multimodal models for generating semantic embeddingsfood-delivery2026-02
Leveraging Commerce Data for Outcome-Based Relevancy in Agentic Recommendation Systemsadtech2026-02
PinLanding: Turn Billions of Products into Instant Shopping Collections with Multimodal AIe-commerce2026-01
Dense Retrievalmarketplace2025-11
Building Multimodal Product Embeddings to Improve Matching at Yandex Marketmarketplace2025-11
Experience with Transformer-Based Personalization at Avitoclassifieds2025-05
How Large Language Models Help Users Formulate Queries in Yandex Mapssearch-engine2025-05
Developing Multimodal Content Models for Recommendationssocial2025-05
On the Deep Value of Useless Email Campaignsgeneral2025-05
RuModernBERT and USERv2: The Evolution of Russian-Language Encodersgeneral2025-05
Embedders 2025: A New Vector in Text Representationretail2025-05
Aggregating Representations of User Action Sequencesbanking2025-05
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
Fine-tuned SigLIP Image Embeddings for Similar Looks Recommendation in a Japanese C2C Marketplacemarketplace2024-11
Learning From Structure: Discord’s Entity-Relationship Embeddingssocial2024-09
Applying NLP Models to Automate Call Center Workbanking2024-05
Shazam for Full Names: Finding a Match When a Name Can Be Written 20 Different Waysfintech2024-05
Neural Filter with YaGPT: Helping Users Sort Out Their Mailboxgeneral2024-05
Optimizing Marketplace Operations with Language Modelsmarketplace2024-05
Generative and Contrastive Learning for Event Sequences: Which Is Better and Can They Be Combined?banking2024-05
How to Build Multilingual Product Matching Without Dataretail2024-05
Multimodal Geo Embeddings: Methods, Results and Deploymentbanking2024-05
How We Trained a Content Video Encoder Without Labeled Datasocial2024-05
RetrievalRecsys: An Architecture Modification to Account for Text in Personal Recommender Systemsmarketplace2024-05
LinkSage: GNN-based Pinterest Off-site Content Understandingsocial2024-03
Fine-Tuned CLIP: Better Listing Experience and 80% More Budget-Friendlymarketplace2023-12
Building In-Video Searchmedia-streaming2023-11
From Pre-trained to Fine-tuned: Nextdoor’s Path to Effective Embedding Applicationssocial2023-09
Deploying Transformers in Yandex Musicmedia-streaming2023-05
Neural Network Outfit Compositione-commerce2023-05
NLP at Alfa-Bank, Part 8: The Development History of an Operator Assistantbanking2023-05
How eBay's New Search Feature Was Inspired By Window Shoppinge-commerce2023-03
Deep Learning for Infinite (Multi-Lingual) Keywordsmarketplace2023-01
How eBay Created a Language Model With Three Billion Item Titlesmarketplace2023-01
When a picture is worth more than wordstravel2022-12
Transformers for Personalizationsearch-engine2022-06