Click-Through-Rate Prediction and Sequential Recommendation
Neural models that predict clicks from sparse user and item features or predict the next item from a user's recent history.
Tier Grades
Tier A: Frontier AI labs
feasible
Google, Meta and Alibaba built and deployed these models (DIN in Alibaba display advertising).
Tier B: Indonesian AI labs
feasible
Models of the SASRec kind are published with code and need modest compute.
Tier C: Institution units such as ITB
feasible
SASRec was produced by a 2-author team, well within an institution unit.
Tier D: Organisations such as SEMAI (PoT, PoC, PoV)
feasible
CTR and next-item models fit e-commerce and media proofs of value with a client's own logs.
Where this comes from
Tiers: a) frontier AI labs, b) Indonesian AI labs, c) institution units such as ITB, d) organisations such as SEMAI delivering proof of technology, concept and value. Each rationale cites the anchors below.
Anchors
- ComputeDIN and SASRec abstracts give no compute; SASRec is a self-attention model sized for single-machine training on public datasets (not stated numerically).
- DataDIN: deployed on Alibaba online display advertising data; SASRec: user action histories from public datasets, sparse and dense.
- Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): 10 authors on DIN; 2 on SASRec.
- Time horizonYears (mature direction).
- Shelf-life signalsGenerative recommenders (HSTU, OneRec) displacing feature-interaction models at the largest platforms, Graph collaborative filtering (LightGCN) as a standard baseline
- Pipeline roleTier A publishes and deploys architectures; Tier C benchmarks them on local data; Tier D builds recommenders for clients in proofs of value.
Where this comes from
Team size is a proxy (author counts of exemplar papers), not a headcount.
Who and What
Exemplars
Observed actors
Google
University of Science and Technology of China (USTC)
Alibaba (Qwen)
Alibaba Group
National University of Singapore
Harbin Institute of Technology
Huawei
Microsoft ResearchLinked models
DCN V2recommendationexemplar
DLRM (Deep Learning Recommendation Model)recommendationexemplar
SASRecrecommendationexemplarDIN (Deep Interest Network)recommendationexemplarSIM (Search-based Interest Model)recommendationproduced byLightGCNrecommendationproduced byPLE (Progressive Layered Extraction)recommendationproduced byNGCF (Neural Graph Collaborative Filtering)recommendationproduced byBERT4Recrecommendationproduced byDIEN (Deep Interest Evolution Network)recommendationproduced byxDeepFMrecommendationproduced by
Mult-VAE (variational autoencoder for collaborative filtering)recommendationproduced byMMoE (Multi-gate Mixture-of-Experts)recommendationproduced byDCN (Deep and Cross Network)recommendationproduced byNCF (Neural Collaborative Filtering) and NeuMFrecommendationproduced byDeepFMrecommendationproduced byWide and Deeprecommendationproduced byGRU4Recrecommendationproduced byBPR (Bayesian Personalized Ranking)recommendationproduced byDLRM-dcnv2recommendationproduced byLinks
R&D TopicClick-Through-Rate Prediction and Sequential Recommendation
ModelsDIN (Deep Interest Network)Alibaba (Qwen), recommendationDCN V2Google, recommendationDLRM (Deep Learning Recommendation Model)Meta, recommendationSASRecUC San Diego, recommendation+16 more
DIN (Deep Interest Network)Alibaba (Qwen), recommendationDCN V2Google, recommendationDLRM (Deep Learning Recommendation Model)Meta, recommendationSASRecUC San Diego, recommendation+16 moreWho Runs It
GrabIn production, 6 Oct 2026
GrabIn production, 6 Oct 2026Models 20
DCN (Deep and Cross Network)Google, Stanford University, recommendationproduced byNGCF (Neural Graph Collaborative Filtering)National University of Singapore, University of Science and Technology of China (USTC), recommendationproduced byDCN V2Google, recommendationexemplarDeepFMHarbin Institute of Technology, Huawei, recommendationproduced byLightGCNUniversity of Science and Technology of China (USTC), recommendationproduced byWide and DeepGoogle, recommendationproduced byxDeepFMUniversity of Science and Technology of China (USTC), Microsoft Research, recommendationproduced byBERT4RecAlibaba Group, recommendationproduced byDIEN (Deep Interest Evolution Network)Alibaba Group, recommendationproduced byDIN (Deep Interest Network)Alibaba (Qwen), recommendationexemplarDLRM (Deep Learning Recommendation Model)Meta, recommendationexemplarDLRM-dcnv2Google, recommendationproduced by+8
GRU4RecGravity R&D, recommendationproduced byMult-VAE (variational autoencoder for collaborative filtering)Netflix, recommendationproduced byNCF (Neural Collaborative Filtering) and NeuMFNational University of Singapore, recommendationproduced bySASRecUC San Diego, recommendationexemplarSIM (Search-based Interest Model)Alibaba (Qwen), recommendationproduced byBPR (Bayesian Personalized Ranking)recommendationproduced byMMoE (Multi-gate Mixture-of-Experts)recommendationproduced byPLE (Progressive Layered Extraction)recommendationproduced by
Mult-VAE (variational autoencoder for collaborative filtering)Netflix, recommendationproduced byNCF (Neural Collaborative Filtering) and NeuMFNational University of Singapore, recommendationproduced bySASRecUC San Diego, recommendationexemplarSIM (Search-based Interest Model)Alibaba (Qwen), recommendationproduced byBPR (Bayesian Personalized Ranking)recommendationproduced byMMoE (Multi-gate Mixture-of-Experts)recommendationproduced byPLE (Progressive Layered Extraction)recommendationproduced byPapers
- Session-based Recommendations with Recurrent Neural NetworksBalázs Hidasi et al., arXiv preprint, 2015 founding
- DCN V2: Improved Deep & Cross Network and Practical Lessons for Web-scale Learning to Rank SystemsRuoxi Wang et al., In Proceedings of the Web Conference 2021 (WWW '21), 2020 state of the art
- BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from TransformerFei Sun et al., CIKM 2019 (per arXiv comment), 2019 state of the art
- Deep Learning Recommendation Model for Personalization and Recommendation SystemsMaxim Naumov et al., arXiv preprint, 2019 state of the art
- Deep Learning based Recommender System: A Survey and New PerspectivesShuai Zhang et al., arXiv preprint, 2017 survey
Where this comes from
The papers to read first: where the direction began, its current best, and a survey when one exists.
Sources
- arxiv.org/abs/1706.06978 https://arxiv.org/abs/1706.06978
- arxiv.org/abs/1808.09781 https://arxiv.org/abs/1808.09781
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Click-Through-Rate Prediction and Sequential Recommendation", https://aurion.semai.id/topics/click-through-rate-and-sequential-recommendation/, records of October 2026.