pith. sign in

arxiv: 2412.18735 · v3 · submitted 2024-12-25 · 💻 cs.IR · cs.LG

Automatic Self-supervised Learning for Social Recommendations

Xin He , Wenqi Fan , Mingchen Sun , Ying Wang , Xin Wang This is my paper

Pith reviewed 2026-05-23 07:28 UTC · model grok-4.3

classification 💻 cs.IR cs.LG
keywords social recommendationsself-supervised learningmeta-learningautomatic weightingauxiliary tasksrepresentation learningrecommendation systems
0
0 comments X

The pith

AusRec uses meta-learning to automatically weight multiple self-supervised auxiliary tasks and improve social recommendation performance without manual task design.

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

The paper presents AusRec as a method that combines several self-supervised auxiliary tasks for social recommendations and lets a meta-learning framework decide how much each task should contribute. This removes the need for experts to hand-craft task weights for each new scenario. The approach is tested on real-world datasets and shown to beat prior methods across different recommendation settings. The core idea is that adaptive weighting during training leads to stronger user and item representations.

Core claim

AusRec integrates multiple self-supervised auxiliary tasks with an automatic weighting mechanism that uses a meta-learning optimization framework to adaptively balance their contributions, enabling the model to automatically learn the optimal importance of each auxiliary task and thereby enhance representation learning in social recommendations.

What carries the argument

Meta-learning optimization framework that automatically learns and applies weights to balance multiple self-supervised auxiliary tasks during training.

If this is right

  • Social recommendation models no longer require scenario-specific manual design of auxiliary tasks.
  • Representation quality improves because task contributions are tuned automatically during optimization.
  • The same framework can be reused on new datasets without repeating the weight-tuning process.
  • Performance gains hold across multiple real-world social recommendation collections.

Where Pith is reading between the lines

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

  • The method could be tested on non-social recommendation tasks that also rely on auxiliary self-supervised signals.
  • If the meta-learner itself needs little extra data, the approach may lower the barrier for deploying self-supervised recsys in new domains.
  • Combining this weighting scheme with graph-based social models might further reduce sensitivity to noisy social links.
  • The automatic balance could reveal which auxiliary tasks are most useful in practice, guiding future task design even if manual intervention is later reintroduced.

Load-bearing premise

The meta-learning optimization framework can reliably determine optimal task weights that generalize across different recommendation scenarios and datasets without introducing instability or requiring extensive additional tuning.

What would settle it

On a new dataset, training with the meta-learned weights produces lower recommendation accuracy than training with fixed equal weights or with weights chosen by grid search.

Figures

Figures reproduced from arXiv: 2412.18735 by Mingchen Sun, Wenqi Fan, Xin He, Xin Wang, Ying Wang.

Figure 1
Figure 1. Figure 1: Examples of various social relations in social networks. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Performance of multiple self-supervised auxiliary tasks for the advanced social [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The overall framework of the proposed method. The parameters updating [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The change of automatic weights on various SS-A tasks during training (Best [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗
read the original abstract

In recent years, researchers have leveraged social relations to enhance recommendation performance. However, most existing social recommendation methods require carefully designed auxiliary social tasks tailored to specific scenarios, which depend heavily on domain knowledge and expertise. To address this limitation, we propose Automatic Self-supervised Learning for Social Recommendations (AusRec), which integrates multiple self-supervised auxiliary tasks with an automatic weighting mechanism to adaptively balance their contributions through a meta-learning optimization framework. This design enables the model to automatically learn the optimal importance of each auxiliary task, thereby enhancing representation learning in social recommendations. Extensive experiments on several real-world datasets demonstrate that AusRec consistently outperforms state-of-the-art baselines, validating its effectiveness and robustness across different recommendation scenarios.

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 manuscript proposes AusRec, which integrates multiple self-supervised auxiliary tasks for social recommendation with an automatic weighting mechanism based on a meta-learning optimization framework. The central claim is that this design allows the model to automatically learn optimal task importances, enhancing representation learning, and that extensive experiments on real-world datasets show consistent outperformance over state-of-the-art baselines along with robustness across scenarios.

Significance. If the meta-learning component reliably produces stable, generalizable task weights without collapse or extra tuning, the work would meaningfully reduce reliance on manual, domain-specific auxiliary task design in social recommendation systems.

major comments (2)
  1. [Abstract] Abstract: the assertion of 'consistent outperformance' and 'robustness' is load-bearing for the central claim yet supplies no experimental details, error bars, baseline descriptions, ablation results, or dataset statistics; without these the claim cannot be evaluated.
  2. [Method] Meta-learning optimization framework (described in the method): the claim that the framework 'automatically learn[s] the optimal importance of each auxiliary task' and generalizes without instability is unsupported; no meta-objective formulation, stabilization techniques (e.g., gradient surgery), or ablation replacing the meta-learner with static/random weights is provided, leaving the weakest assumption untested.
minor comments (2)
  1. [Method] Notation for the auxiliary tasks and the meta-objective should be introduced with explicit equations rather than prose only.
  2. [Abstract] The abstract would be clearer if it named the specific self-supervised tasks employed.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and indicate the revisions we will make to the next version.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the assertion of 'consistent outperformance' and 'robustness' is load-bearing for the central claim yet supplies no experimental details, error bars, baseline descriptions, ablation results, or dataset statistics; without these the claim cannot be evaluated.

    Authors: We acknowledge that abstracts are concise summaries and do not contain the detailed statistics, error bars, or ablation results that appear in the body of the paper. Sections 4 and 5 of the manuscript report these elements across multiple real-world datasets, including baseline descriptions, ablation studies, and performance metrics with standard deviations. To make the abstract more self-contained while remaining within length limits, we have revised it to briefly reference the datasets used and the consistent improvements observed. revision: partial

  2. Referee: [Method] Meta-learning optimization framework (described in the method): the claim that the framework 'automatically learn[s] the optimal importance of each auxiliary task' and generalizes without instability is unsupported; no meta-objective formulation, stabilization techniques (e.g., gradient surgery), or ablation replacing the meta-learner with static/random weights is provided, leaving the weakest assumption untested.

    Authors: The method section presents the meta-learning framework for automatic task weighting. To directly address the concern, we will add an explicit mathematical formulation of the meta-objective in the revision. We will also include a new ablation that replaces the meta-learner with static and random weights to demonstrate its contribution. Gradient clipping is already applied during meta-optimization to mitigate instability; we will clarify this and add discussion of generalization behavior. These additions will strengthen the empirical support for the automatic weighting claim. revision: yes

Circularity Check

0 steps flagged

No circularity: meta-learning weighting described at high level without reduction to fitted inputs or self-citations

full rationale

The abstract and description present AusRec as integrating self-supervised tasks with meta-learning for adaptive weighting, but no equations, derivations, or self-citations are provided that would allow any claim to reduce by construction to its inputs. The central mechanism is presented as an independent optimization framework rather than a renaming or post-hoc fit. This is the common case of a self-contained proposal whose validity rests on external experiments rather than definitional equivalence.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The paper is an empirical machine learning proposal with no explicit mathematical axioms or derivations visible in the abstract; it relies on standard assumptions of meta-learning frameworks and self-supervised learning being beneficial for representation learning.

pith-pipeline@v0.9.0 · 5640 in / 1018 out tokens · 22663 ms · 2026-05-23T07:28:31.253394+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

63 extracted references · 63 canonical work pages · 3 internal anchors

  1. [1]

    Y. Liu, X. Zhou, H. Kou, Y. Zhao, X. Xu, X. Zhang, L. Qi, Privacy- preserving point-of-interest recommendation based on simplified graph convolutional network for geological traveling, ACM Transactions on Intelligent Systems and Technology 15 (4) (2024) 1–17

    work page 2024

  2. [2]

    W. Fan, Y. Ma, Q. Li, J. Wang, G. Cai, J. Tang, D. Yin, A graph neural network framework for social recommendations, IEEE Transactions on Knowledge and Data Engineering (2020)

    work page 2020

  3. [3]

    L. Meng, Z. Liu, D. Chu, Q. Z. Sheng, J. Yu, X. Song, Poi recommen- dation for occasional groups based on hybrid graph neural networks, Expert Systems with Applications 237 (2024) 121583

    work page 2024

  4. [4]

    McPherson, L

    M. McPherson, L. Smith-Lovin, J. M. Cook, Birds of a feather: Ho- mophily in social networks, Annual review of sociology (2001) 415–444

    work page 2001

  5. [5]

    Koren, R

    Y. Koren, R. Bell, C. Volinsky, Matrix factorization techniques for rec- ommender systems, Computer 42 (8) (2009) 30–37

    work page 2009

  6. [6]

    W. Fan, Y. Ma, D. Yin, J. Wang, J. Tang, Q. Li, Deep social col- laborative filtering, in: Proceedings of the 13th ACM Conference on Recommender Systems, 2019, pp. 305–313

    work page 2019

  7. [7]

    S. Ni, W. Zhou, J. Wen, L. Hu, S. Qiao, Enhancing sequential recom- mendation with contrastive generative adversarial network, Information Processing & Management 60 (3) (2023) 103331

    work page 2023

  8. [8]

    P. R. de Souza, F. A. Dur˜ ao, Exploiting social capital for improving personalized recommendations in online social networks, Expert Systems with Applications 246 (2024) 123098. 25

    work page 2024

  9. [9]

    W. Fan, Y. Ma, Q. Li, Y. He, E. Zhao, J. Tang, D. Yin, Graph neural networks for social recommendation, in: The world wide web conference, 2019, pp. 417–426

    work page 2019

  10. [10]

    L. Wu, P. Sun, Y. Fu, R. Hong, X. Wang, M. Wang, A neural influence diffusion model for social recommendation, in: Proceedings of the 42nd international ACM SIGIR conference on research and development in information retrieval, 2019

    work page 2019

  11. [11]

    J. Yu, H. Yin, J. Li, M. Gao, Z. Huang, L. Cui, Enhance social recom- mendation with adversarial graph convolutional networks, IEEE Trans- actions on Knowledge and Data Engineering (2020)

    work page 2020

  12. [12]

    Y. A. Kim, J. Srivastava, Impact of social influence in e-commerce de- cision making, in: Proceedings of the ninth international conference on Electronic commerce, 2007, pp. 293–302

    work page 2007

  13. [13]

    Jamali, M

    M. Jamali, M. Ester, A matrix factorization technique with trust prop- agation for recommendation in social networks, in: Proceedings of the fourth ACM conference on Recommender systems, 2010

    work page 2010

  14. [14]

    J. Tang, X. Hu, H. Liu, Social recommendation: a review, Social Net- work Analysis and Mining 3 (4) (2013) 1113–1133

    work page 2013

  15. [15]

    Z. Wu, X. Yang, X. Wei, P. Yuan, Y. Zhang, J. Bai, A self-supervised anomaly detection algorithm with interpretability, Expert Systems with Applications 237 (2024) 121539

    work page 2024

  16. [16]

    Hashmi, S

    E. Hashmi, S. Y. Yayilgan, M. M. Yamin, M. Abomhara, M. Ullah, Self- supervised hate speech detection in norwegian texts with lexical and semantic augmentations, Expert Systems with Applications 264 (2025) 125843

    work page 2025

  17. [17]

    J. Wu, X. Wang, F. Feng, X. He, L. Chen, J. Lian, X. Xie, Self- supervised graph learning for recommendation, in: Proceedings of the 44th International ACM SIGIR Conference on Research and Develop- ment in Information Retrieval, 2021

    work page 2021

  18. [18]

    X. Xia, H. Yin, J. Yu, Q. Wang, L. Cui, X. Zhang, Self-supervised hypergraph convolutional networks for session-based recommendation, 26 in: Proceedings of the AAAI conference on artificial intelligence, Vol. 35, 2021, pp. 4503–4511

    work page 2021

  19. [19]

    J. Yu, H. Yin, J. Li, Q. Wang, N. Q. V. Hung, X. Zhang, Self-supervised multi-channel hypergraph convolutional network for social recommenda- tion, in: Proceedings of the Web Conference 2021, 2021

    work page 2021

  20. [20]

    R. Tian, J. Cai, C. Li, J. Wang, Self-supervised pre-training model based on multi-view for mooc recommendation, Expert Systems with Appli- cations 252 (2024) 124143

    work page 2024

  21. [21]

    BPR: Bayesian Personalized Ranking from Implicit Feedback

    S. Rendle, C. Freudenthaler, Z. Gantner, L. Schmidt-Thieme, Bpr: Bayesian personalized ranking from implicit feedback, arXiv preprint arXiv:1205.2618 (2012)

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  22. [22]

    X. Juan, M. Peng, X. Wang, Dynamic self-training with less uncertainty for graph imbalance learning, Expert Systems with Applications (2025) 126643

    work page 2025

  23. [23]

    X. Juan, X. Liang, H. Xue, X. Wang, Multi-strategy adaptive data aug- mentation for graph neural networks, Expert Systems with Applications 258 (2024) 125076

    work page 2024

  24. [24]

    Y. Wang, H. Xue, K. Zhou, T. Chen, X. Wang, A two-stage co- adversarial perturbation to mitigate out-of-distribution generalization of large-scale graph, Expert Systems with Applications (2024) 124472

    work page 2024

  25. [25]

    W. Fan, X. Liu, W. Jin, X. Zhao, J. Tang, Q. Li, Graph trend networks for recommendations, arXiv preprint arXiv:2108.05552 (2021)

    work page arXiv 2021

  26. [26]

    X. He, K. Deng, X. Wang, Y. Li, Y. Zhang, M. Wang, Lightgcn: Sim- plifying and powering graph convolution network for recommendation, in: Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval, 2020

    work page 2020

  27. [27]

    W. Fan, Q. Li, M. Cheng, Deep modeling of social relations for rec- ommendation, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 32, 2018

    work page 2018

  28. [28]

    W. Zhou, A. U. Haq, L. Qiu, J. Akbar, Multi-view social recommenda- tion via matrix factorization with sub-linear convergence rate, Expert Systems with Applications 237 (2024) 121687. 27

    work page 2024

  29. [29]

    Z. Peng, W. Huang, M. Luo, Q. Zheng, Y. Rong, T. Xu, J. Huang, Graph representation learning via graphical mutual information maximization, in: Proceedings of The Web Conference 2020, 2020, pp. 259–270

    work page 2020

  30. [30]

    J. Qiu, Q. Chen, Y. Dong, J. Zhang, H. Yang, M. Ding, K. Wang, J. Tang, Gcc: Graph contrastive coding for graph neural network pre- training, in: Proceedings of the 26th ACM SIGKDD International Con- ference on Knowledge Discovery & Data Mining, 2020, pp. 1150–1160

    work page 2020

  31. [31]

    K. Sun, Z. Lin, Z. Zhu, Multi-stage self-supervised learning for graph convolutional networks on graphs with few labeled nodes, in: Proceed- ings of the AAAI Conference on Artificial Intelligence, Vol. 34, 2020, pp. 5892–5899

    work page 2020

  32. [32]

    Velickovic, W

    P. Velickovic, W. Fedus, W. L. Hamilton, P. Li` o, Y. Bengio, R. D. Hjelm, Deep graph infomax., ICLR (Poster) 2 (3) (2019) 4

    work page 2019

  33. [33]

    Sankar, J

    A. Sankar, J. Wang, A. Krishnan, H. Sundaram, Beyond localized graph neural networks: An attributed motif regularization framework, in: 2020 IEEE International Conference on Data Mining (ICDM), IEEE, 2020, pp. 472–481

    work page 2020

  34. [34]

    Hwang, J

    D. Hwang, J. Park, S. Kwon, K. Kim, J.-W. Ha, H. J. Kim, Self- supervised auxiliary learning with meta-paths for heterogeneous graphs, Advances in Neural Information Processing Systems 33 (2020) 10294– 10305

    work page 2020

  35. [35]

    L. Wu, H. Lin, C. Tan, Z. Gao, S. Z. Li, Self-supervised learning on graphs: Contrastive, generative, or predictive, IEEE Transactions on Knowledge and Data Engineering (2021)

    work page 2021

  36. [36]

    W. Jin, T. Derr, H. Liu, Y. Wang, S. Wang, Z. Liu, J. Tang, Self- supervised learning on graphs: Deep insights and new direction, arXiv preprint arXiv:2006.10141 (2020)

    work page arXiv 2006

  37. [37]

    J. Yu, H. Yin, X. Xia, T. Chen, J. Li, Z. Huang, Self-supervised learning for recommender systems: A survey, arXiv preprint arXiv:2203.15876 (2022). 28

    work page arXiv 2022

  38. [38]

    H. Liu, H. Lin, W. Fan, Y. Ren, B. Xu, X. Zhang, D. Wen, N. Zhao, Y. Lin, L. Yang, Self-supervised learning for fair recommender systems, Applied Soft Computing (2022) 109126

    work page 2022

  39. [39]

    Jaiswal, A

    A. Jaiswal, A. R. Babu, M. Z. Zadeh, D. Banerjee, F. Makedon, A survey on contrastive self-supervised learning, Technologies 9 (1) (2020) 2

    work page 2020

  40. [40]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    J. Devlin, M.-W. Chang, K. Lee, K. Toutanova, Bert: Pre-training of deep bidirectional transformers for language understanding, arXiv preprint arXiv:1810.04805 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  41. [41]

    Joshi, D

    M. Joshi, D. Chen, Y. Liu, D. S. Weld, L. Zettlemoyer, O. Levy, Spanbert: Improving pre-training by representing and predicting spans, Transactions of the Association for Computational Linguistics 8 (2020) 64–77

    work page 2020

  42. [42]

    Conneau, G

    A. Conneau, G. Lample, Cross-lingual language model pretraining, Ad- vances in neural information processing systems 32 (2019)

    work page 2019

  43. [43]

    K. Song, X. Tan, T. Qin, J. Lu, T.-Y. Liu, Mass: Masked se- quence to sequence pre-training for language generation, arXiv preprint arXiv:1905.02450 (2019)

    work page internal anchor Pith review Pith/arXiv arXiv 1905

  44. [44]

    J. Yu, H. Yin, M. Gao, X. Xia, X. Zhang, N. Q. Viet Hung, Socially- aware self-supervised tri-training for recommendation, in: Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2021, pp. 2084–2092

    work page 2021

  45. [45]

    M. Li, S. Hu, F. Zhu, Q. Zhu, Few-shot learning for cold-start rec- ommendation, in: Proceedings of the 2024 Joint International Confer- ence on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), 2024, pp. 7185–7195

    work page 2024

  46. [46]

    N. Li, J. Wang, X. Liu, L. Wang, L. Zhang, Contrastive learning-based meta-learning sequential recommendation, in: 2024 International Con- ference on Distributed Computing and Optimization Techniques (ICD- COT), IEEE, 2024, pp. 1–5

    work page 2024

  47. [47]

    X. Pan, Y. Chen, C. Tian, Z. Lin, J. Wang, H. Hu, W. X. Zhao, Multi- modal meta-learning for cold-start sequential recommendation, in: Pro- ceedings of the 31st ACM international conference on information & knowledge management, 2022, pp. 3421–3430. 29

    work page 2022

  48. [48]

    J. Chen, H. Dong, Y. Qiu, X. He, X. Xin, L. Chen, G. Lin, K. Yang, Autodebias: Learning to debias for recommendation, in: Proceedings of the 44th International ACM SIGIR Conference on Research and Devel- opment in Information Retrieval, 2021, pp. 21–30

    work page 2021

  49. [49]

    C. Finn, P. Abbeel, S. Levine, Model-agnostic meta-learning for fast adaptation of deep networks, in: International conference on machine learning, PMLR, 2017, pp. 1126–1135

    work page 2017

  50. [50]

    H. Lee, J. Im, S. Jang, H. Cho, S. Chung, Melu: Meta-learned user pref- erence estimator for cold-start recommendation, in: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019, pp. 1073–1082

    work page 2019

  51. [51]

    Bharadhwaj, Meta-learning for user cold-start recommendation, in: 2019 International Joint Conference on Neural Networks (IJCNN), IEEE, 2019, pp

    H. Bharadhwaj, Meta-learning for user cold-start recommendation, in: 2019 International Joint Conference on Neural Networks (IJCNN), IEEE, 2019, pp. 1–8

    work page 2019

  52. [52]

    Y. Chen, B. Chen, X. He, C. Gao, Y. Li, J.-G. Lou, Y. Wang, λopt: Learn to regularize recommender models in finer levels, in: Proceed- ings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019, pp. 978–986

    work page 2019

  53. [53]

    Zhang, F

    Y. Zhang, F. Feng, C. Wang, X. He, M. Wang, Y. Li, Y. Zhang, How to retrain recommender system? a sequential meta-learning method, in: Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, 2020, pp. 1479– 1488

    work page 2020

  54. [54]

    R. Miao, K. Zhou, Y. Wang, N. Liu, Y. Wang, X. Wang, Rethinking independent cross-entropy loss for graph-structured data, arXiv preprint arXiv:2405.15564 (2024)

    work page arXiv 2024

  55. [55]

    R. Milo, S. Shen-Orr, S. Itzkovitz, N. Kashtan, D. Chklovskii, U. Alon, Network motifs: simple building blocks of complex networks, Science (2002)

    work page 2002

  56. [56]

    Newman, Networks, Oxford university press, 2018

    M. Newman, Networks, Oxford university press, 2018. 30

    work page 2018

  57. [57]

    C. Shi, Y. Li, J. Zhang, Y. Sun, S. Y. Philip, A survey of heterogeneous information network analysis, IEEE Transactions on Knowledge and Data Engineering (2016)

    work page 2016

  58. [58]

    A. R. Benson, D. F. Gleich, J. Leskovec, Higher-order organization of complex networks, Science 353 (6295) (2016) 163–166

    work page 2016

  59. [59]

    H. Wang, H. Zhao, B. Li, Bridging multi-task learning and meta- learning: Towards efficient training and effective adaptation, in: In- ternational Conference on Machine Learning, PMLR, 2021, pp. 10991– 11002

    work page 2021

  60. [60]

    J. Shu, Q. Xie, L. Yi, Q. Zhao, S. Zhou, Z. Xu, D. Meng, Meta-weight- net: Learning an explicit mapping for sample weighting, Advances in neural information processing systems 32 (2019)

    work page 2019

  61. [61]

    X. He, L. Liao, H. Zhang, L. Nie, X. Hu, T.-S. Chua, Neural collabo- rative filtering, in: Proceedings of the 26th international conference on world wide web, 2017

    work page 2017

  62. [62]

    X. Wang, X. He, M. Wang, F. Feng, T.-S. Chua, Neural graph collab- orative filtering, in: Proceedings of the 42nd international ACM SIGIR conference on Research and development in Information Retrieval, 2019

    work page 2019

  63. [63]

    M. Chen, C. Huang, L. Xia, W. Wei, Y. Xu, R. Luo, Heterogeneous graph contrastive learning for recommendation, in: Proceedings of the sixteenth ACM international conference on web search and data mining, 2023, pp. 544–552. 31

    work page 2023