Pith. sign in

REVIEW 5 major objections 5 minor 67 references

AdaF^2M^2: Comprehensive Learning and Responsive Leveraging Features in Recommendation System

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read AdaF2M2 claims that randomly masking feature embeddings during training and reweighting features with a state-aware adapter lets recommendation models learn non-ID features and improve online engagement metrics such as active days and app…

desk verdict Plausible industrial feature-mask/adapter framework with consistent small offline gains, but the online evidence as reported does not support the headline cumulative improvements. read the letter →

arxiv 2501.15816 v1 pith:C7WRHIE6 submitted 2025-01-27 cs.IR cs.AI

Yongchun Zhu , Jingwu Chen , Ling Chen , Yitan Li , Feng Zhang , Xiao Yang , Zuotao Liu This is my paper · ORCID
classification cs.IRcs.AI
keywords recommendationsystemsfeaturerepresentationlearningmaskingstate-awareadapterlong-taildistributioncold-startCTRpredictionmodel-agnosticframework
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

Recommendation models trained on skewed real-world data tend to lean on user and item IDs, leaving age, gender, genre, and other non-ID features under-trained. AdaF2M2 is a model-agnostic wrapper that attacks this with two coordinated pieces: random feature masking during training, which forces the model to predict from non-ID features, and a state-aware adapter that re-weights features for new versus old users and hot versus cold items. The paper reports that adding the wrapper to five base models improves offline AUC and UAUC, and that online A/B tests in a music recommendation service show cumulative gains of +1.37% in user active days and +1.89% in app duration. If these results hold, the framework offers a deployment path for better cold-start and long-tail performance without changing the serving-time architecture.

What carries the argument

The framework has two coordinated mechanisms. The feature-mask mechanism generates k augmented copies of each training sample by randomly replacing feature embeddings with distinct learned [MASK] embeddings, then runs each copy through the base model and supervises it with the original label, so the model learns to make predictions from non-ID features when IDs are masked. The state-aware adapter takes four empirical state signals (active days, concatenated ID embeddings, norms of ID embeddings, and interaction counts) and produces Sigmoid weights that rescale feature embeddings before prediction, allowing different features to matter for new versus old users and cold versus hot items. Keeping the two paths separate prevents the adapter's low weights from shrinking gradients for non-ID features.

What would settle it

Re-run the online A/B test on the same DCN-V2 ranking task with a two-week horizon; if Active Days and Duration show no statistically significant gain, the central claim fails. Alternatively, construct a toy dataset where IDs perfectly determine the label and non-ID features carry no signal, then check whether the feature-mask auxiliary loss still improves test AUC; if it does, the claimed mechanism of learning non-ID features is not what drives the gain.

Watch

Extended reading notes

Core claim

The paper's central claim is that the harmful effects of popularity bias on feature learning can be countered by training a base recommendation model on masked-feature augmented samples while separately learning per-feature weights from empirical user and item state signals. During training, each sample is copied k times, and in each copy every feature embedding is replaced by its own learned [MASK] embedding with probability drawn from the range [0.1, 0.5]; all copies are scored by the same base model with the same click or finish label, so the model must learn to use whatever non-masked features remain. The state-aware adapter takes active days, concatenated ID embeddings, ID-embedding norms, and interaction counts as input and produces Sigmoid weights applied to all feature embeddings before the final prediction. The mask path and the adapter path are trained with separate losses, so low adapter weights cannot starve non-ID embeddings of gradient. The paper claims that this combination, applied on top of FM, YouTube DNN, AFN, DCN-V2, and EulerNet, improves AUC and UAUC on public and on 4-billion- and 20-billion-sample industrial datasets and produces statistically significant online gains.

Load-bearing premise

The method assumes that a prediction made from masked, non-ID features for a high-active user should match that same user's behavior in an earlier, less-active stage; if this counterfactual is wrong, the extra training signal teaches a distribution that does not match how the model is actually served.

Editorial extensions

If this is right

  • Feature-mask training should reduce reliance on user and item IDs, so cold-start users and items should get better predictions from age, gender, genre, and similar meta features.
  • Because adapter weights come from state signals rather than from the features themselves, the weight distribution should shift appropriately as a user moves from new to active or an item from cold to hot.
  • Serving latency is unchanged by the mask mechanism, since masked forwards run only during training; the deployed model is still a single forward pass with adapter-weighted embeddings.
  • The framework is model-agnostic, so any ranking or two-tower retrieval model that takes embeddings as input can be wrapped without redesigning the feature interaction layer.
  • If the reported online numbers generalize, a production recommender can expect engagement gains of a few tenths of a percent per model upgrade, which is substantial relative to typical production improvements.

Reading between the lines

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

  • A plausible reading not tested in the paper is that the feature-mask auxiliary loss acts partly as a regularizer or noise-injection mechanism; comparing it with standard embedding dropout at matched compute would show how much of the gain is specific to masking IDs versus general noise robustness.
  • The four state signals are treated as sufficient, but the paper gives no evidence that they exhaust relevant user and item state; adding session context, content embeddings, or recency signals to the adapter could produce even sharper weights.
  • The counterfactual premise could be tested directly with logged data: for users with complete histories, compare a masked-feature prediction to the same user's actual early-stage logged behavior; divergence would suggest tuning the mask probability range or using state-conditioned masks.
  • The framework's success on engagement metrics suggests it could complement popularity-bias correction and debiasing methods, since it addresses the representation side of long-tail data rather than only re-ranking or sampling.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper introduces AdaF2M2, a model-agnostic framework for recommender-system feature modeling. It combines a feature-mask mechanism, which generates k randomly masked augmented copies of each sample and trains them with a task-oriented cross-entropy auxiliary loss, with a state-aware adapter that multiplies each feature embedding by a sigmoid-gated weight computed from active-day statistics, ID embeddings, embedding norms, and interaction counts. The framework is applied to five base models (FM, YouTube DNN, AFN, DCN-V2, EulerNet) and evaluated offline on MovieLens-1M and two industrial datasets (DouyinMusic-4B and -20B), as well as online through A/B tests in ranking, retrieval, and item cold-start tasks. The paper reports cumulative online improvements of +1.37% on user active days and +1.89% on app duration, and claims deployment in multiple applications of Douyin Group.

Significance. The contribution is empirical and engineering-oriented. If the reported gains hold up, AdaF2M2 is practically valuable because it is model-agnostic, imposes no additional serving latency (the mask forward passes are training-only), and shows consistent offline gains across five base models and two data scales. The inclusion of a public dataset alongside billion-scale industrial logs and online A/B tests gives the evaluation unusual breadth for a conference paper. However, the paper's central quantitative claims currently rest on incomplete statistical reporting: the confidence intervals for the online lifts are not interpretable as presented, and the cumulative headline numbers are not derivable from the tables. These issues are fixable in revision, but they are not merely cosmetic.

major comments (5)
  1. [Table 1 (and Tables 2, 3, 7)] Table 1's caption says the square brackets are 95% confidence intervals for online metrics, but the Overall row reports Active Days +0.212% with interval [-0.063%, +0.063%]. If these brackets are CIs for the relative lift, the interval includes 0, which contradicts the claim in Section 4.2 that the improvement is statistically significant; if they are CIs for the underlying absolute metric, then the CI for the relative lift is not reported. The same pattern appears in Tables 2, 3, and 7, and Table 4 gives no intervals at all. Please state precisely what the intervals denote and report the CI (or a test) for the relative improvement itself, along with sample sizes and experiment duration.
  2. [Abstract / Section 5] The headline numbers +1.37% and +1.89% cumulative improvements on Active Days and Duration are repeated in the abstract, contributions, and conclusion, but no table or equation explains how they are aggregated from the per-experiment lifts in Tables 1-4. For example, Table 1 reports +0.212%/+0.442% for ranking, Table 3 reports +0.066%/+0.195% and +0.073%/+0.184% for two retrieval models, and Table 4 reports additional cold-start lifts. Without the aggregation formula, weighting, or experimental periods, the central quantitative claim cannot be verified or reproduced from the paper. Please document the cumulative calculation and report the component lifts with their uncertainties.
  3. [Section 4.3 / Tables 5-6] The text states 'We report the offline results via three random runs' and 'Statistically significant improvement is marked with bold font in the tables,' yet Tables 5 and 6 show only point estimates (RelaImpr values) with no standard deviations, confidence intervals, or p-values, and no bold markers are visible in the plain-text rendering. Relative AUC improvements of 0.13%-0.71% may be smaller than run-to-run variance. Please add dispersion measures or significance tests, and ensure the bold marking is present and defined.
  4. [Section 3.2] The feature-mask mechanism explicitly relies on the counterfactual assumption that a high-active user with only the basic features available at the new-user stage would receive a prediction consistent with the same user's later-stage behavior. This assumption is named but not tested. A concrete way to test it is to compare masked-sample predictions against actual new-user labels for users who later become active, or to evaluate a variant trained only on unmasked samples on a new-user holdout. Without such evidence, the auxiliary masked loss remains a plausible but unvalidated mechanism for improving cold-start generalization.
  5. [Section 3.3] The state-aware adapter is built on four empirical signals (active days, ID embeddings, embedding norms, interaction counts), but the paper gives no evidence that these signals capture the state information that matters for feature weighting. The ablation in Tables 7-8 shows the adapter helps offline but not online in isolation, so the mechanism's contribution is not fully characterized. Please add ablations that remove or replace each state signal, or otherwise demonstrate that the chosen signals are sufficient for the claimed responsiveness.
minor comments (5)
  1. [Section 1] The phrase 'The main contributions of our work are summarized into four folds' should be 'four aspects' or 'fourfold'; the current wording is nonstandard.
  2. [Table 1] The negative value for Dislike (-1.594% overall) is not discussed; please clarify the direction of improvement for constraint metrics and whether a decrease in Dislike is the desired outcome.
  3. [Section 4.1] The industrial dataset paragraphs report sampling time spans but not how the two DouyinMusic datasets differ in feature sets; a sentence on feature compatibility would help readers assess the effect of dataset size.
  4. [Figure 2] The heatmaps have no color scale or numeric range for the reported average feature weights, which makes the comparison across features hard to interpret; the phrase 'the limitation of pages' also appears to be a leftover editing note.
  5. [Algorithm 1] The serving stage returns only the adapter prediction; it would be helpful to state explicitly that the [MASK] embeddings are not used at serving time, though this is implied in Section 3.4.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the framework is an empirical architecture with external benchmarks, and all self-citations are peripheral related-work references.

full rationale

The paper does not contain a derivation in which an output quantity is defined in terms of its own fitted parameters. The feature-mask mechanism (Section 3.2) creates augmented samples by random masking and optimizes an auxiliary cross-entropy loss (Eq. 5); the state-aware adapter (Section 3.3) computes per-feature weights from four explicit state signals via Eq. 8, and the final prediction is Eq. 11 trained with the main loss Eq. 12. None of the reported metrics is defined as the output of these equations themselves; the evidence is external: MovieLens-1M (Table 6), large industrial datasets (Table 5), and online A/B tests (Tables 1-4, 7). The only self-citations, e.g., reference [58] in related-work lists of DNN and sequential-feature methods, are illustrative and not load-bearing; no uniqueness theorem, ansatz, or prior result is imported from the authors' own work to force the proposed design. The counterfactual assumption stated in Section 3.2 is an explicitly declared modeling motivation for the feature-mask training objective, not a conclusion derived from the objective, so it is not circular. Two non-circularity concerns should be flagged as reporting/correctness issues: the abstract/conclusion cumulative gains of +1.37% and +1.89% are not explicitly aggregated from Tables 1-4, and the 95% confidence intervals in those tables, such as +0.212% with [-0.063%, +0.063%], are internally inconsistent as presented. These concern verifiability, not circular reasoning.

Assumptions & free parameters 3 free parameters · 7 assumptions · 1 invented entities

The framework depends on several hand-set hyperparameters and domain assumptions about long-tail behavior, masking as noise simulation, and state-signal sufficiency. The mask embedding is a new learned component. No external physical entity is introduced.

free parameters (3)
  • alpha (auxiliary loss weight) = 0.2
    Weight on the masked-sample auxiliary loss in Equation (13); chosen by hand with no reported sensitivity analysis.
  • k (number of augmented samples) = 1
    Number of random masks per training sample in Section 3.2; set to 1 without ablation.
  • mask probability range [beta, gamma] = [0.1, 0.5]
    Range for sampling the per-feature masking probability in Section 3.2; chosen by hand, no sensitivity tests.
assumptions (7)
  • domain assumption Industrial recommendation data follow a highly skewed long-tail distribution, causing models to over-rely on ID-based features and under-learn non-ID meta features.
    Stated in the Introduction and used to motivate the whole framework; no quantitative evidence is provided in this paper.
  • domain assumption Randomly masking features during training simulates real-world missing or noisy features and improves robustness and generalization.
    Section 3.2, second advantage of feature mask; this is a modeling assumption not proven.
  • domain assumption Counterfactual consistency: if a high-active user only had the basic features available at the new-user stage, the model would produce a consistent prediction across stages.
    Explicitly named in Section 3.2 as the counterfactual assumption underlying the feature-mask mechanism.
  • domain assumption The empirical state signals (active days, ID embeddings, embedding norms, interaction counts) are sufficient for the adapter to assign appropriate feature weights.
    Section 3.3 lists four signals without evidence that they exhaust relevant state information.
  • domain assumption Task-oriented supervised loss on masked samples is preferable to self-supervised contrastive loss for recommendation data.
    Section 3.2 argues labeled samples are abundant in recommendations; supported by reasoning but not by controlled comparison in this paper.
  • ad hoc to paper Sigmoid output for feature weights is better than Softmax for the adapter.
    Selected based on online experiments in Section 3.3; no theoretical justification.
  • standard math Standard backpropagation with cross-entropy loss and Adam optimizer is a valid training procedure.
    Used throughout Section 3 and experiments; standard background.
invented entities (1)
  • Per-feature mask embedding [MASK]_i
    purpose: A learned default embedding that replaces a feature embedding when the feature is randomly masked during training.
    Introduced in Section 3.2 as a trainable component; it has no falsifiable handle outside the paper and its value is learned from the same data as the model.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AdaF^2M^2: Comprehensive Learning and Responsive Leveraging Features in Recommendation System." pith.science (2026). https://pith.science/paper/C7WRHIE6

@misc{pith2026250115816,
  author       = {Pith},
  title        = {Pith review of: AdaF^2M^2: Comprehensive Learning and Responsive Leveraging Features in Recommendation System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C7WRHIE6}},
  note         = {Machine review of arXiv:2501.15816}
}
abstract

Feature modeling, which involves feature representation learning and leveraging, plays an essential role in industrial recommendation systems. However, the data distribution in real-world applications usually follows a highly skewed long-tail pattern due to the popularity bias, which easily leads to over-reliance on ID-based features, such as user/item IDs and ID sequences of interactions. Such over-reliance makes it hard for models to learn features comprehensively, especially for those non-ID meta features, e.g., user/item characteristics. Further, it limits the feature leveraging ability in models, getting less generalized and more susceptible to data noise. Previous studies on feature modeling focus on feature extraction and interaction, hardly noticing the problems brought about by the long-tail data distribution. To achieve better feature representation learning and leveraging on real-world data, we propose a model-agnostic framework AdaF^2M^2, short for Adaptive Feature Modeling with Feature Mask. The feature-mask mechanism helps comprehensive feature learning via multi-forward training with augmented samples, while the adapter applies adaptive weights on features responsive to different user/item states. By arming base models with AdaF^2M^2, we conduct online A/B tests on multiple recommendation scenarios, obtaining +1.37% and +1.89% cumulative improvements on user active days and app duration respectively. Besides, the extended offline experiments on different models show improvements as well. AdaF$^2$M$^2$ has been widely deployed on both retrieval and ranking tasks in multiple applications of Douyin Group, indicating its superior effectiveness and universality.

Figures

Figures reproduced from arXiv: 2501.15816 by the authors.

Figure 1
Figure 1. Adaptive Feature Modeling with Feature Mask [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Heatmaps with the average feature weights of [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

67 extracted references · 45 canonical work pages

  1. [1]

    Rahul Bhagat, Srevatsan Muralidharan, Alex Lobzhanidze, and Shankar Vish- wanath. 2018. Buy it again: Modeling repeat purchase recommendations. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining . 62–70

    work page 2018

  2. [2]

    Jiangxia Cao, Xixun Lin, Xin Cong, Jing Ya, Tingwen Liu, and Bin Wang. 2022. Disencdr: Learning disentangled representations for cross-domain recommenda- tion. In Proceedings of the 45th International ACM SIGIR conference on research and development in information retrieval . 267–277

    2022

  3. [3]

    Jiangxia Cao, Jiawei Sheng, Xin Cong, Tingwen Liu, and Bin Wang. 2022. Cross- domain recommendation to cold-start users via variational information bottle- neck. In 2022 IEEE 38th International Conference on Data Engineering (ICDE). IEEE, 2209–2223

    work page 2022

  4. [4]

    Allison JB Chaney, Brandon M Stewart, and Barbara E Engelhardt. 2018. How algorithmic confounding in recommendation systems increases homogeneity and decreases utility. In Proceedings of the 12th ACM Conference on Recommender Systems. 224–232

    work page 2018

  5. [5]

    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 International Conference on Knowledge Discovery & Data Mining

    work page 2023

  6. [6]

    Jingwu Chen, Fuzhen Zhuang, Tianxin Wang, Leyu Lin, Feng Xia, Lihuan Du, and Qing He. 2019. Follow the Title Then Read the Article: Click-Guide Network for Dwell Time Prediction. IEEE Transactions on Knowledge and Data Engineering 33, 7 (2019), 2903–2913

    work page 2019

  7. [7]

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A simple framework for contrastive learning of visual representations. In Proceed- ings of the 37th International Conference on Machine Learning . 1597–1607

    work page 2020

  8. [8]

    Weiyu Cheng, Yanyan Shen, and Linpeng Huang. 2020. Adaptive factorization network: Learning adaptive-order feature interactions. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 34. 3609–3616

    2020

Show all 67 references
  1. [9]

    Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep neural networks for youtube recommendations. In Proceedings of the 10th ACM Conference on Recommender Systems. 191–198

    2016

  2. [10]

    Wei Deng, Junwei Pan, Tian Zhou, Deguang Kong, Aaron Flores, and Guang Lin. 2021. Deeplight: Deep lightweight feature interactions for accelerating ctr predictions in ad serving. In Proceedings of the 14th ACM International Conference on Web Search and Data Mining . 922–930

    2021

  3. [11]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human...

    2019

  4. [12]

    Tom Fawcett. 2006. An introduction to ROC analysis. Pattern recognition letters 27, 8 (2006), 861–874

    2006

  5. [13]

    Huifeng Guo, TANG Ruiming, Yunming Ye, Zhenguo Li, and Xiuqiang He. 2017. DeepFM: A Factorization-Machine based Neural Network for CTR Prediction. In Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelli- gence. International Joint Conferences on...

    2017

  6. [14]

    Xiaobo Hao, Yudan Liu, Ruobing Xie, Kaikai Ge, Linyao Tang, Xu Zhang, and Leyu Lin. 2021. Adversarial feature translation for multi-domain recommendation. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. 2964–2973

    2021

  7. [15]

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. 2020. Momen- tum contrast for unsupervised visual representation learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 9729–9738

    2020

  8. [16]

    Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, and Tat-Seng Chua. 2017. Neural collaborative filtering. In Proceedings of the 26th International Conference on World Wide Web. 173–182

    2017

  9. [17]

    Tongwen Huang, Zhiqi Zhang, and Junlin Zhang. 2019. FiBiNET: combining fea- ture importance and bilinear feature interaction for click-through rate prediction. In Proceedings of the 13th ACM Conference on Recommender Systems . 169–177

    2019

  10. [18]

    Yuchin Juan, Yong Zhuang, Wei-Sheng Chin, and Chih-Jen Lin. 2016. Field- aware factorization machines for CTR prediction. In Proceedings of the 10th ACM conference on recommender systems . 43–50

    2016

  11. [19]

    Diederik P Kingma and Jimmy Ba. 2015. Adam: A method for stochastic opti- mization. In International Conference on Learning Representations , Vol. 5

    2015

  12. [20]

    Yehuda Koren, Robert Bell, and Chris Volinsky. 2009. Matrix factorization tech- niques for recommender systems. Computer 42, 8 (2009), 30–37

    2009

  13. [21]

    Jianxun Lian, Xiaohuan Zhou, Fuzheng Zhang, Zhongxia Chen, Xing Xie, and Guangzhong Sun. 2018. xdeepfm: Combining explicit and implicit feature in- teractions for recommender systems. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data ...

    2018

  14. [22]

    Bin Liu, Niannan Xue, Huifeng Guo, Ruiming Tang, Stefanos Zafeiriou, Xiuqiang He, and Zhenguo Li. 2020. AutoGroup: Automatic feature grouping for modelling explicit high-order feature interactions in CTR prediction. In Proceedings of the 43rd International ACM SIGIR Conference...

    2020

  15. [23]

    Bin Liu, Chenxu Zhu, Guilin Li, Weinan Zhang, Jincai Lai, Ruiming Tang, Xi- uqiang He, Zhenguo Li, and Yong Yu. 2020. Autofis: Automatic feature interaction selection in factorization models for click-through rate prediction. In proceedings of the 26th ACM SIGKDD International...

    2020

  16. [24]

    Yudan Liu, Kaikai Ge, Xu Zhang, and Leyu Lin. 2019. Real-time attention based look-alike model for recommender system. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining . 2765–2773

    2019

  17. [25]

    Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692 (2019)

    2019 arXiv

  18. [26]

    Tong Man, Huawei Shen, Xiaolong Jin, and Xueqi Cheng. 2017. Cross-domain recommendation: an embedding and mapping approach. In Proceedings of the 26th International Joint Conference on Artificial Intelligence . 2464–2470

    2017

  19. [27]

    Wentao Ouyang, Xiuwu Zhang, Shukui Ren, Li Li, Kun Zhang, Jinmei Luo, Zhaojie Liu, and Yanlong Du. 2021. Learning graph meta embeddings for cold-start ads in click-through rate prediction. In Proceedings of the 44th International ACM SIGIR Conference on Research and Developmen...

    2021

  20. [28]

    Feiyang Pan, Shuokai Li, Xiang Ao, Pingzhong Tang, and Qing He. 2019. Warm up cold-start advertisements: Improving ctr predictions via learning to learn id embeddings. In Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retr...

    2019

  21. [29]

    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. In Proceedings of the 29th ACM International Conference on Informat...

    2020

  22. [30]

    Steffen Rendle. 2010. Factorization Machines. In Proceedings of the 2010 IEEE International Conference on Data Mining . 995–1000

    2010

  23. [31]

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

  24. [32]

    Ajit P Singh and Geoffrey J Gordon. 2008. Relational learning via collective matrix factorization. In Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . 650–658

    2008

  25. [33]

    Weiping Song, Chence Shi, Zhiping Xiao, Zhijian Duan, Yewen Xu, Ming Zhang, and Jian Tang. 2019. Autoint: Automatic feature interaction learning via self- attentive neural networks. InProceedings of the 28th ACM International Conference on Information and Knowledge Management ...

    2019

  26. [34]

    Rui Sun, Xuezhi Cao, Yan Zhao, Junchen Wan, Kun Zhou, Fuzheng Zhang, Zhongyuan Wang, and Kai Zheng. 2020. Multi-modal knowledge graphs for recommender systems. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management . 1405–1414

    2020

  27. [35]

    Yang Sun, Junwei Pan, Alex Zhang, and Aaron Flores. 2021. FM2: Field-matrixed factorization machines for recommender systems. In Proceedings of the Web Conference 2021. 2828–2837

    2021

  28. [36]

    Ying Sun, Hengshu Zhu, Chuan Qin, Fuzhen Zhuang, Qing He, and Hui Xiong

  29. [37]

    Ying Sun, Hengshu Zhu, Lu Wang, Le Zhang, and Hui Xiong. 2024. Large-scale online job search behaviors reveal labor market shifts amid COVID-19. Nature Cities 1, 2 (2024), 150–163. DASFAA ’25, May 26-29, 2025, Singapore Zhu, et al

    2024

  30. [38]

    Ying Sun, Fuzhen Zhuang, Hengshu Zhu, Qi Zhang, Qing He, and Hui Xiong

  31. [39]

    Jiaxi Tang and Ke Wang. 2018. Personalized top-n sequential recommenda- tion via convolutional sequence embedding. In Proceedings of the eleventh ACM International Conference on Web Search and Data Mining . 565–573

    2018

  32. [40]

    Zhen Tian, Ting Bai, Wayne Xin Zhao, Ji-Rong Wen, and Zhao Cao. 2023. Euler- Net: Adaptive Feature Interaction Learning via Euler’s Formula for CTR Predic- tion. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval

    2023

  33. [41]

    Nature communications 12, 1 (2021), 1992

    Market-oriented job skill valuation with cooperative composition neural network. Nature communications 12, 1 (2021), 1992

    2021

  34. [42]

    Jianling Wang, Haokai Lu, Sai Zhang, Bart Locanthi, Haoting Wang, Dylan Greaves, Benjamin Lipshitz, Sriraj Badam, Ed H Chi, Cristos J Goodrow, et al

  35. [43]

    Ruoxi Wang, Bin Fu, Gang Fu, and Mingliang Wang. 2017. Deep & cross network for ad click predictions. In Proceedings of the ADKDD’17 . 1–7

    2017

  36. [44]

    Maksims Volkovs, Guangwei Yu, and Tomi Poutanen. 2017. DropoutNet: address- ing cold start in recommender systems. In Proceedings of the 31st International Conference on Neural Information Processing Systems . 4964–4973

    2017

  37. [45]

    Yinwei Wei, Xiang Wang, Liqiang Nie, Xiangnan He, Richang Hong, and Tat-Seng Chua. 2019. MMGCN: Multi-modal graph convolution network for personalized recommendation of micro-video. In Proceedings of the 27th ACM International Conference on Multimedia. 1437–1445

    2019

  38. [46]

    Yiqing Wu, Ruobing Xie, Yongchun Zhu, Xiang Ao, Xin Chen, Xu Zhang, Fuzhen Zhuang, Leyu Lin, and Qing He. 2022. Multi-view multi-behavior contrastive learning in recommendation. In International conference on database systems for advanced applications. Springer, 166–182

    2022

  39. [47]

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

  40. [48]

    Ruoxi Wang, Rakesh Shivanna, Derek Cheng, Sagar Jain, Dong Lin, Lichan Hong, and Ed Chi. 2021. Dcn v2: Improved deep & cross network and practical lessons for web-scale learning to rank systems. In Proceedings of the Web Conference 2021. 1785–1797

    2021

  41. [49]

    Tiansheng Yao, Xinyang Yi, Derek Zhiyuan Cheng, Felix Yu, Ting Chen, Aditya Menon, Lichan Hong, Ed H Chi, Steve Tjoa, Jieqi Kang, et al. 2021. Self-supervised learning for large-scale item recommendations. In Proceedings of the 30th ACM International Conference on Information ...

    2021

  42. [50]

    Fuzheng Zhang, Nicholas Jing Yuan, Defu Lian, Xing Xie, and Wei-Ying Ma

  43. [51]

    Honglei Zhang, Fangyuan Luo, Jun Wu, Xiangnan He, and Yidong Li. 2023. LightFR: Lightweight federated recommendation with privacy-preserving matrix factorization. ACM Transactions on Information Systems 41, 4 (2023), 1–28

    2023

  44. [52]

    He Zhang, Ying Sun, Weiyu Guo, Yafei Liu, Haonan Lu, Xiaodong Lin, and Hui Xiong. 2023. Interactive interior design recommendation via coarse-to-fine multimodal reinforcement learning. In Proceedings of the 31st ACM International Conference on Multimedia. 6472–6480

    2023

  45. [53]

    Haoran Xin, Ying Sun, Chao Wang, and Hui Xiong. 2025. LLMCDSR: Enhancing Cross-Domain Sequential Recommendation with Large Language Models. ACM Transactions on Information Systems (2025)

    2025

  46. [54]

    Kai Zhang, Hao Qian, Qing Cui, Qi Liu, Longfei Li, Jun Zhou, Jianhui Ma, and Enhong Chen. 2021. Multi-interactive attention network for fine-grained feature learning in ctr prediction. In Proceedings of the 14th ACM International Conference on Web Search and Data Mining . 984–992

    2021

  47. [55]

    Qi Zhang, Hengshu Zhu, Ying Sun, Hao Liu, Fuzhen Zhuang, and Hui Xiong. 2021. Talent demand forecasting with attentive neural sequential model. InProceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining . 3906–3916

    2021

  48. [56]

    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. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining ...

    2018

  49. [57]

    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 sequential recommendation with mutual information maximization. In Pro- ceedings of the 29th ACM International Conference on In...

    2020

  50. [58]

    Yongchun Zhu, Jingwu Chen, Ling Chen, Yitan Li, Feng Zhang, and Zuotao Liu

  51. [59]

    Honglei Zhang, Shuyi Wang, Haoxuan Li, Chunyuan Zheng, Xu Chen, Li Liu, Shanshan Luo, and Peng Wu. 2024. Uncovering the propensity identification problem in debiased recommendations. In2024 IEEE 40th International Conference on Data Engineering (ICDE) . IEEE, 653–666

    2024

  52. [60]

    Yongchun Zhu, Ruobing Xie, Fuzhen Zhuang, Kaikai Ge, Ying Sun, Xu Zhang, Leyu Lin, and Juan Cao. 2021. Learning to warm up cold item embeddings for cold- start recommendation with meta scaling and shifting networks. In Proceedings of the 44th International ACM SIGIR Conference...

    2021

  53. [66]

    Yongchun Zhu, Zhenwei Tang, Yudan Liu, Fuzhen Zhuang, Ruobing Xie, Xu Zhang, Leyu Lin, and Qing He. 2022. Personalized transfer of user preferences for cross-domain recommendation. In Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining . 1507–1515

    2022

  54. [2009]

    In Proceedings of the Twenty-Fifth Conference on Uncertainty in Artificial Intelligence

    BPR: Bayesian personalized ranking from implicit feedback. In Proceedings of the Twenty-Fifth Conference on Uncertainty in Artificial Intelligence . 452–461

  55. [2016]

    In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

    Collaborative knowledge base embedding for recommender systems. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . 353–362

  56. [2017]

    In Proceedings of the 26th International Joint Conference on Artificial Intelligence

    Attentional factorization machines: learning the weight of feature in- teractions via attention networks. In Proceedings of the 26th International Joint Conference on Artificial Intelligence. 3119–3125

  57. [2021]

    Advances in Neural Information Processing Systems 34 (2021), 17221–17234

    Discerning decision-making process of deep neural networks with hierar- chical voting transformation. Advances in Neural Information Processing Systems 34 (2021), 17221–17234

    2021

  58. [2023]

    In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining

    Fresh Content Needs More Attention: Multi-funnel Fresh Content Recom- mendation. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 5082–5091

  59. [2024]

    In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval

    Interest clock: Time perception in real-time streaming recommendation system. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval . 2915–2919

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.