pith. sign in

arxiv: 2605.19651 · v1 · pith:4UWHA274new · submitted 2026-05-19 · 💻 cs.IR

Divergence Meets Consensus: A Multi-Source Negative Sampling Framework for Sequential Recommendation

Yuanzi Li , Lingjie Wang , Jingyu Zhao , Zihang Tian , Yuhan Wang , Lei Wang , Xu Chen This is my paper

Pith reviewed 2026-05-20 02:18 UTC · model grok-4.3

classification 💻 cs.IR
keywords negative samplingsequential recommendationmulti-source frameworkdivergence re-rankingconsensus distillationimplicit feedbacklocal optimateacher-peer-self
0
0 comments X

The pith

Multi-source negative sampling breaks self-reinforcement loops in sequential recommendation training.

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

The paper introduces MDCNS, a Teacher-Peer-Self framework for negative sampling in sequential recommendation under implicit feedback. Standard self-guided hard negative sampling couples model updates to sampling and restricts choices to a narrow region of items, creating a cycle that pushes models into local optima while wasting computation on low-gain scores. MDCNS counters this by pulling negative signals from peer models and ensemble teachers, re-ranking candidates according to prediction differences, and distilling consensus knowledge through KL divergence. A sympathetic reader would see this as a way to train more stable and diverse models without proportional increases in cost.

Core claim

The authors claim that a multi-source scoring process drawing on peer and teacher models, followed by divergence re-ranking and consensus distillation, breaks the vicious cycle of self-guided negative sampling, increases item-space coverage, and improves the information yield of each scored candidate, yielding better performing sequential recommenders.

What carries the argument

The Teacher-Peer-Self framework that combines multi-source scoring to inject external signals, divergence re-ranking to promote diversity, and consensus distillation to align the model while lowering computational overhead.

If this is right

  • Sequential models can escape local optima by replacing purely internal negative selection with external signals from peers and teachers.
  • Divergence-based re-ranking expands the sampled item region and thereby supports better generalization to unseen sequences.
  • Consensus distillation converts the full-pool scoring step into a higher-value training signal rather than a pure cost.
  • The same three-component structure works across different backbone architectures without architecture-specific redesign.

Where Pith is reading between the lines

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

  • The same divergence-plus-consensus pattern could be tested in non-sequential recommendation or ranking settings where self-reinforcement is also observed.
  • If peer models can be chosen or updated cheaply, the extra scoring cost might remain small enough to allow larger candidate pools at training time.
  • Dynamic weighting between sources might further reduce any residual bias introduced by a fixed teacher ensemble.

Load-bearing premise

Peer models and ensemble teachers supply sufficiently independent negative signals that do not simply replicate or amplify the main model's biases.

What would settle it

Run the same five backbone models on the six datasets using only self-guided hard negatives versus MDCNS and measure whether recall, NDCG, or HR improve; absence of consistent gains would falsify the central performance claim.

Figures

Figures reproduced from arXiv: 2605.19651 by Jingyu Zhao, Lei Wang, Lingjie Wang, Xu Chen, Yuanzi Li, Yuhan Wang, Zihang Tian.

Figure 1
Figure 1. Figure 1: Illustration of the ZPD. It utilizes a "Teacher-Peer [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The overall framework of MDCNS. (1) Multi-source Scoring, which constructs a candidate pool and generates relevance [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Loss dynamics of MDCNS on four datasets. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
read the original abstract

Negative sampling is significant for training sequential recommendation models under implicit feedback. The predominant strategy, self-guided hard negative sampling, selects negatives based on the model's current state but suffers from three limitations: (1) the coupling between sampling and model updates triggers a vicious cycle that drives the model into local optima; (2) relying on current model parameters narrows sampling to a small region of the item space, reducing diversity and harming generalization; (3) identifying a hard negative requires scoring the entire candidate pool, causing substantial computational overhead with minimal information gain. To address these challenges, we propose MDCNS (Multi-source Divergence-Consensus for Negative Sampling), a novel "Teacher-Peer-Self" framework inspired by Vygotsky's Zone of Proximal Development (ZPD) theory. The proposed method comprises three components, including multi-source scoring, divergence re-ranking, and consensus distillation. Firstly, multi-source scoring incorporates peer and ensemble teacher models to inject external negative signals and break the self-reinforcement loop. Then, divergence re-ranking exploits prediction discrepancy between self and peer models to enhance sampling diversity. Finally, consensus distillation aligns the self model with the teacher via KL divergence, simultaneously improving computational cost utilization. Extensive experiments on six real-world datasets and five backbone models show that MDCNS consistently outperforms state-of-the-art negative sampling methods, demonstrating strong effectiveness and generalization.

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.

Referee Report

2 major / 2 minor

Summary. The paper proposes MDCNS, a 'Teacher-Peer-Self' negative sampling framework for sequential recommendation under implicit feedback. It uses multi-source scoring from peer and ensemble teacher models to break self-reinforcement, divergence re-ranking based on prediction discrepancies to increase diversity, and consensus distillation via KL divergence to improve efficiency. Experiments across six real-world datasets and five backbone models report consistent outperformance over state-of-the-art negative sampling methods.

Significance. If the reported gains are robust, the framework offers a practical route to mitigate local-optima issues and sampling narrowness in sequential recommenders by injecting external signals. The multi-dataset, multi-backbone evaluation is a positive empirical strength that supports claims of generalization, though the ZPD framing functions more as conceptual motivation than as a source of formal derivations or parameter-free results.

major comments (2)
  1. [Section 3.2] Section 3.2: The divergence re-ranking step assumes that prediction discrepancies between the self model and peer models supply sufficiently independent negative signals. Because peer and teacher models are trained on identical implicit-feedback data with comparable sequential architectures, score distributions are likely to correlate strongly; this risks reducing the re-ranking benefit to little more than generic ensembling, which would weaken the central claim that multi-source scoring breaks the self-reinforcement loop.
  2. [Experimental evaluation] Experimental evaluation: The abstract states consistent outperformance on six datasets and five backbones, yet the manuscript does not appear to report statistical significance tests (e.g., paired t-tests across runs) or component-wise ablations that isolate the contribution of divergence re-ranking versus simple multi-model averaging. Without these, it remains unclear whether gains arise specifically from the proposed ZPD-inspired mechanisms or from broader ensemble effects.
minor comments (2)
  1. [Section 3.3] The computational-cost argument for consensus distillation would benefit from an explicit comparison of wall-clock time or FLOPs against full-pool scoring baselines.
  2. [Section 3.3] Notation for the KL term in consensus distillation should be defined with respect to the exact output distributions of the self and teacher models.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We sincerely thank the referee for the thoughtful and constructive comments on our manuscript. We have carefully considered the concerns regarding the independence of negative signals in the divergence re-ranking and the need for additional experimental rigor. Below, we provide point-by-point responses and indicate the revisions we plan to make.

read point-by-point responses

  1. Referee: [Section 3.2] Section 3.2: The divergence re-ranking step assumes that prediction discrepancies between the self model and peer models supply sufficiently independent negative signals. Because peer and teacher models are trained on identical implicit-feedback data with comparable sequential architectures, score distributions are likely to correlate strongly; this risks reducing the re-ranking benefit to little more than generic ensembling, which would weaken the central claim that multi-source scoring breaks the self-reinforcement loop.

    Authors: We appreciate the referee's insightful observation on potential correlations among the models. In the MDCNS framework, the peer models are trained separately with distinct random seeds and training schedules, which introduces variability in their learned representations despite sharing the same data and architecture base. The divergence re-ranking is designed to capitalize on these discrepancies by re-ranking candidates based on the difference in prediction scores, thereby selecting negatives that are particularly challenging for the self-model but less so for the peers. This mechanism is intended to provide signals that help escape local optima beyond what a simple ensemble average would achieve. To substantiate this, we will include a new subsection in the revised manuscript analyzing the pairwise score correlations between the self, peer, and teacher models across datasets, along with examples of selected negatives to illustrate the diversity introduced. revision: yes

  2. Referee: [Experimental evaluation] Experimental evaluation: The abstract states consistent outperformance on six datasets and five backbones, yet the manuscript does not appear to report statistical significance tests (e.g., paired t-tests across runs) or component-wise ablations that isolate the contribution of divergence re-ranking versus simple multi-model averaging. Without these, it remains unclear whether gains arise specifically from the proposed ZPD-inspired mechanisms or from broader ensemble effects.

    Authors: We agree that reporting statistical significance and conducting component ablations are important for rigorously validating the contributions of each component. The current version reports mean performance metrics over multiple independent runs but omits formal statistical tests and detailed ablations. In the revision, we will add paired t-tests to assess the significance of improvements over baselines across all datasets and backbones. Additionally, we will include ablation studies that compare the full MDCNS against variants without divergence re-ranking (to isolate its effect from multi-source scoring) and without consensus distillation. These additions will help clarify that the performance gains stem from the proposed mechanisms rather than generic ensembling. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical framework with independent experimental validation

full rationale

The paper proposes MDCNS as a practical multi-source negative sampling method for sequential recommenders, consisting of multi-source scoring, divergence re-ranking, and consensus distillation. No mathematical derivation chain exists that reduces a claimed result to fitted parameters or self-citations by construction. The central claims rest on empirical outperformance across six datasets and five backbones rather than any closed-form prediction or uniqueness theorem. External benchmarks (real-world datasets) are used to evaluate the framework, and the method does not invoke self-citations as load-bearing justifications for its core components. This is a standard empirical contribution whose validity can be assessed directly from the reported experiments without circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on standard assumptions of implicit-feedback recommendation (user-item interactions are positive signals and unobserved items can serve as negatives) plus the domain assumption that external models can provide useful diversity signals. No new physical entities or ad-hoc mathematical constants are introduced in the abstract.

axioms (2)
  • domain assumption Implicit feedback provides reliable positive signals and unobserved items are valid negatives
    Standard premise in the sequential recommendation literature that the method inherits.
  • domain assumption Peer and teacher models supply sufficiently independent negative signals
    Invoked when claiming that multi-source scoring breaks the self-reinforcement loop.

pith-pipeline@v0.9.0 · 5790 in / 1441 out tokens · 32848 ms · 2026-05-20T02:18:14.008571+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages · 4 internal anchors

  1. [1]

    Jin Chen, Defu Lian, Binbin Jin, Kai Zheng, and Enhong Chen. 2022. Learning recommenders for implicit feedback with importance resampling. InProceedings of the ACM Web Conference 2022. 1997–2005

    work page 2022

  2. [2]

    Jiawei Chen, Can Wang, Sheng Zhou, Qihao Shi, Yan Feng, and Chun Chen. 2019. Samwalker: Social recommendation with informative sampling strategy. InThe world wide web conference. 228–239

    work page 2019

  3. [3]

    Ting Chen, Yizhou Sun, Yue Shi, and Liangjie Hong. 2017. On sampling strategies for neural network-based collaborative filtering. InProceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 767– 776

    work page 2017

  4. [4]

    Yongjun Chen, Zhiwei Liu, Jia Li, Julian McAuley, and Caiming Xiong. 2022. Intent Contrastive Learning for Sequential Recommendation.arXiv preprint arXiv:2202.02519(2022)

    work page arXiv 2022

  5. [5]

    Jingtao Ding, Yuhan Quan, Xiangnan He, Yong Li, and Depeng Jin. 2019. Re- inforced negative sampling for recommendation with exposure data.. InIJCAI. Macao, 2230–2236

    work page 2019

  6. [6]

    Jingtao Ding, Yuhan Quan, Quanming Yao, Yong Li, and Depeng Jin. 2020. Simplify and robustify negative sampling for implicit collaborative filtering.Advances in Neural Information Processing Systems33 (2020), 1094–1105

    work page 2020

  7. [7]

    Lu Fan, Jiashu Pu, Rongsheng Zhang, and Xiao-Ming Wu. 2023. Neighborhood- based hard negative mining for sequential recommendation. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. 2042–2046

    work page 2023

  8. [8]

    Chongming Gao, Shijun Li, Yuan Zhang, Jiawei Chen, Biao Li, Wenqiang Lei, Peng Jiang, and Xiangnan He. 2022. Kuairand: An unbiased sequential recom- mendation dataset with randomly exposed videos. InProceedings of the 31st ACM international conference on information & knowledge management. 3953–3957

    work page 2022

  9. [9]

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

    work page

  10. [10]

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

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  11. [11]

    Balázs Hidasi, Massimo Quadrana, Alexandros Karatzoglou, and Domonkos Tikk. 2016. Parallel recurrent neural network architectures for feature-rich session-based recommendations. InProceedings of the 10th ACM conference on recommender systems. 241–248

    work page 2016

  12. [12]

    Balázs Hidasi and Domonkos Tikk. 2016. General factorization framework for context-aware recommendations.Data Mining and Knowledge Discovery30, 2 (2016), 342–371

    work page 2016

  13. [13]

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. 2015. Distilling the knowledge in a neural network.arXiv preprint arXiv:1503.02531(2015)

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  14. [14]

    Yujing Hu, Qing Da, Anxiang Zeng, Yang Yu, and Yinghui Xu. 2018. Reinforce- ment learning to rank in e-commerce search engine: Formalization, analysis, and application. InProceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining. 368–377

    work page 2018

  15. [15]

    Jui-Ting Huang, Ashish Sharma, Shuying Sun, Li Xia, David Zhang, Philip Pronin, Janani Padmanabhan, Giuseppe Ottaviano, and Linjun Yang. 2020. Embedding- based retrieval in facebook search. InProceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2553–2561

    work page 2020

  16. [16]

    Tinglin Huang, Yuxiao Dong, Ming Ding, Zhen Yang, Wenzheng Feng, Xinyu Wang, and Jie Tang. 2021. Mixgcf: An improved training method for graph neural network-based recommender systems. InProceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining. 665–674

    work page 2021

  17. [17]

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

    work page 2018

  18. [18]

    Diederik P Kingma. 2014. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980(2014)

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  19. [19]

    Yuanzi Li, Xuri Ge, Jingyu Zhao, Yidan Wang, Jiyuan Yang, Zhumin Chen, Zhaochun Ren, and Xin Xin. 2026. R2NS: Recall and Re-ranking of Negative Sam- ples for Sequential Recommendation. InProceedings of the ACM Web Conference 2026(United Arab Emirates)(WWW ’26). Association for Computing Machinery, New York, NY, USA, 6331–6341. https://doi.org/10.1145/3774...

    work page doi:10.1145/3774904.3792360 2026

  20. [20]

    Chengkai Liu, Jianghao Lin, Jianling Wang, Hanzhou Liu, and James Caverlee

    work page

  21. [21]

    Mamba4rec: Towards efficient sequential recommendation with selective state space models.arXiv preprint arXiv:2403.03900(2024)

    work page arXiv 2024

  22. [22]

    Langming Liu, Liu Cai, Chi Zhang, Xiangyu Zhao, Jingtong Gao, Wanyu Wang, Yifu Lv, Wenqi Fan, Yiqi Wang, Ming He, et al. 2023. Linrec: Linear attention mechanism for long-term sequential recommender systems. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. 289–299

    work page 2023

  23. [23]

    Julian McAuley, Christopher Targett, Qinfeng Shi, and Anton Van Den Hengel

    work page

  24. [24]

    InProceedings of the 38th international ACM SIGIR conference on research and development in information retrieval

    Image-based recommendations on styles and substitutes. InProceedings of the 38th international ACM SIGIR conference on research and development in information retrieval. 43–52

    work page

  25. [25]

    Dae Hoon Park and Yi Chang. 2019. Adversarial sampling and training for semi- supervised information retrieval. InThe World Wide Web Conference. 1443–1453

    work page 2019

  26. [26]

    Tao Qi, Fangzhao Wu, Chuhan Wu, Yongfeng Huang, and Xing Xie. 2020. Privacy- preserving news recommendation model learning.arXiv preprint arXiv:2003.09592 (2020)

    work page arXiv 2020

  27. [27]

    Ruihong Qiu, Zi Huang, Hongzhi Yin, and Zijian Wang. 2022. Contrastive learning for representation degeneration problem in sequential recommendation. InProceedings of the fifteenth ACM international conference on web search and data mining. 813–823

    work page 2022

  28. [28]

    Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme

    work page

  29. [29]

    BPR: Bayesian personalized ranking from implicit feedback.arXiv preprint arXiv:1205.2618(2012)

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  30. [30]

    Steffen Rendle, Christoph Freudenthaler, and Lars Schmidt-Thieme. 2010. Factor- izing personalized markov chains for next-basket recommendation. InProceedings of the 19th international conference on World wide web. 811–820

    work page 2010

  31. [31]

    Karim Shabani, Mohamad Khatib, and Saman Ebadi. 2010. Vygotsky’s zone of proximal development: Instructional implications and teachers’ professional development.English language teaching3, 4 (2010), 237–248

    work page 2010

  32. [32]

    Claude E Shannon. 1948. A mathematical theory of communication.The Bell system technical journal27, 3 (1948), 379–423

    work page 1948

  33. [33]

    Wentao Shi, Jiawei Chen, Fuli Feng, Jizhi Zhang, Junkang Wu, Chongming Gao, and Xiangnan He. 2023. On the theories behind hard negative sampling for recommendation. InProceedings of the ACM Web Conference 2023. 812–822

    work page 2023

  34. [34]

    Fei Sun, Jun Liu, Jian Wu, Changhua Pei, Xiao Lin, Wenwu Ou, and Peng Jiang. 2019. BERT4Rec: Sequential Recommendation with Bidirectional En- coder Representations from Transformer. arXiv:1904.06690 [cs.IR] https: //arxiv.org/abs/1904.06690

    work page arXiv 2019

  35. [35]

    Jiaxi Tang and Ke Wang. 2018. Personalized Top-N Sequential Recommen- dation via Convolutional Sequence Embedding.Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining(2018). https: //api.semanticscholar.org/CorpusID:39847715

    work page 2018

  36. [36]

    Tim Van Erven and Peter Harremos. 2014. Rényi divergence and Kullback-Leibler divergence.IEEE Transactions on Information Theory60, 7 (2014), 3797–3820

    work page 2014

  37. [37]

    Jun Wang, Lantao Yu, Weinan Zhang, Yu Gong, Yinghui Xu, Benyou Wang, Peng Zhang, and Dell Zhang. 2017. Irgan: A minimax game for unifying generative and discriminative information retrieval models. InProceedings of the 40th In- ternational ACM SIGIR conference on Research and Development in Information Retrieval. 515–524

    work page 2017

  38. [38]

    Xin Wang, Yudong Chen, and Wenwu Zhu. 2021. A Survey on Curriculum Learning. arXiv:2010.13166 [cs.LG] https://arxiv.org/abs/2010.13166

    work page arXiv 2021

  39. [39]

    Yidan Wang, Xuri Ge, Xin Chen, Ruobing Xie, Su Yan, Xu Zhang, Zhumin Chen, Jun Ma, and Xin Xin. 2025. Exploration and Exploitation of Hard Negative Samples for Cross-Domain Sequential Recommendation. InProceedings of the Eighteenth ACM International Conference on Web Search and Data Mining(Han- nover, Germany)(WSDM ’25). Association for Computing Machiner...

    work page doi:10.1145/3701551.3703535 2025

  40. [40]

    Zhenlei Wang, Shiqi Shen, Zhipeng Wang, Bo Chen, Xu Chen, and Ji-Rong Wen. 2022. Unbiased sequential recommendation with latent confounders. In Proceedings of the ACM web conference 2022. 2195–2204

    work page 2022

  41. [41]

    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

    work page 2022

  42. [42]

    Lanling Xu, Jianxun Lian, Wayne Xin Zhao, Ming Gong, Linjun Shou, Daxin Jiang, Xing Xie, and Ji-Rong Wen. 2022. Negative sampling for contrastive representation learning: A review.arXiv preprint arXiv:2206.00212(2022)

    work page arXiv 2022

  43. [43]

    Jiyuan Yang, Yue Ding, Yidan Wang, Pengjie Ren, Zhumin Chen, Fei Cai, Jun Ma, Rui Zhang, Zhaochun Ren, and Xin Xin. 2024. Debiasing Sequential Recom- menders through Distributionally Robust Optimization over System Exposure. InProceedings of the 17th ACM International Conference on Web Search and Data Mining. 882–890

    work page 2024

  44. [44]

    Jiyuan Yang, Yuanzi Li, Jingyu Zhao, Hanbing Wang, Muyang Ma, Jun Ma, Zhaochun Ren, Mengqi Zhang, Xin Xin, Zhumin Chen, et al. 2024. Uncovering Se- lective State Space Model’s Capabilities in Lifelong Sequential Recommendation. arXiv preprint arXiv:2403.16371(2024)

    work page arXiv 2024

  45. [45]

    Rex Ying, Ruining He, Kaifeng Chen, Pong Eksombatchai, William L Hamilton, and Jure Leskovec. 2018. Graph convolutional neural networks for web-scale recommender systems. InProceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining. 974–983

    work page 2018

  46. [46]

    Fajie Yuan, Alexandros Karatzoglou, Ioannis Arapakis, Joemon M Jose, and Xi- angnan He. 2019. A simple convolutional generative network for next item recommendation. InProceedings of the twelfth ACM international conference on web search and data mining. 582–590

    work page 2019

  47. [47]

    Shuai Zhang, Lina Yao, Aixin Sun, and Yi Tay. 2019. Deep learning based recom- mender system: A survey and new perspectives.ACM computing surveys (CSUR) 52, 1 (2019), 1–38

    work page 2019

  48. [48]

    Weinan Zhang, Tianqi Chen, Jun Wang, and Yong Yu. 2013. Optimizing top-n collaborative filtering via dynamic negative item sampling. InProceedings of the 36th international ACM SIGIR conference on Research and development in information retrieval. 785–788

    work page 2013

  49. [49]

    Wayne Xin Zhao, Yupeng Hou, Xingyu Pan, Chen Yang, Zeyu Zhang, Zihan Lin, Jingsen Zhang, Shuqing Bian, Jiakai Tang, Wenqi Sun, et al. 2022. Recbole 2.0: SIGIR ’26, July 20–24, 2026, Melbourne, VIC, Australia Yuanzi Li et al. Towards a more up-to-date recommendation library. InProceedings of the 31st ACM international conference on information & knowledge ...

    work page 2022

  50. [50]

    Wayne Xin Zhao, Shanlei Mu, Yupeng Hou, Zihan Lin, Yushuo Chen, Xingyu Pan, Kaiyuan Li, Yujie Lu, Hui Wang, Changxin Tian, et al. 2021. Recbole: Towards a unified, comprehensive and efficient framework for recommendation algorithms. Inproceedings of the 30th acm international conference on information & knowledge management. 4653–4664

    work page 2021

  51. [51]

    Jin Peng Zhou, Ga Wu, Zheda Mai, and Scott Sanner. 2020. Noise contrastive estimation for autoencoding-based one-class collaborative filtering.arXiv preprint arXiv:2008.01246(2020)

    work page arXiv 2020

  52. [52]

    Kun Zhou, Hui Wang, Wayne Xin Zhao, Yutao Zhu, Sirui Wang, Fuzheng Zhang, Zhongyuan Wang, and Ji-Rong Wen. 2020. S3-rec: Self-supervised learning for se- quential recommendation with mutual information maximization. InProceedings of the 29th ACM international conference on information & knowledge management. 1893–1902

    work page 2020

  53. [53]

    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

    work page 2022

  54. [54]

    Rui Zhou, Xian Wu, Zhaopeng Qiu, Yefeng Zheng, and Xu Chen. 2023. Dis- tributionally robust sequential recommnedation. InProceedings of the 46th In- ternational ACM SIGIR Conference on Research and Development in Information Retrieval. 279–288

    work page 2023