AutoCTR employs evolutionary algorithms and learning-to-rank guidance to automatically discover neural interaction architectures for CTR prediction.

Qingquan Song, Dehua Cheng, Hanning Zhou et al.

Proposes Search-based Interest Model (SIM) with two-stage search for long user behavior sequences, boosting CTR prediction accuracy.

Pi Qi, Xiaoqiang Zhu, Guorui Zhou et al.

The study shows that by considering user interaction time, RecentPop and DecayPop methods can improve recommendation accuracy by over 70% on the MovieLens dataset.

Yitong Ji, Aixin Sun, Jie Zhang et al.

This study compares neural collaborative filtering (MLP) with matrix factorization (dot product), showing that after hyperparameter tuning, dot product outperforms MLP in recommendation tasks with lower complexity.

Steffen Rendle, Walid Krichene, Li Zhang et al.

DeepBidirectional RNN classifier automates identification of operating segments in financial reports, achieving over 97% F1 score.

Zhiqiang Ma, Steven Pomerville, Mingyang Di et al.

ColBERT employs a late interaction architecture with offline document encoding, enabling fast and effective passage retrieval with BERT embeddings.

Omar Khattab, Matei Zaharia

TREC 2019 deep learning track demonstrates that large-scale labeled data enables deep models like BERT to outperform traditional IR methods by up to 29.4%.

Nick Craswell, Bhaskar Mitra, Emine Yilmaz et al.

DeepCT leverages BERT to generate context-aware term weights, improving first-stage retrieval accuracy by 27-46% on MS MARCO and TREC-CAR.

Zhuyun Dai, Jamie Callan

Proposes Fi-GNN, a graph neural network model that explicitly models feature interactions, improving CTR prediction by 4.2% AUC on Criteo.

Zekun Li, Zeyu Cui, Shu Wu et al.

Systematic analysis of 18 deep learning recommendation algorithms shows only 7 reproducible; most are outperformed by simple heuristics, exposing current research limitations.

Maurizio Ferrari Dacrema, Paolo Cremonesi, Dietmar Jannach

DLRM integrates embedding tables, MLP, and feature interactions, employing model and data parallelism to scale recommendation systems efficiently.

Maxim Naumov, Dheevatsa Mudigere, Hao-Jun Michael Shi et al.

FUNSD dataset for form understanding in noisy scanned documents, includes 199 fully annotated forms.

Guillaume Jaume, Hazim Kemal Ekenel, Jean-Philippe Thiran

This paper introduces the Behavior Sequence Transformer (BST) for e-commerce CTR prediction, achieving significant online CTR gains.

Qiwei Chen, Huan Zhao, Wei Li et al.

DiffNet models recursive social influence diffusion, improving recommendation accuracy by over 13% on real datasets.

Le Wu, Peijie Sun, Yanjie Fu et al.

BERT4Rec employs bidirectional Transformer with Cloze training, outperforming SOTA in sequential recommendation.

Fei Sun, Jun Liu, Jian Wu et al.

FGCNN combines CNN and MLP for automatic feature generation, significantly improving CTR prediction accuracy.

Bin Liu, Ruiming Tang, Yingzhi Chen et al.

Proposed DANSER model uses dual graph attention networks to dynamically model multifaceted social effects, achieving significant accuracy improvements in recommendation systems.

Qitian Wu, Hengrui Zhang, Xiaofeng Gao et al.

A projected-gradient attack on graph recommenders raised unpopular-item exposure by about 580× with only 1% fake users.

Minghong Fang, Guolei Yang, Neil Zhenqiang Gong et al.

Constructed MedSTS with 174,629 sentence pairs from Mayo Clinic clinical notes; experts annotated 1068 pairs with 0-5 similarity scores for clinical STS research.

Yanshan Wang, Naveed Afzal, Sunyang Fu et al.

Proposes SASRec, a self-attention-based sequential recommendation model outperforming SOTA in accuracy and efficiency.

Wang-Cheng Kang, Julian McAuley