SDM: Sequential Deep Matching Model for Online Large-scale Recommender System

Changlong Yu; Fei Sun; Fuyu Lv; Keping Yang; Quan Lin; Taiwei Jin; Wilfred Ng

REVIEW 3 major objections 5 minor 39 references

SDM: Sequential Deep Matching Model for Online Large-scale Recommender System

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read SDM combines multi-head self-attention with a gated fusion of short- and long-term behavior to outperform prior matching methods and improve Taobao's online metrics.

desk verdict SDM is a plausible industrial matching model with a real evaluation flaw: the best variant is trained on next-5 targets while baselines predict next-1, and the best epochs are picked on the test set—so the claimed superiority over baselines is not cleanly isolated, though the paper is stronger than the reader's rationale suggests. read the letter →

arxiv 1909.00385 v2 pith:3XCG3ONX submitted 2019-09-01 cs.IR

Fuyu Lv , Taiwei Jin , Changlong Yu , Fei Sun , Quan Lin , Keping Yang , Wilfred Ng This is my paper · ORCID

classification cs.IR

keywords deepmatchingsequentialrecommendationmulti-headself-attentiongatedfusionlong-termandshort-termpreferencessession-basede-commercesampledsoftmax

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to show that a matching-stage recommender can capture a user's evolving preferences by modeling the user's latest browsing session and long-term history separately and then fusing them with a learned gate. The proposed SDM model uses multi-head self-attention over an LSTM-encoded session to represent several simultaneous interest tendencies, and aggregates long-term behaviors across item, category, brand, and shop features before a gated fusion combines the two views. Offline experiments on Taobao and JD data report that SDM and its variants beat item-based collaborative filtering and several sequential recommendation baselines, and the deployed variant PSDMMAL-N improved pCTR, pGMV, and discovery in an online A/B test. The practical stakes are large because the matching stage determines which candidates a ranker can even consider.

What carries the argument

The mechanism that carries the argument is a two-path encoder with a gated fusion. The short-term path embeds each item with side information (item ID, leaf category, first-level category, brand, shop), runs the sequence through an LSTM, applies multi-head self-attention to represent several simultaneous interest foci, and adds a user-profile attention layer. The long-term path aggregates up to twenty entries in each of the same feature types using attention with the user profile as query, then maps them through a dense layer. A gate vector $G_t^u = \sigma(W_1 e^u + W_2 s_t^u + W_3 p^u + b)$ computed from profile, short-term, and long-term vectors decides element-wise how much of each dimension comes from short-term versus long-term representation, giving $o_t^u = (1 - G_t^u) \odot p^u + G_t^u \odot s_t^u$. Training uses sampled softmax with the next item (or next five items for the best variant) as targets, and serving retrieves candidates by nearest-neighbor search over item embeddings.

What would settle it

Retrain every baseline under the same next-N objective, the same early-stopping rule, and the same hardware, and rerun the offline comparison on the released Taobao and JD datasets; if SDM's PSDMMAL-N no longer leads on HitRate and Recall, the paper's central claim of superiority collapses.

Watch

Extended reading notes

Core claim

The paper's central claim is that SDM, a sequential deep matching model, outperforms existing matching and sequence-aware recommendation methods on real-world e-commerce data, and that its best variant PSDMMAL-N improves commercial metrics when deployed in Taobao's matching stage. The claimed online gains, averaged over seven days, are 7.04% in pCTR, 4.50% in pGMV, and 24.37% in discovery, compared with the item-based CF baseline. Offline, PSDMMAL-N reaches 73.13% HitRate@100 on Taobao and 74.33% HitRate@20 on JD, leading the tables on the primary metric while not winning every auxiliary metric. The authors attribute the gains to two design choices: multi-head self-attention for multiple in-session interests and a gated fusion that keeps only long-term preferences correlated with the current session.

Load-bearing premise

The reported superiority assumes the baselines were compared fairly: the best SDM variant is trained to predict the next five items while baselines predict the next one item, and the best epoch is chosen on the test set, so the architecture alone may not be the cause.

Editorial extensions

If this is right

In the matching stage, replacing item-based CF with a sequence model that blends short- and long-term behavior can raise both relevance and novelty of retrieved candidates without a ranking-side change.
The multi-head attention is claimed to specialize: different heads track different product aspects within one session, so session representations can preserve competing intents rather than averaging them away.
The gated fusion is claimed to beat addition, concatenation, and multiplication fusions, because the gate can suppress long-term preferences unrelated to the current session.
Training the model to predict the next five items instead of the next one item is reported to improve offline recall and is used in the deployed variant.
The online A/B test shows that adding long-term behaviors to the short-term-only sequential model improves pCTR, pGMV, and discovery, meaning long-term preferences add signal beyond the current session.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A testable inference not settled by the paper: the next-N training target may account for a large share of the reported offline gains, since all baselines are trained for next-one prediction while the best variant PSDMMAL-N is trained with N=5; ablating the target length while holding the architecture fixed would separate the two effects.
The paper's own JD results, where the BINN baseline beats several SDM variants on some metrics, suggest the architecture's edge is not uniform across datasets; the gated fusion may help most where long-term behavior is rich, as in Taobao.
The serving design (precomputed item embeddings plus user-side network and KNN search) implies the approach transfers to other large-scale retrieval settings where a short context and a long profile must be combined online.
The head-specialization visualization suggests multi-head attention could serve as an interpretability tool for session-based recommenders, but the paper does not quantify whether specialization persists across users.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

You should know two things about this paper before deciding what to do with it. First, the core architecture is a reasonable industrial contribution: LSTM over the short session, multi-head self-attention to capture multiple interest facets, user attention on top, and a learned gate to fuse long-term preferences with the short-term vector. That specific combination for a matching task is not in the cited prior work, and the authors deploy it successfully at Taobao with reported gains in pCTR, pGMV, and discovery. Second, the offline comparison that supports the central claim is not fully controlled, and the reader's rationale contains a factual slip you should correct before trusting it.

What the paper does well: it releases code and offline datasets on GitHub, runs ablations on head count and fusion methods, and gives a clean problem formulation with session rules and evaluation metrics. The design choices are motivated with real examples rather than hand-waving. That is more reproducible than most CIKM industrial papers from this period.

The soft spot is in Table 2 and Section 5.1. PSDMMAL-N, the deployed variant, is trained with the next five items as target classes (N=5, following Tang and Wang), while all baselines and the core variant PSDMMAL are trained to predict only the next item. The offline evaluation uses a multi-item ground truth, so the next-5 objective is better aligned with the metric than the baselines' next-1 objective. The within-model comparison between PSDMMAL and PSDMMAL-N shows a meaningful jump in HitRate, which means part of the margin over baselines could come from the training objective, not the architecture. On top of that, Section 5.1 says the best results are selected across all training epochs on the test set. That is test-set selection, which inflates all models but may inflate the more flexible ones more. The online A/B test lacks confidence intervals or significance tests, so the 7.04% pCTR and 4.50% pGMV numbers are point estimates from seven days.

One correction to the reader's take: the claim that BINN beats PSDMMAL on JD is wrong. Table 2 shows PSDMMAL at 73.25% HitRate@20 vs BINN at 72.19% on JD, and PSDMMAL leads on all four JD metrics. The paper is not internally contradictory on that point; the issue is the unmatched training objective and the test-set selection, not an architecture reversal.

Overall, the paper is solid enough to deserve serious referee time. The evaluation protocol is fixable: retrain baselines with the same next-5 target, use a validation set for early stopping, and add error bars to the online results. The architecture is not circular and the reproducibility assets are real. I would engage with it.

For peer review: accept, with heavy revision requested on the experimental protocol.

Referee Report

3 major / 5 minor

Summary. The paper proposes SDM, a sequential deep matching model for large-scale recommender systems, combining a multi-head self-attention network over short-term session behaviors with a gated fusion of long-term behavior representations. The model is trained with sampled softmax and deployed in Taobao's matching stage. Offline experiments on two real-world e-commerce datasets (Taobao and JD) compare SDM variants against item-based CF, DNN, GRU4REC, NARM, SHAN, and BINN, reporting improvements in HitRate, Precision, Recall, and F1. An online A/B test reports average gains of 7.04% pCTR, 4.50% pGMV, and 24.37% discovery over seven days. The authors also provide ablation studies for the number of attention heads and fusion methods, and release their code and datasets.

Significance. If the reported superiority holds, the work is significant for industrial recommender matching: it introduces a practical architecture that addresses multiple interests in a session and a gated mechanism for fusing long- and short-term preferences, with deployment evidence on a large-scale platform. The release of code and datasets is a concrete strength that supports reproducibility. The multi-head attention and gated fusion ideas are plausible and well-motivated by the failure modes of single-attention and simple combination baselines. However, the central comparative claim is currently not fully supported because the key variant PSDMMAL-N is trained with a different target (next N items) than the baselines (next single item), and because the epoch selection procedure is performed on the test set. These issues need to be addressed before the superiority claim can be accepted.

major comments (3)

[§4.3 and Table 2] The comparison in Table 2 is not controlled for the training target: PSDMMAL-N is trained with the following N=5 items as target classes (as stated in §4.3), while all baselines (Item-based CF, DNN, GRU4REC, NARM, SHAN, BINN) are trained to predict only the next single item. Since the offline evaluation (Eq. 4.2.1) computes HitRate, Precision, Recall, and F1 against a multi-item ground-truth set, the improved numbers of PSDMMAL-N may reflect better alignment between its training objective and the evaluation protocol rather than the SDM architecture. To establish the claimed superiority, the authors should either train the baselines with the same next-N objective or report PSDMMAL under the single-next-item setting.
[§5.1] The sentence "We select the best results from all the training epochs of these models" indicates that the best epoch is chosen on the test set. This is a test-set selection procedure that inflates the reported performance and compromises a fair comparison across methods. The authors should select epochs and hyperparameters on a validation split and report test performance for the selected models.
[§5.2 and Figure 3] The online A/B test reports average improvements of 7.04%, 4.50%, and 24.37% for pCTR, pGMV, and discovery over seven days, but no confidence intervals, significance tests, or traffic-split information are provided. Without such statistical details, the online claims are not supported beyond descriptive trends; the authors should report standard errors and a significance test, or at least the number of users and the experiment duration.

minor comments (5)

[§5.3] The statement "the results become worse dramatically" is imprecise; please provide the exact numerical differences or a statistical comparison for the head-count ablation.
[§5.4] The text refers to "shown in Figure 4" when discussing fusion methods, but the corresponding comparison appears to be Table 4; please correct the cross-reference.
[§3.4 and §3.6] The symbol o_u^t is used both for the LSTM output gate in Eq. (5) and for the final user behavior vector in Eq. (14); consider renaming one of them to avoid confusion.
[§4.2.2] The definition of the discovery metric would benefit from a clear equation and an explicit statement of the time window over which "new" is defined.
[Table 3] The head-number ablation is performed on PSDMMAL only; please clarify whether the conclusion about the optimal number of heads transfers to PSDMMAL-N, the deployed variant.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: SDM is a trained neural matching model evaluated on held-out data, with no target quantity reused as an input constraint.

full rationale

The paper's central claim is empirical rather than derivational: a proposed matching architecture (LSTM-encoded short-term session, multi-head self-attention, user attention, gated fusion of long-term behavior, sampled-softmax training) is trained on logged user interactions and then evaluated on held-out test-day data. The user behavior vector o_u^t in Eq. (14) is computed from learned representations and item embeddings, and the scoring rule in Eq. (1) is an inner product; neither is defined in terms of HitRate, Recall, Precision, pCTR, pGMV, or discovery. No parameter is fitted to the evaluation metric and then renamed as a prediction, and no target quantity is built into the model by construction. The only self-citation in the modeling discussion, Ref. [6] for the claim that ignoring session structure hurts performance, is contextual and not load-bearing; the model's components are justified by the paper's own experiments and by standard published techniques (Transformer-style attention, LSTM, gating). The offline comparison does have a real methodological weakness, namely that PSDMMAL-N is trained with a next-5-item target while baselines predict the next single item, and the best results are selected across training epochs on the test set, but this is an experimental-control and selection issue, not circularity: it does not make the reported gains equivalent to the training objective or to the baseline by construction. The online A/B results are reported on the authors' own production platform without independent audit, which is an external-validity concern, not an internal circular-derivation concern. No step in the paper's derivation chain reduces to its own inputs.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard deep-learning background and on several domain assumptions about session structure and fusion, plus hand-set hyperparameters. There are no invented physical entities. The learned weights are not counted as free parameters; instead the listed choices are those the paper tunes by hand and that materially affect the results.

free parameters (5)

Number of attention heads h = 4 offline, 8 online
Hand-tuned; Table 3 shows results vary with h and the paper concludes 4 heads are best offline.
N for next-N target classes = 5
Chosen by hand for the best variant PSDMMAL-N; this changes the training task relative to baselines and is central to the reported gains.
Session merge time threshold = 10 minutes
Hand-set rule in Section 3.1 defining short-term sessions; affects all inputs.
Maximum session length = 50
Hand-set in Section 3.1; sessions longer than 50 are split, changing sequence structure.
Long-term window and max subset size = 7 days; 20
Hand-set in Sections 3.1 and 4.1; defines the scope of long-term behavior used in the model.

assumptions (5)

domain assumption Users' interests in a shopping session are multiple and can be captured by multi-head self-attention.
Section 3.5.1 and Section 5.3; the short-term module of the model relies on this assumption.
domain assumption Long-term behaviors from the past 7 days, summarized by attention over side information subsets, contain information relevant to the current session.
Section 3.6; the gated fusion module assumes this representation is sufficient.
domain assumption The gate vector Gu_t = sigmoid(W1 eu + W2 su_t + W3 pu + b) interpolating between su_t and pu learns a better combination than concatenation or addition.
Section 3.6, Equations 13 and 14; this is a modeling assumption evaluated only empirically.
domain assumption Sampled softmax with the log-uniform sampler and cross-entropy loss provides a valid training signal for the top-N matching objective.
Section 3.2; training uses sampled softmax, a standard but approximate method.
domain assumption The offline test protocol, first 25% of sessions for input, remaining interactions as ground truth, duplicates removed, evaluates matching quality.
Section 4.1; the reported metrics depend on this protocol.

how reviews work

0 comments

Cite this review

Pith. "Pith review of SDM: Sequential Deep Matching Model for Online Large-scale Recommender System." pith.science (2026). https://pith.science/paper/3XCG3ONX

@misc{pith2026190900385,
  author       = {Pith},
  title        = {Pith review of: SDM: Sequential Deep Matching Model for Online Large-scale Recommender System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3XCG3ONX}},
  note         = {Machine review of arXiv:1909.00385}
}

read the original abstract

Capturing users' precise preferences is a fundamental problem in large-scale recommender system. Currently, item-based Collaborative Filtering (CF) methods are common matching approaches in industry. However, they are not effective to model dynamic and evolving preferences of users. In this paper, we propose a new sequential deep matching (SDM) model to capture users' dynamic preferences by combining short-term sessions and long-term behaviors. Compared with existing sequence-aware recommendation methods, we tackle the following two inherent problems in real-world applications: (1) there could exist multiple interest tendencies in one session. (2) long-term preferences may not be effectively fused with current session interests. Long-term behaviors are various and complex, hence those highly related to the short-term session should be kept for fusion. We propose to encode behavior sequences with two corresponding components: multi-head self-attention module to capture multiple types of interests and long-short term gated fusion module to incorporate long-term preferences. Successive items are recommended after matching between sequential user behavior vector and item embedding vectors. Offline experiments on real-world datasets show the superior performance of the proposed SDM. Moreover, SDM has been successfully deployed on online large-scale recommender system at Taobao and achieves improvements in terms of a range of commercial metrics.

Figures

Figures reproduced from arXiv: 1909.00385 by the authors.

**Figure 2.** Each i u t ∈ Su is embedded into a vector ei u t . Short-term representation s u t is encoded by LSTM and attention mechanism. We describe long-term behaviors Lu from various side information, i.e., item ID (Lu id ), first level category (Lu cate ), leaf category (Lu leaf ), brand (Lu br and ) and shop (Lu shop ). Long-term representation p u is encoded through attention and dense fullyconnected networks. s u t an… view at source ↗

**Figure 3.** Online performances of our models compared with baseline in 7 days during December 2018. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Visualization of attention weights (the last hidden [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

**Figure 5.** Figure 5: A short-term session S u and long-term behaviors Lu from a sampled user on our online system. R u is the set of items recommended by our model, which are also clicked by the user. generate the recommended item red wine decanter. The case shows our gate module has effec…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 20 canonical work pages

[1]

Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2014. Neural ma- chine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473 (2014)

arXiv 2014
[2]

Ting Bai, Pan Du, Wayne Xin Zhao, Ji-Rong Wen, and Jian-Yun Nie. 2019. A Long- Short Demands-Aware Model for Next-Item Recommendation. arXiv preprint arXiv:1903.00066 (2019)

work page Pith review arXiv 2019
[3]

Xu Chen, Hongteng Xu, Yongfeng Zhang, Jiaxi Tang, Yixin Cao, Zheng Qin, and Hongyuan Zha. 2018. Sequential recommendation with user memory networks. In WSDM. ACM, 108–116

2018
[4]

Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep neural networks for youtube recommendations. In RecSys. ACM, 191–198

2016
[5]

Disheng Dong, Xiaolin Zheng, Ruixun Zhang, and Yan Wang. 2018. Recurrent Collaborative Filtering for Unifying General and Sequential Recommender. In IJCAI. 3350–3356

work page 2018
[6]

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 IJCAI. 2301–2307

work page 2019
[7]

Balázs Hidasi, Alexandros Karatzoglou, Linas Baltrunas, and Domonkos Tikk
[8]

Jin Huang, Zhaochun Ren, Wayne Xin Zhao, Gaole He, Ji-Rong Wen, and Daxiang Dong. 2019. Taxonomy-aware multi-hop reasoning networks for sequential recommendation. In WSDM. ACM, 573–581

work page 2019

Show all 39 references

[9]

Jin Huang, Wayne Xin Zhao, Hongjian Dou, Ji-Rong Wen, and Edward Y Chang
[10]

Sébastien Jean, Kyunghyun Cho, Roland Memisevic, and Yoshua Bengio. 2014. On using very large target vocabulary for neural machine translation. arXiv preprint arXiv:1412.2007 (2014)

2014 arXiv
[11]

Jeff Johnson, Matthijs Douze, and Hervé Jégou. 2017. Billion-scale similarity search with GPUs. arXiv preprint arXiv:1702.08734 (2017)

2017 arXiv
[12]

Wang-Cheng Kang and Julian McAuley. 2018. Self-attentive sequential recom- mendation. In ICDE. IEEE, 197–206

2018
[13]

Yehuda Koren, Robert Bell, and Chris Volinsky. 2009. Matrix factorization tech- niques for recommender systems. Computer 8 (2009), 30–37

2009
[14]

Jing Li, Pengjie Ren, Zhumin Chen, Zhaochun Ren, Tao Lian, and Jun Ma. 2017. Neural attentive session-based recommendation. In CIKM. ACM, 1419–1428

2017
[15]

Zhi Li, Hongke Zhao, Qi Liu, Zhenya Huang, Tao Mei, and Enhong Chen. 2018. Learning from history and present: next-item recommendation via discrimina- tively exploiting user behaviors. In KDD. ACM, 1734–1743

2018
[16]

Greg Linden, Brent Smith, and Jeremy York. 2003. Amazon. com recommen- dations: Item-to-item collaborative filtering. IEEE Internet computing 1 (2003), 76–80

2003
[17]

Qiao Liu, Yifu Zeng, Refuoe Mokhosi, and Haibin Zhang. 2018. STAMP: short- term attention/memory priority model for session-based recommendation. In KDD. ACM, 1831–1839

2018
[18]

Minh-Thang Luong, Hieu Pham, and Christopher D Manning. 2015. Effec- tive approaches to attention-based neural machine translation. arXiv preprint arXiv:1508.04025 (2015)

2015 arXiv
[19]

Stephen Merity, Nitish Shirish Keskar, and Richard Socher. 2017. Regularizing and optimizing LSTM language models. arXiv preprint arXiv:1708.02182 (2017)

2017 arXiv
[20]

Massimo Quadrana, Alexandros Karatzoglou, Balázs Hidasi, and Paolo Cremonesi
[21]

Steffen Rendle, Christoph Freudenthaler, and Lars Schmidt-Thieme. 2010. Factor- izing personalized markov chains for next-basket recommendation. In WWW. ACM, 811–820

2010
[22]

Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl. 2001. Item-based collaborative filtering recommendation algorithms. In WWW. ACM, 285–295

2001
[23]

Jiaxi Tang, Francois Belletti, Sagar Jain, Minmin Chen, Alex Beutel, Can Xu, and Ed H Chi. 2019. Towards Neural Mixture Recommender for Long Range Dependent User Sequences. arXiv preprint arXiv:1902.08588 (2019)

2019 arXiv
[24]

Jiaxi Tang and Ke Wang. 2018. Personalized top-n sequential recommendation via convolutional sequence embedding. In WSDM. ACM, 565–573

2018
[25]

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In NIPS. 5998–6008

2017
[26]

Jizhe Wang, Pipei Huang, Huan Zhao, Zhibo Zhang, Binqiang Zhao, and Dik Lun Lee. 2018. Billion-scale Commodity Embedding for E-commerce Recommendation in Alibaba. arXiv preprint arXiv:1803.02349 (2018)

2018 arXiv
[27]

Pengfei Wang, Jiafeng Guo, Yanyan Lan, Jun Xu, Shengxian Wan, and Xueqi Cheng. 2015. Learning hierarchical representation model for nextbasket recom- mendation. In SIGIR. ACM, 403–412

2015
[28]

Yonghui Wu, Mike Schuster, Zhifeng Chen, Quoc V Le, Mohammad Norouzi, Wolfgang Macherey, Maxim Krikun, Yuan Cao, Qin Gao, Klaus Macherey, et al
[29]

Haochao Ying, Fuzhen Zhuang, Fuzheng Zhang, Yanchi Liu, Guandong Xu, Xing Xie, Hui Xiong, and Jian Wu. 2018. Sequential Recommender System based on Hierarchical Attention Networks. In IJCAI

2018
[30]

Rex Ying, Ruining He, Kaifeng Chen, Pong Eksombatchai, William L Hamilton, and Jure Leskovec. 2018. Graph Convolutional Neural Networks for Web-Scale Recommender Systems. arXiv preprint arXiv:1806.01973 (2018)

2018 arXiv
[31]

Fajie Yuan, Alexandros Karatzoglou, Ioannis Arapakis, Joemon M Jose, and Xi- angnan He. 2019. A Simple Convolutional Generative Network for Next Item Recommendation. In WSDM. ACM, 582–590

2019
[32]

Shuai Zhang, Yi Tay, Lina Yao, and Aixin Sun. 2018. Next item recommendation with self-attention. arXiv preprint arXiv:1808.06414 (2018)

2018 arXiv
[33]

Wei Zhao, Benyou Wang, Jianbo Ye, Yongqiang Gao, Min Yang, and Xiaojun Chen. 2018. PLASTIC: Prioritize Long and Short-term Information in Top-n Recommendation using Adversarial Training. In IJCAI. 3676–3682

2018
[34]

Han Zhu, Xiang Li, Pengye Zhang, Guozheng Li, Jie He, Han Li, and Kun Gai
[39]

Learning Tree-based Deep Model for Recommender Systems. InKDD. ACM, 1079–1088
[2015]

arXiv preprint arXiv:1511.06939 (2015)

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

2015 arXiv
[2016]

arXiv preprint arXiv:1609.08144 (2016)

Google’s neural machine translation system: Bridging the gap between human and machine translation. arXiv preprint arXiv:1609.08144 (2016)

2016 arXiv
[2017]

In RecSys

Personalizing session-based recommendations with hierarchical recurrent neural networks. In RecSys. ACM, 130–137
[2018]

In SIGIR

Improving sequential recommendation with knowledge-enhanced memory networks. In SIGIR. ACM, 505–514

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2014. Neural ma- chine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473 (2014)

arXiv 2014

[2] [2]

Ting Bai, Pan Du, Wayne Xin Zhao, Ji-Rong Wen, and Jian-Yun Nie. 2019. A Long- Short Demands-Aware Model for Next-Item Recommendation. arXiv preprint arXiv:1903.00066 (2019)

work page Pith review arXiv 2019

[3] [3]

Xu Chen, Hongteng Xu, Yongfeng Zhang, Jiaxi Tang, Yixin Cao, Zheng Qin, and Hongyuan Zha. 2018. Sequential recommendation with user memory networks. In WSDM. ACM, 108–116

2018

[4] [4]

Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep neural networks for youtube recommendations. In RecSys. ACM, 191–198

2016

[5] [5]

Disheng Dong, Xiaolin Zheng, Ruixun Zhang, and Yan Wang. 2018. Recurrent Collaborative Filtering for Unifying General and Sequential Recommender. In IJCAI. 3350–3356

work page 2018

[6] [6]

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 IJCAI. 2301–2307

work page 2019

[7] [7]

Balázs Hidasi, Alexandros Karatzoglou, Linas Baltrunas, and Domonkos Tikk

[8] [8]

Jin Huang, Zhaochun Ren, Wayne Xin Zhao, Gaole He, Ji-Rong Wen, and Daxiang Dong. 2019. Taxonomy-aware multi-hop reasoning networks for sequential recommendation. In WSDM. ACM, 573–581

work page 2019