arxiv: 2604.23810 · v1 · submitted 2026-04-26 · 💻 cs.IR

Recognition: unknown

Similar Users-Augmented Interest Network

Xiaolong Chen , Haoyi Zhao , Xu Huang , Defu Lian

Authors on Pith no claims yet

Pith reviewed 2026-05-08 05:13 UTC · model grok-4.3

classification 💻 cs.IR

keywords CTR predictionrecommender systemsuser behavior sequencessimilar user retrievalsequence augmentationattention mechanismssparsity handlingsequential modeling

0 comments

The pith

Augmenting a target user's behavior sequence with sequences from similar users improves click-through rate prediction despite data sparsity.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows how to retrieve users with similar past behaviors through embeddings, then combine their sequences with the target user's own history to form a longer augmented sequence for modeling. Special position encodings mark which user each behavior came from and its relation to the target item, while a joint attention mechanism weighs both item similarities and user similarities to limit noise from the added data. This addresses the common problem of short or incomplete personal histories that limit accurate preference profiling. If the approach works as described, models can deliver better predictions without needing every user to have extensive individual interaction logs.

Core claim

The SUIN method retrieves similar users from a pool using sequence-encoded behavior embeddings, concatenates their sequences to the target user's in descending similarity order to build an augmented sequence, applies user-specific target-aware position encoding to distinguish sources and relative positions, and uses user-aware target attention that models both item-item and user-user correlations to exploit the added behaviors while reducing noise, yielding higher CTR prediction accuracy than prior sequential models.

What carries the argument

User-aware target attention that jointly accounts for item-item and user-user correlations within the augmented multi-user sequence to leverage similar users' data while filtering noise.

If this is right

The augmented sequence enables more comprehensive user profiling even when individual behavior histories are short or sparse.
User-specific position encoding distinguishes the source of each behavior and its position relative to the target item in the combined sequence.
User-aware attention reduces the effect of noisy behaviors contributed by similar users.
The full method produces significantly higher CTR prediction accuracy than existing sequential models on both short-term and long-term benchmark datasets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The retrieval-plus-augmentation pattern could be tested in other sequential recommendation settings such as session-based or next-item prediction where personal histories are also limited.
If embedding-based similarity proves brittle on certain data, alternative similarity measures like explicit collaborative filtering scores might serve as drop-in replacements for the retrieval step.
Distinguishing users inside the sequence via encoding and attention offers a general route for injecting collaborative signals into deep sequential models without full retraining.

Load-bearing premise

Behavior embeddings can reliably locate similar users whose added sequences supply useful preference signal rather than noise that the attention cannot sufficiently control.

What would settle it

An ablation experiment on the benchmark datasets where similar-user sequences are still retrieved and concatenated but the user-aware attention is replaced by standard target attention, and the resulting model performs no better than or worse than non-augmented baselines.

Figures

Figures reproduced from arXiv: 2604.23810 by Defu Lian, Haoyi Zhao, Xiaolong Chen, Xu Huang.

**Figure 1.** Figure 1: Logloss for user groups with various lengths of view at source ↗

**Figure 2.** Figure 2: Overall framework of SUIN. The left part depicts similar-user retrieval. Similar users’ behavior sequences are view at source ↗

**Figure 3.** Figure 3: A Toy example illustrating different positional en view at source ↗

**Figure 5.** Figure 5: Illustration of the user-aware target attention view at source ↗

**Figure 6.** Figure 6: Ablation study on Electronics and Kindle Store. view at source ↗

**Figure 7.** Figure 7: Performance of SUIN with 1 to 6 similar users on view at source ↗

**Figure 8.** Figure 8: Performance improvement over TIN across differ view at source ↗

read the original abstract

Click-through rate (CTR) prediction is one of the core tasks in recommender systems. User behavior sequences, as one of the most effective features, can accurately reflect user preferences and significantly improve prediction accuracy. Richer behavior sequences often enable more comprehensive user profiling, and recent studies have shown that scaling the length of user behavior sequence can yield substantial gains in CTR. However, due to the widespread sparsity in recommender systems, incomplete behavior sequences are common in real-world scenarios. Existing sequential modeling methods often rely solely on the target user's own behavior, and therefore struggle in such scenarios. This paper proposes a novel method called SUIN (Similar Users-augmented Interest Network), which enhances the target user's behavior sequence with behaviors from similar users to enhance the user profile for CTR prediction. Specifically, we use behavior embeddings encoded by a sequence encoder to retrieve users with similar behaviors from a user retrieval pool. The behavior sequences of these similar users are then concatenated with that of the target user in descending order of similarity to construct an augmented sequence. Given that the augmented sequence contains behaviors from multiple users, we propose a user-specific target-aware position encoding, which identifies the source user of each behavior and captures its relative position to the target item. Furthermore, to mitigate the empirically observed noise in similar users' behaviors, we design a user-aware target attention that jointly considers item-item and user-user correlations, fully exploiting the potential of the augmented behavior sequence. Comprehensive experiments on widely-used short-term and long-term sequence benchmark datasets demonstrate that our method significantly outperforms state-of-the-art sequential CTR models.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

SUIN augments target user sequences with similar users' behaviors via embedding retrieval, then uses custom position encoding and joint attention to handle the mix, but the gains rest on whether that retrieval adds aligned preference signal rather than noise.

read the letter

The paper's main move is to retrieve users whose behavior sequences look similar in embedding space, concatenate those sequences to the target user's in similarity order, and then run a modified transformer-style model on the longer combined sequence for CTR prediction. The two concrete additions are a user-specific target-aware position encoding that tags each behavior with its source user and its distance to the target item, plus a user-aware target attention that mixes item-item and user-user correlations to down-weight mismatched behaviors from the added users. This is a direct response to the sparsity problem that limits most sequential CTR models when histories are short. The architecture is described clearly enough that someone could reimplement the retrieval-plus-augmentation step without much guesswork. They evaluate on standard short-term and long-term sequence benchmarks and state that the method beats prior sequential CTR baselines. That is the part worth noting: it is a practical engineering proposal rather than a purely theoretical tweak. The central risk is exactly the one in the stress-test note. If the behavior embeddings mostly surface users who share popular items or temporal patterns rather than overlapping latent preferences, the added tokens are mostly noise. The attention layer is then asked to recover the signal without any explicit supervision on which behaviors are useful for the target user. The abstract does not show an ablation that replaces the similarity retrieval with random users or with a popularity baseline, nor does it report how often the retrieved sequences actually improve the target user's prediction on held-out clicks. Without those checks, the reported gains could be driven more by longer effective sequence length than by genuinely relevant augmentation. The number of similar users retrieved is also a free parameter that likely needs per-dataset tuning. For someone building production CTR models that already use sequence encoders, the paper supplies a concrete pattern to try when histories are too short. It is not a foundational result, but the design choices are specific enough to be worth testing. I would send it to peer review so the experiments can be examined for ablations and failure cases, rather than desk-rejecting it outright.

Referee Report

2 major / 2 minor

Summary. The paper proposes SUIN (Similar Users-augmented Interest Network) for CTR prediction in recommender systems. It retrieves similar users via sequence embeddings from a user retrieval pool, concatenates their behavior sequences with the target user's in descending similarity order to form an augmented sequence, introduces user-specific target-aware position encoding to identify user sources and relative positions to the target item, and designs user-aware target attention that jointly models item-item and user-user correlations to mitigate noise. Comprehensive experiments on short-term and long-term sequence benchmarks are claimed to show significant outperformance over state-of-the-art sequential CTR models.

Significance. If the results hold and the retrieval step adds net signal, the approach could meaningfully advance handling of sparse behavior sequences in CTR models by leveraging cross-user information, extending attention-based sequential models with multi-user augmentations and tailored encodings/attentions.

major comments (2)

[§3] §3 (Method): The central claim depends on behavior embeddings retrieving users whose sequences contain CTR-aligned preference information rather than noise; however, no independent verification, correlation analysis, or ablation isolating the retrieval quality (e.g., comparing embedding similarity to held-out CTR overlap) is provided, leaving the downstream position encoding and attention to implicitly filter without explicit supervision or evidence.
[Experiments] Experiments section (and abstract claim): While outperformance is asserted, the absence of specific quantitative metrics, baseline details, ablation studies on the number of similar users, or error analysis in the provided description makes it impossible to evaluate the magnitude, statistical reliability, or robustness of the reported gains.

minor comments (2)

[§3] Notation for the augmented sequence construction and the joint attention mechanism could be clarified with explicit equations or pseudocode to improve reproducibility.
[Abstract] The abstract would benefit from at least one quantitative result (e.g., relative improvement on a key metric) to substantiate the 'significantly outperforms' claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We appreciate the referee's thorough review and valuable suggestions for improving our manuscript on SUIN. We address the major comments point by point below, and indicate where revisions will be made to the paper.

read point-by-point responses

Referee: [§3] §3 (Method): The central claim depends on behavior embeddings retrieving users whose sequences contain CTR-aligned preference information rather than noise; however, no independent verification, correlation analysis, or ablation isolating the retrieval quality (e.g., comparing embedding similarity to held-out CTR overlap) is provided, leaving the downstream position encoding and attention to implicitly filter without explicit supervision or evidence.

Authors: We thank the referee for highlighting this important aspect. While the user-aware target attention mechanism is specifically designed to jointly model item-item and user-user correlations to mitigate noise from similar users' behaviors, we acknowledge that the manuscript lacks an explicit independent verification of the retrieval quality. In the revised version, we will add a new subsection or analysis in §3 or the experiments section that includes correlation analysis between embedding similarities and held-out CTR overlap, as well as an ablation study isolating the retrieval component. This will provide direct evidence supporting the central claim. revision: yes
Referee: [Experiments] Experiments section (and abstract claim): While outperformance is asserted, the absence of specific quantitative metrics, baseline details, ablation studies on the number of similar users, or error analysis in the provided description makes it impossible to evaluate the magnitude, statistical reliability, or robustness of the reported gains.

Authors: The full manuscript in Section 4 presents comprehensive experimental results, including specific quantitative metrics (e.g., AUC and LogLoss improvements), detailed baseline comparisons (such as against DIN, DIEN, BST, and other sequential models), and initial ablations on key components. However, we agree that additional studies on the number of similar users and error analysis would enhance the evaluation of robustness. We will revise the experiments section to include ablations varying the number of similar users (e.g., 1, 3, 5), statistical significance tests, and error analysis with case studies. These additions will allow for a clearer assessment of the gains' magnitude and reliability. revision: yes

Circularity Check

0 steps flagged

No circularity: architectural proposal validated on external benchmarks

full rationale

The paper presents SUIN as a new model architecture that retrieves similar users via behavior embeddings, augments the target sequence, applies user-specific position encoding, and uses joint item-user attention. All central claims are supported by empirical results on standard short-term and long-term CTR benchmark datasets rather than any internal derivation, fitted parameter renamed as prediction, or self-referential equation. No load-bearing step reduces the reported gains to quantities defined solely by the model's own inputs or prior self-citations; the retrieval and attention components are independently testable via the external evaluations.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 2 invented entities

Central claim rests on the domain assumption that similar-user behaviors provide net-positive signal after noise mitigation; no independent evidence for this is supplied in the abstract.

free parameters (1)

number of similar users retrieved
Controls length of augmented sequence; value chosen to balance added signal against noise.

axioms (1)

domain assumption Behavior embeddings from a sequence encoder can be used to retrieve meaningfully similar users
Invoked in the user retrieval step from the pool.

invented entities (2)

user-specific target-aware position encoding no independent evidence
purpose: Marks source user identity and relative position within the mixed-user augmented sequence
New component introduced to handle multi-user input.
user-aware target attention no independent evidence
purpose: Jointly models item-item and user-user correlations to reduce noise from similar users
New attention variant proposed for the augmented setting.

pith-pipeline@v0.9.0 · 5579 in / 1234 out tokens · 32910 ms · 2026-05-08T05:13:56.683883+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

66 extracted references · 12 canonical work pages · 6 internal anchors

[1]

Andreas Blattmann, Robin Rombach, Kaan Oktay, Jonas Müller, and Björn Ommer
[2]

Retrieval-augmented diffusion models.Advances in Neural Information Processing Systems35 (2022), 15309–15324

2022
[3]

Tesfaye Fenta Boka, Zhendong Niu, and Rama Bastola Neupane. 2024. A sur- vey of sequential recommendation systems: Techniques, evaluation, and future directions.Information Systems125 (2024), 102427

2024
[4]

Sebastian Borgeaud, Arthur Mensch, Jordan Hoffmann, Trevor Cai, Eliza Ruther- ford, Katie Millican, George Bm Van Den Driessche, Jean-Baptiste Lespiau, Bog- dan Damoc, Aidan Clark, et al. 2022. Improving language models by retrieving from trillions of tokens. InInternational conference on machine learning. PMLR, 2206–2240

2022
[5]

Yue Cao, Xiaojiang Zhou, Jiaqi Feng, Peihao Huang, Yao Xiao, Dayao Chen, and Sheng Chen. 2022. Sampling is all you need on modeling long-term user behaviors for CTR prediction. InProceedings of the 31st ACM International Conference on Information & Knowledge Management. 2974–2983

2022
[6]

Jianxin Chang, Chenbin Zhang, Zhiyi Fu, Xiaoxue Zang, Lin Guan, Jing Lu, Yiqun Hui, Dewei Leng, Yanan Niu, Yang Song, et al. 2023. TWIN: TWo-stage interest network for lifelong user behavior modeling in CTR prediction at kuaishou. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 3785–3794

2023
[7]

Junxuan Chen, Baigui Sun, Hao Li, Hongtao Lu, and Xian-Sheng Hua. 2016. Deep CTR Prediction in Display Advertising. InProceedings of the 24th ACM International Conference on Multimedia (MM ’16). Association for Computing Machinery, 811–820

2016
[8]

Qiwei Chen, Changhua Pei, Shanshan Lv, Chao Li, Junfeng Ge, and Wenwu Ou
[9]

End-to-end user behavior retrieval in click-through rateprediction model.arXiv preprint arXiv:2108.04468, 2021

End-to-end user behavior retrieval in click-through rateprediction model. arXiv preprint arXiv:2108.04468(2021)

work page arXiv 2021
[10]

Qiwei Chen, Huan Zhao, Wei Li, Pipei Huang, and Wenwu Ou. 2019. Behavior sequence transformer for e-commerce recommendation in alibaba. InProceedings of the 1st international workshop on deep learning practice for high-dimensional sparse data. 1–4

2019
[11]

Wenhu Chen, Hexiang Hu, Chitwan Saharia, and William W Cohen. 2022. Re-imagen: Retrieval-augmented text-to-image generator.arXiv preprint arXiv:2209.14491(2022)

work page arXiv 2022
[12]

Heng-Tze Cheng, Levent Koc, Jeremiah Harmsen, Tal Shaked, Tushar Chandra, Hrishi Aradhye, Glen Anderson, Greg Corrado, Wei Chai, Mustafa Ispir, Rohan Anil, Zakaria Haque, Lichan Hong, Vihan Jain, Xiaobing Liu, and Hemal Shah
[13]

InProceedings of the 1st Workshop on Deep Learning for Recommender Systems (DLRS 2016)

Wide & Deep Learning for Recommender Systems. InProceedings of the 1st Workshop on Deep Learning for Recommender Systems (DLRS 2016). Association for Computing Machinery, 7–10

2016
[14]

Kounianhua Du, Weinan Zhang, Ruiwen Zhou, Yangkun Wang, Xilong Zhao, Jiarui Jin, Quan Gan, Zheng Zhang, and David P Wipf. 2022. Learning enhanced representation for tabular data via neighborhood propagation.Advances in neural information processing systems35 (2022), 16373–16384

2022
[15]

Shereen Elsayed, Ahmed Rashed, and Lars Schmidt-Thieme. 2024. Multi- Behavioral Sequential Recommendation. InProceedings of the 18th ACM con- ference on recommender systems. 902–906

2024
[16]

Wenqi Fan, Yujuan Ding, Liangbo Ning, Shijie Wang, Hengyun Li, Dawei Yin, Tat-Seng Chua, and Qing Li. 2024. A survey on rag meeting llms: Towards retrieval-augmented large language models. InProceedings of the 30th ACM SIGKDD conference on knowledge discovery and data mining. 6491–6501

2024
[17]

Yufei Feng, Fuyu Lv, Weichen Shen, Menghan Wang, Fei Sun, Yu Zhu, and Keping Yang. 2019. Deep session interest network for click-through rate prediction. In Proceedings of the 28th International Joint Conference on Artificial Intelligence (Macao, China)(IJCAI’19). AAAI Press, 2301–2307

2019
[18]

Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yixin Dai, Jiawei Sun, Haofen Wang, and Haofen Wang. 2023. Retrieval-augmented generation for large language models: A survey.arXiv preprint arXiv:2312.10997 2, 1 (2023)

work page internal anchor Pith review arXiv 2023
[19]

Xavier Glorot, Antoine Bordes, and Yoshua Bengio. 2011. Deep sparse rectifier neural networks. InProceedings of the fourteenth international conference on artificial intelligence and statistics. JMLR Workshop and Conference Proceedings, 315–323

2011
[20]

Lin Guan, Jia-Qi Yang, Zhishan Zhao, Beichuan Zhang, Bo Sun, Xuanyuan Luo, Jinan Ni, Xiaowen Li, Yuhang Qi, Zhifang Fan, et al. 2025. Make It Long, Keep It Fast: End-to-End 10k-Sequence Modeling at Billion Scale on Douyin.arXiv preprint arXiv:2511.06077(2025)

work page internal anchor Pith review arXiv 2025
[21]

Huifeng Guo, Ruiming Tang, Yunming Ye, Zhenguo Li, and Xiuqiang He. 2017. DeepFM: a factorization-machine based neural network for CTR prediction. In Proceedings of the 26th International Joint Conference on Artificial Intelligence (IJCAI’17). AAAI Press, 1725–1731

2017
[22]

Liang Guo, Wei Li, Lucy Liao, Huihui Cheng, Rui Zhang, Yu Shi, Yueming Wang, Yanzun Huang, Keke Zhai, Pengchao Wang, et al. 2025. Request-Only Optimiza- tion for Recommendation Systems.arXiv preprint arXiv:2508.05640(2025)

work page arXiv 2025
[23]

Wei Guo, Rong Su, Renhao Tan, Huifeng Guo, Yingxue Zhang, Zhirong Liu, Ruiming Tang, and Xiuqiang He. 2021. Dual graph enhanced embedding neural network for CTR prediction. InProceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining. 496–504

2021
[24]

Balázs Hidasi, Alexandros Karatzoglou, Linas Baltrunas, and Domonkos Tikk
[25]

Session-based recommendations with recurrent neural networks.arXiv preprint arXiv:1511.06939(2015)

work page internal anchor Pith review arXiv 2015
[26]

Guoqing Hu, An Zhang, Shuo Liu, Zhibo Cai, Xun Yang, and Xiang Wang. 2025. Alphafuse: Learn id embeddings for sequential recommendation in null space of language embeddings. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval. 1614–1623

2025
[27]

Junjie Huang, Guohao Cai, Jieming Zhu, Zhenhua Dong, Ruiming Tang, Weinan Zhang, and Yong Yu. 2024. Recall-Augmented Ranking: Enhancing Click-Through Rate Prediction Accuracy with Cross-Stage Data. InCompanion Proceedings of the ACM Web Conference 2024. 830–833

2024
[28]

Xu Huang, Defu Lian, Jin Chen, Liu Zheng, Xing Xie, and Enhong Chen. 2023. Cooperative retriever and ranker in deep recommenders. InProceedings of the ACM Web Conference 2023. 1150–1161

2023
[29]

Gautier Izacard, Patrick Lewis, Maria Lomeli, Lucas Hosseini, Fabio Petroni, Timo Schick, Jane Dwivedi-Yu, Armand Joulin, Sebastian Riedel, and Edouard Grave. 2023. Atlas: Few-shot learning with retrieval augmented language models. Journal of Machine Learning Research24, 251 (2023), 1–43

2023
[30]

Yuchin Juan, Yong Zhuang, Wei-Sheng Chin, and Chih-Jen Lin. 2016. Field- aware Factorization Machines for CTR Prediction. InProceedings of the 10th ACM Conference on Recommender Systems (RecSys ’16). Association for Computing Machinery, 43–50

2016
[31]

Wang-Cheng Kang and Julian McAuley. 2018. Self-attentive sequential recom- mendation. In2018 IEEE international conference on data mining (ICDM). IEEE, 197–206

2018
[32]

Yuxuan Lei, Xiaolong Chen, Defu Lian, Peiyan Zhang, Jianxun Lian, Chaozhuo Li, and Xing Xie. 2023. Practical Content-aware Session-based Recommendation: deep retrieve then shallow rank. InAmazon KDD Cup 2023 Workshop

2023
[33]

Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. 2020. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. arXiv:2005.11401 [cs.CL]

work page internal anchor Pith review arXiv 2020
[34]

Yushen Li, Jinpeng Wang, Tao Dai, Jieming Zhu, Jun Yuan, Rui Zhang, and Shu- Tao Xia. 2024. RAT: Retrieval-Augmented Transformer for Click-Through Rate Prediction. InCompanion Proceedings of the ACM Web Conference 2024 (WWW ’24). Association for Computing Machinery, 867–870

2024
[35]

Jianxun Lian, Xiaohuan Zhou, Fuzheng Zhang, Zhongxia Chen, Xing Xie, and Guangzhong Sun. 2018. xDeepFM: Combining Explicit and Implicit Feature Interactions for Recommender Systems. InProceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD ’18). Asso- ciation for Computing Machinery, 1754–1763

2018
[36]

Qi Liu, Kai Zheng, Rui Huang, Wuchao Li, Kuo Cai, Yuan Chai, Yanan Niu, Yiqun Hui, Bing Han, Na Mou, et al . 2024. RecFlow: An Industrial Full Flow Recommendation Dataset.arXiv preprint arXiv:2410.20868(2024)

work page arXiv 2024
[37]

Xiao Ma, Liqin Zhao, Guan Huang, Zhi Wang, Zelin Hu, Xiaoqiang Zhu, and Kun Gai. 2018. Entire space multi-task model: An effective approach for estimating post-click conversion rate. InThe 41st International ACM SIGIR Conference on Research & Development in Information Retrieval. 1137–1140

2018
[38]

Erxue Min, Yu Rong, Tingyang Xu, Yatao Bian, Da Luo, Kangyi Lin, Junzhou Huang, Sophia Ananiadou, and Peilin Zhao. 2022. Neighbour interaction based click-through rate prediction via graph-masked transformer. InProceedings of the 45th international ACM SIGIR conference on research and development in information retrieval. 353–362

2022
[39]

Qi Pi, Weijie Bian, Guorui Zhou, Xiaoqiang Zhu, and Kun Gai. 2019. Practice on Long Sequential User Behavior Modeling for Click-Through Rate Prediction. InProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD ’19). Association for Computing Machinery, 2671–2679

2019
[40]

Qi Pi, Guorui Zhou, Yujing Zhang, Zhe Wang, Lejian Ren, Ying Fan, Xiaoqiang Zhu, and Kun Gai. 2020. Search-based user interest modeling with lifelong sequential behavior data for click-through rate prediction. InProceedings of the 29th ACM International Conference on Information & Knowledge Management. 2685–2692

2020
[41]

Jiarui Qin, Weinan Zhang, Rong Su, Zhirong Liu, Weiwen Liu, Ruiming Tang, Xiuqiang He, and Yong Yu. 2021. Retrieval & Interaction Machine for Tabular Data Prediction.Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining(2021)

2021
[42]

Jiarui Qin, Jiachen Zhu, Bo Chen, Zhirong Liu, Weiwen Liu, Ruiming Tang, Rui Zhang, Yong Yu, and Weinan Zhang. 2022. Rankflow: Joint optimization of multi- stage cascade ranking systems as flows. InProceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. 814–824

2022
[43]

Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks.arXiv preprint arXiv:1908.10084(2019)

work page internal anchor Pith review arXiv 2019
[44]

Steffen Rendle. 2010. Factorization Machines. In2010 IEEE International Confer- ence on Data Mining. 995–1000. Similar Users-Augmented Interest Network Conference acronym ’XX, June 03–05, 2018, Woodstock, NY

2010
[45]

Zihua Si, Lin Guan, Zhongxiang Sun, Xiaoxue Zang, Jing Lu, Yiqun Hui, Xingchao Cao, Zeyu Yang, Yichen Zheng, Dewei Leng, Kai Zheng, Chenbin Zhang, Yanan Niu, Yang Song, and Kun Gai. 2024. TWIN V2: Scaling Ultra-Long User Behav- ior Sequence Modeling for Enhanced CTR Prediction at Kuaishou(CIKM ’24). Association for Computing Machinery, 4890–4897

2024
[46]

Oumaima Stitini, Soulaimane Kaloun, and Omar Bencharef. 2022. Investigating different similarity metrics used in various recommender systems types: Scenario cases.The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences48 (2022), 187–193

2022
[47]

Fei Sun, Jun Liu, Jian Wu, Changhua Pei, Xiao Lin, Wenwu Ou, and Peng Jiang
[48]

InProceedings of the 28th ACM international conference on information and knowledge management

BERT4Rec: Sequential recommendation with bidirectional encoder rep- resentations from transformer. InProceedings of the 28th ACM international conference on information and knowledge management. 1441–1450
[49]

Ruoxi Wang, Bin Fu, Gang Fu, and Mingliang Wang. 2017. Deep & Cross Net- work for Ad Click Predictions. InProceedings of the ADKDD’17 (ADKDD’17). Association for Computing Machinery, Article 12

2017
[50]

Shoujin Wang, Liang Hu, Yan Wang, Longbing Cao, Quan Z Sheng, and Mehmet Orgun. 2019. Sequential recommender systems: challenges, progress and prospects.arXiv preprint arXiv:2001.04830(2019)

work page arXiv 2019
[51]

Wei Wei, Chao Huang, Lianghao Xia, Yong Xu, Jiashu Zhao, and Dawei Yin. 2022. Contrastive meta learning with behavior multiplicity for recommendation. In Proceedings of the fifteenth ACM international conference on web search and data mining. 1120–1128

2022
[52]

Jiancan Wu, Xiang Wang, Fuli Feng, Xiangnan He, Liang Chen, Jianxun Lian, and Xing Xie. 2021. Self-supervised graph learning for recommendation. InProceed- ings of the 44th international ACM SIGIR conference on research and development in information retrieval. 726–735

2021
[53]

Jun Xiao, Hao Ye, Xiangnan He, Hanwang Zhang, Fei Wu, and Tat-Seng Chua
[54]

AAAI Press, 3119–3125

Attentional factorization machines: learning the weight of feature interac- tions via attention networks(IJCAI’17). AAAI Press, 3119–3125
[55]

Zhibo Xiao, Luwei Yang, Wen Jiang, Yi Wei, Yi Hu, and Hao Wang. 2020. Deep multi-interest network for click-through rate prediction. InProceedings of the 29th ACM international conference on information & knowledge management. 2265–2268

2020
[56]

Xu Xie, Fei Sun, Zhaoyang Liu, Shiwen Wu, Jinyang Gao, Jiandong Zhang, Bolin Ding, and Bin Cui. 2022. Contrastive learning for sequential recommendation. In 2022 IEEE 38th international conference on data engineering (ICDE). IEEE, 1259– 1273

2022
[57]

Yueqi Xie, Peilin Zhou, and Sunghun Kim. 2022. Decoupled side information fusion for sequential recommendation. InProceedings of the 45th international ACM SIGIR conference on research and development in information retrieval. 1611– 1621

2022
[58]

Xiaoyong Yang, Yadong Zhu, Yi Zhang, Xiaobo Wang, and Quan Yuan. 2020. Large scale product graph construction for recommendation in e-commerce. arXiv preprint arXiv:2010.05525(2020)

work page arXiv 2020
[59]

Jun Yuan, Guohao Cai, and Zhenhua Dong. 2025. A Contextual-Aware Position Encoding for Sequential Recommendation. InCompanion Proceedings of the ACM on Web Conference 2025. 577–585

2025
[60]

Yanzhao Zhang, Mingxin Li, Dingkun Long, Xin Zhang, Huan Lin, Baosong Yang, Pengjun Xie, An Yang, Dayiheng Liu, Junyang Lin, et al. 2025. Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models.arXiv preprint arXiv:2506.05176(2025)

work page internal anchor Pith review arXiv 2025
[61]

Lei Zheng, Ning Li, Xianyu Chen, Quan Gan, and Weinan Zhang. 2023. Dense Representation Learning and Retrieval for Tabular Data Prediction(KDD ’23). Association for Computing Machinery, 3559–3569

2023
[62]

Guorui Zhou, Na Mou, Ying Fan, Qi Pi, Weijie Bian, Chang Zhou, Xiaoqiang Zhu, and Kun Gai. 2019. Deep interest evolution network for click-through rate prediction. InProceedings of the AAAI conference on artificial intelligence, Vol. 33. 5941–5948

2019
[63]

Guorui Zhou, Xiaoqiang Zhu, Chenru Song, Ying Fan, Han Zhu, Xiao Ma, Yanghui Yan, Junqi Jin, Han Li, and Kun Gai. 2018. Deep interest network for click-through rate prediction. InProceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining. 1059–1068

2018
[64]

Haolin Zhou, Junwei Pan, Xinyi Zhou, Xihua Chen, Jie Jiang, Xiaofeng Gao, and Guihai Chen. 2024. Temporal Interest Network for User Response Prediction. In Companion Proceedings of the ACM Web Conference 2024. 413–422

2024
[65]

Kun Zhou, Hui Yu, Wayne Xin Zhao, and Ji-Rong Wen. 2022. Filter-enhanced MLP is all you need for sequential recommendation. InProceedings of the ACM web conference 2022. 2388–2399

2022
[66]

Jieming Zhu, Jinyang Liu, Shuai Yang, Qi Zhang, and Xiuqiang He. 2021. Open benchmarking for click-through rate prediction. InProceedings of the 30th ACM international conference on information & knowledge management. 2759–2769. Received 20 February 2007; revised 12 March 2009; accepted 5 June 2009

2021