Pith. sign in

REVIEW 5 major objections 5 minor 69 references

When personalized features are stripped away, RAMP keeps ad CTR/CVR accuracy near its full-feature level, outperforming best baselines by 0.10–0.87% AUC.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 17:51 UTC pith:ZFPUWVBU

load-bearing objection Useful architecture with honest caveats, but the headline robustness claim leans on a masking protocol that overstates real-world transfer; the natural-traffic evidence is much weaker. the 5 major comments →

arxiv 2607.17473 v1 pith:ZFPUWVBU submitted 2026-07-20 cs.IR

RAMP: Robust Ad Recommendation Under Limited Personalized-Feature Availability via Masking and Alignment Pathways

Dairui Liu , Zhongyi Lu , Roger Zhe Li , Changhong Jin , Jitao Lu , Xinyang Shao , Bichen Shi , Mete Sertkan
show 6 more authors
Aghiles Salah Aonghus Lawlor Barry Smyth Tri Kurniawan Wijaya Ruihai Dong Xingsheng Guo
This is my paper
classification cs.IR
keywords click-through rate predictionconversion rate predictionnon-personalized recommendationprivacy constraintsfeature maskingknowledge distillationdual-tower architectureonline advertising
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Online ad systems predict clicks and conversions using user-linked signals such as IDs, device attributes, and behavior history, but privacy regulations increasingly make those signals unavailable for some or all traffic. The paper argues that accuracy under this constraint does not have to collapse: it proposes RAMP, a training scheme that separates the learning signal for personalized and non-personalized traffic and then aligns them. On four datasets spanning CTR and CVR prediction, RAMP improves non-personalized AUC by 0.10% to 0.87% over the best baseline, with all gains statistically significant, and it stays competitive when full features are available. The practical point is that a serving system can keep using a single low-latency model while a training-time auxiliary pathway transfers knowledge from rich, consented data to the restricted setting. The catch is that the public-benchmark experiments simulate missing features by masking, and the paper openly notes this may not capture all real differences of non-consenting traffic.

Core claim

RAMP's central claim is that a model can be trained to serve accurately when personalized features are absent by using a dual-tower personalized pathway with output masking—tower A learns on personalized traffic, tower B on non-personalized traffic using the same inputs—plus a separate non-personalized pathway trained only on non-personalized features, while a distillation-inspired L1 logit-level alignment loss pulls the two pathways' predictions together. During training only, the non-personalized pathway absorbs knowledge from the richer personalized pathway; at inference it is discarded and the dual-tower component alone makes predictions, choosing the appropriate tower according to featu

What carries the argument

The load-bearing mechanism is the masking-and-alignment pathway pair: a personalized pathway containing two parallel towers with shared embeddings but independent parameters, whose outputs are gated by a per-sample mask so that tower A only receives supervision on personalized traffic and tower B only on non-personalized traffic; a separate non-personalized pathway that consumes non-personalized features only; and a distance loss computed as the mean absolute difference between the two pathways' logits on non-personalized samples, formulated through a cross-entropy-style objective to align their final predictions. The key design choice is logit-level rather than feature-level alignment, whic

Load-bearing premise

The paper's headline result rests on assuming that removing personalized features from otherwise ordinary traffic is a faithful stand-in for traffic from users who actually withheld consent; the authors themselves state that masking 'may not fully capture all characteristics of real non-consented traffic.'

What would settle it

Log real traffic with true consent labels, run RAMP and the best baseline on the non-consenting subset only, and compare non-personalized AUC; if RAMP's advantage disappears, reverses, or falls below roughly 0.1% AUC, the transfer effect demonstrated under synthetic masking would not carry over to actual deployment.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • RAMP is model-agnostic: it improves non-personalized prediction when plugged into each of the evaluated backbone models.
  • The gains span task types and scales: CTR on two public datasets and CVR on one public and one industrial dataset, with the largest AUC gain on the largest dataset.
  • Inference is unchanged: only the dual-tower component is needed at serving time, so the privacy-constrained regime adds no extra latency or model complexity at runtime.
  • Even with heavy personalized-feature loss during training, non-personalized AUC degrades only slightly, suggesting graceful handling of partial feature availability.
  • An industry A/B test reported in the paper shows total advertiser value improving by more than 3% with RAMP.

Where Pith is reading between the lines

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

  • If masking faithfully mimics consent-constrained traffic, the method implies that platforms could offer near-personalized ad quality on non-consenting traffic without storing user-linked data—a direct improvement in the privacy–quality trade-off.
  • The success of logit-level over feature-level alignment suggests a transferable design principle: when two input regimes differ, align final predictions rather than internal representations; this is testable in other missing-feature prediction problems.
  • The mild degradation under progressive feature removal hints that RAMP could support gradual tightening of consent policies rather than only binary all-or-nothing feature availability.
  • The smaller gain on the one natural-traffic dataset compared with the synthetic-masking benchmarks suggests real-world improvements may be quieter than headline numbers, pointing to the value of testing on naturally non-consenting traffic.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes RAMP, a model-agnostic training framework for CTR/CVR prediction when personalized features are unavailable at inference. RAMP combines (i) a dual-tower 'personalized pathway' whose two towers are trained with output masking so that one specializes in personalized traffic and the other in non-personalized traffic; (ii) a separate non-personalized pathway trained only on non-personalized features; and (iii) a logit-level prediction-alignment loss that distills knowledge from the personalized pathway into the non-personalized side. At inference, the non-personalized tower is intended to serve traffic without personalized features. Experiments on Avazu, TaobaoAd, CriteoPrivateAd, and a private IndustryAd dataset report consistent non-personalized AUC gains over strong baselines (from +0.10 to +0.87 percentage points), competitive personalized AUC, and ablations showing the contributions of the dual-tower and alignment components. The code is publicly available.

Significance. If the results hold, RAMP is a practically useful, model-agnostic technique for ad ranking under consent-driven feature restrictions, with no extra inference-time cost. The paper's strengths include public code, evaluation on four datasets (including a natural-traffic industrial dataset), multiple backbone integrations, and a clear ablation of the two pathway components. The central idea—aligning a non-personalized predictor with a personalized predictor at the logit level—is simple and plausible, and the reported gains are consistent across datasets. However, the strongest public-benchmark evidence relies on a synthetic masking protocol whose validity for real non-consented traffic is the main open question. The natural-traffic IndustryAd gain is much smaller, and several reporting issues (missing error bars, a missing appendix for an A/B claim, and incomplete feature-category details) currently prevent full confidence in the quantitative claims.

major comments (5)
  1. [Section 4.1.2 / Abstract] The public-benchmark evidence is obtained by masking personalized fields of the same samples/users that also appear in personalized form in the training data. This evaluates robustness to feature absence under a matching population, not generalization to real non-consented traffic, where users may have different behavior. The paper acknowledges this in Section 4.1.2 and Section 5, but the abstract's claim 'when personalized features are missing' is broader. The natural-traffic IndustryAd result is much smaller (+0.11 AUC over the best baseline vs +0.87 on CriteoPrivateAd). Please either temper the abstract to 'under synthetic feature-masking and one natural-traffic deployment' or provide additional analysis (e.g., stratified results by user overlap or a covariate-shift experiment) to support the broader claim.
  2. [Section 4.1.2 / Table 1] The validity of the masking simulation depends entirely on which fields are labeled personalized vs non-personalized, but Table 1 lists only example field names. For Avazu, if a quasi-identifier such as device_id is retained in the non-personalized set, the 'non-personalized' input can still carry user/device identity and the reported gains would be inflated. Provide a complete list of personalized and non-personalized fields for all datasets and justify the categorization, especially for Avazu and TaobaoAd where device-level identifiers exist.
  3. [Section 4.2 / Table 2] The paper states that all RAMP improvements are statistically significant at p<1e-3, but no standard deviations, confidence intervals, or test details are reported. Several gains are 0.10–0.22 percentage points, which may or may not be significant across 5 seeds. Report mean and variance (or per-seed values) for all tables, state the exact test procedure, and clarify whether the comparisons are paired over seeds.
  4. [Section 4.2] The claim that 'RAMP was A/B tested in industry production for CVR prediction, the total advertiser value (TAV) improved over 3% (See Appendix??)' points to a missing appendix. This is an externally validated claim that cannot be checked. Either include the appendix with experimental details (duration, traffic split, metric definition, significance) or remove the sentence.
  5. [Section 4.1.4 vs Section 4.1.6] There is a direct contradiction about inference. Section 4.1.4 says 'during validation and testing, only the foundational dual-tower component is used ... The non-personalized pathway is thus not required during inference,' while Section 4.1.6 says 'when personalized features are unavailable, only the non-personalized pathway is activated.' This ambiguity affects the architecture description and the privacy-compliance claims. Clarify which component serves non-personalized traffic at inference.
minor comments (5)
  1. [Section 4.1.4] The phrase 'The training cost of the method is introduced in Appendix??' has a dangling appendix reference. Either add the appendix or remove the pointer.
  2. [Figure 3] The x-axis label 'Distance Loss Weight' would be clearer as 'β'; the caption should state which backbone is used in each panel, as it currently only mentions 'RAMP-FCN' and 'RAMP-PNN' in the body text.
  3. [Section 4.1.4] The sentence 'The proposed framework is evaluated in three configurations based on the backbone model: PP ... and RAMP' lists only two configurations. Please rephrase.
  4. [Section 4.5] The t-SNE visualization does not specify which embeddings are plotted—those from the personalized pathway, the non-personalized pathway, or a concatenation. Since the non-personalized pathway is not used at inference, the claim of 'feature separation' needs a precise definition of what is being visualized.
  5. [References] References [16] and [17] are duplicates of the same Hinton et al. work; consolidate.

Circularity Check

0 steps flagged

No significant circularity: RAMP is an empirical architecture paper; no predictive claim reduces by construction to its fitted inputs.

full rationale

RAMP's claims are empirical rather than derived: the training objective is L_overall = alpha * L_base + beta * L_dis, where L_base is BCE against ground-truth labels and L_dis is an L1/logit alignment between the dual-tower output and the non-personalized pathway (Eqs. 4-6). Neither loss is fitted to the reported Non-PER AUC, and the non-personalized pathway is explicitly not used at inference, so the evaluation is not self-referential. The dual-tower output masking (Eqs. 1-3) is a traffic-regime routing choice, not a definition of the evaluation metric. The paper's only overlapping self-citation, [18], is used as background for the energy/distillation interpretation of the alignment loss and is not load-bearing; the design choice is instead supported empirically by ablations against KD, PFD, and HAPFD in Section 4.3.2. Acknowledged limitations, such as the statement that masking 'may not fully capture all characteristics of real non-consented traffic' (Section 4.1.2), and unresolved 'Appendix??' references for A/B-test and training-cost details are external-validity or completeness concerns, not circular reductions. Overall, no derivation step reduces to its own inputs, so the circularity score is 0.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The paper contributes an architecture, so the ledger holds the modeling choices and data assumptions that the empirical claim rests on. No invented physical or formal entities are introduced.

free parameters (4)
  • beta (distance-loss weight) = 150 (TaobaoAd), 80 (CriteoPrivateAd); not reported for Avazu/IndustryAd
    Controls the strength of prediction alignment; performance is sensitive to it (inverted-U in Fig. 3a), so the reported gains depend on this tuned hyperparameter.
  • alpha (base-loss weight) = 1
    Fixed by design to simplify hyperparameter search; effectively a scale choice.
  • Backbone model selection = PNN for TaobaoAd/CriteoPrivateAd; FCN for Avazu/IndustryAd
    The backbone yielding best PP/RAMP validation performance is chosen per dataset; Table 2 results depend on this choice, and the selection criterion is not fully specified.
  • Embedding dimension and OOV frequency threshold = 16 or 32; OOV threshold 2 or 10
    Standard preprocessing hyperparameters tuned per dataset; not a scientific parameter but affects all results.
axioms (4)
  • domain assumption The feature categorization in Table 1 (e.g., device_ip and app_click_id are personalized; app_id, campaign_id, site_id are non-personalized) is correct.
    The entire evaluation and the masking protocol depend on this split; a misclassified field could change the meaning of the reported gains.
  • domain assumption Masking personalized features in public datasets is a valid proxy for non-consented traffic.
    Authors acknowledge that masking simulates feature absence but not distributional differences in real non-consenting users (Section 4.1.2).
  • domain assumption A 0.1% AUC improvement is practically meaningful in computational advertising.
    The paper adopts this convention from prior work [45, 51] and uses it to interpret small AUC gains as significant for practice.
  • ad hoc to paper Independent tower parameters with identical architectures yield the best empirical results.
    Section 3.1.2 states this choice 'yields the best empirical results' but no ablation with parameter sharing is shown; the design is load-bearing for the dual-tower mechanism.

pith-pipeline@v1.3.0-alltime-deepseek · 19915 in / 16214 out tokens · 149412 ms · 2026-08-01T17:51:11.370176+00:00 · methodology

0 comments
read the original abstract

Click-through rate (CTR) and conversion rate (CVR) prediction are fundamental tasks in online advertising, aiming to estimate the likelihood of user interactions based on various features. While personalized attributes such as age and gender can significantly enhance predictive accuracy, their use is increasingly restricted by privacy regulations, thereby limiting available data for both training and inference. To address this challenge, we propose RAMP (Robust Ad Recommendation Under Limited Personalized-Feature Availability via Masking and Alignment Pathways), which is designed to improve CTR/CVR prediction accuracy when personalized features are not accessible, thus supporting deployment in privacy-constrained settings.RAMP consists of (i) a personalized pathway built upon a dual-tower component with identical inputs but independent parameters, where output masking separates predictions for personalized and non-personalized signals, (ii) a separate non-personalized pathway trained with non-personalized features only, and (iii) a distillation-inspired prediction-alignment architecture between (i) and (ii) that improves prediction when personalized features are unavailable. We conduct comprehensive experiments using both public benchmarks and industrial datasets to evaluate the performance of RAMP. Our evaluation spans multiple backbone models and different settings: with and without access to personalized features. The results show that RAMP consistently outperforms state-of-the-art methods when personalized features are missing, while maintaining competitive performance when all features are available. %demonstrating its effectiveness and practicality for real-world advertising systems. Our code is publicly available at https://github.com/Ruixinhua/RAMP.

Figures

Figures reproduced from arXiv: 2607.17473 by Aghiles Salah, Aonghus Lawlor, Barry Smyth, Bichen Shi, Changhong Jin, Dairui Liu, Jitao Lu, Mete Sertkan, Roger Zhe Li, Ruihai Dong, Tri Kurniawan Wijaya, Xingsheng Guo, Xinyang Shao, Zhongyi Lu.

Figure 1
Figure 1. Figure 1: Motivation study based on the CriteoPrivateAd dataset. Removing personalized features causes a substantial performance drop for strong CTR/CVR baselines (PNN [41], FCN [24], FINAL [36]). Applying RAMP on top of each backbone partially recovers the lost accuracy, narrowing the gap under limited personalized features. 1 Introduction Online advertising relies on predicting user responses, such as click￾throug… view at source ↗
Figure 2
Figure 2. Figure 2: The proposed RAMP. The personalized and the non￾personalized pathways are trained jointly. A lightweight loss is adopted to enable the distillation-inspired architecture that allows knowledge to be aligned between the two components. Specifically, both towers use the same input and features for train￾ing. However, the prediction of the towers is selectively masked based on personalized and non-personalized… view at source ↗
Figure 3
Figure 3. Figure 3: Ablation results for RAMP. (a) Effect of the distance-loss weight 𝛽 on Non-PER performance: Non-PER AUC (blue crosses) and LogLoss (red dots) for RAMP-FCN on TaobaoAd (left) and RAMP-PNN on CriteoPrivateAd (right) across 𝛽 values. (b) Effect of dropping an increasing number of personalized feature fields on CriteoPrivateAd: Non-PER AUC (blue crosses) and LogLoss (red dots) for RAMP-PNN. As shown in [PITH_… view at source ↗
Figure 4
Figure 4. Figure 4: Visualization of feature separation on CriteoPrivateAd using PCA and t-SNE with 10,000 random samples. personalized and non-personalized features forming distinct, well￾separated clusters in different regions of the embedding space. Each cluster also exhibits tighter intra-class cohesion with reduced disper￾sion. This enhanced discriminability can be attributed to RAMP’s dual mechanisms: the dual-tower com… view at source ↗

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

69 extracted references · 3 canonical work pages

  1. [1]

    Jianxin Chang, Chenbin Zhang, Zhiyi Fu, Xiaoxue Zang, Lin Guan, Jing Lu, Yiqun Hui, Dewei Leng, Yanan Niu, Yang Song, and Kun Gai. 2023. TWIN: TWo-stage Interest Network for Lifelong User Behavior Modeling in CTR Prediction at Kuaishou. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining(Long Beach, CA, USA)(KDD ’23). As...

    doi:10.1145/3580305 2023

  2. [2]

    Bo Chen, Yichao Wang, Zhirong Liu, Ruiming Tang, Wei Guo, Hongkun Zheng, Weiwei Yao, Muyu Zhang, and Xiuqiang He. 2021. Enhancing Explicit and Implicit Feature Interactions via Information Sharing for Parallel Deep CTR Models. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management(Virtual Event, Queensland, Australia...

    doi:10.1145/3459637 2021

  3. [3]

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

  4. [4]

    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

  5. [5]

    Yuan Cheng. 2022. Dynamic Explicit Embedding Representation for Numerical Features in Deep CTR Prediction. InCIKM. ACM, 3888–3892

    2022

  6. [6]

    Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep neural networks for youtube recommendations. InProceedings of the 10th ACM conference on recommender systems. 191–198

    2016

  7. [7]

    Yongheng Deng, Guanbo Wang, Sheng Yue, Wei Rao, Qin Zu, Wenjie Wang, Shuai Chen, Ju Ren, and Yaoxue Zhang. 2024. RelayRec: Empowering Privacy- Preserving CTR Prediction via Cloud-Device Relay Learning. In2024 23rd ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN). 188–199. doi:10.1109/IPSN61024.2024.00020

    arXiv 2024

  8. [8]

    Yuval Dishi, Ophir Friedler, Yonatan Karni, Natalia Silberstein, and Yulia Stolin

  9. [9]

    Yingpeng Du, Zhu Sun, Ziyan Wang, Haoyan Chua, Jie Zhang, and Yew-Soon Ong

  10. [10]

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

    2019

  11. [11]

    Yu Gong, Ziwen Jiang, Yufei Feng, Binbin Hu, Kaiqi Zhao, Qingwen Liu, and Wenwu Ou. 2020. EdgeRec: recommender system on edge in Mobile Taobao. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management. 2477–2484

    2020

  12. [12]

    Active Large Language Model-Based Knowledge Distillation for Session- Based Recommendation. InThirty-Ninth AAAI Conference on Artificial Intelligence, Thirty-Seventh Conference on Innovative Applications of Artificial Intelligence, Fifteenth Symposium on Educational Advances in Artificial Intelligence, AAAI 2025, Philadelphia, PA, USA, February 25 - March...

    work page doi:10.1609/aaai.v39i11.33263 2025

  13. [13]

    Raia Hadsell, Sumit Chopra, and Yann LeCun. 2006. Dimensionality Reduction by Learning an Invariant Mapping. In2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’06). IEEE, 1735–1742. doi:10. 1109/CVPR.2006.100

    2006

  14. [14]

    Jialiang Han, Yun Ma, Qiaozhu Mei, and Xuanzhe Liu. 2021. DeepRec: On- device Deep Learning for Privacy-Preserving Sequential Recommendation in Mobile Commerce. InProceedings of the Web Conference 2021(Ljubljana, Slovenia) (WWW ’21). Association for Computing Machinery, New York, NY, USA, 900–911. doi:10.1145/3442381.3449942

    arXiv 2021

  15. [15]

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

    2017

  16. [17]

    Hinton, Oriol Vinyals, and Jeffrey Dean

    Geoffrey E. Hinton, Oriol Vinyals, and Jeffrey Dean. 2015. Distilling the Knowl- edge in a Neural Network.CoRRabs/1503.02531 (2015)

    Pith/arXiv arXiv 2015

  17. [18]

    Yu, and Guandong Xu

    Li He, Hongxu Chen, Dingxian Wang, Shoaib Jameel, Philip S. Yu, and Guandong Xu. 2021. Click-Through Rate Prediction with Multi-Modal Hypergraphs. In CIKM. ACM, 690–699

    2021

  18. [19]

    Tongwen Huang, Zhiqi Zhang, and Junlin Zhang. 2019. FiBiNET: combining fea- ture importance and bilinear feature interaction for click-through rate prediction. InProceedings of the 13th ACM Conference on Recommender Systems(Copenhagen, Denmark)(RecSys ’19). Association for Computing Machinery, New York, NY, USA, 169–177. doi:10.1145/3298689.3347043

    arXiv 2019

  19. [20]

    Jiazheng Jing, Yinan Zhang, Xin Zhou, and Zhiqi Shen. 2023. Capturing Pop- ularity Trends: A Simplistic Non-Personalized Approach for Enhanced Item Recommendation. InCIKM. ACM, 1014–1024

    2023

  20. [21]

    Sebastian Hofstätter, Sophia Althammer, Michael Schröder, Mete Sertkan, and Allan Hanbury. 2020. Improving Efficient Neural Ranking Models with Cross- Architecture Knowledge Distillation. arXiv:2010.02666 [cs.IR] doi:10.48550/arXiv. 2010.02666

    Pith/arXiv arXiv doi:10.48550/arxiv.2010.02666 2020

  21. [22]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. 2015. Adam: A Method for Stochastic Opti- mization. InICLR (Poster)

    2015

  22. [23]

    Yann LeCun, Sumit Chopra, Raia Hadsell, Marc’Aurelio Ranzato, and Fu-Jie Huang. 2006. A Tutorial on Energy-Based Learning. InPredicting Structured Data, Gökhan Bakir, Thomas Hofmann, Bernhard Schölkopf, Alexander J. Smola, and Ben Taskar (Eds.). MIT Press, Cambridge, MA. doi:10.7551/mitpress/7443.003.0014

    work page doi:10.7551/mitpress/7443.003.0014 2006

  23. [24]

    SeongKu Kang, Junyoung Hwang, Wonbin Kweon, and Hwanjo Yu. 2020. DE-RRD: A Knowledge Distillation Framework for Recommender System. InCIKM ’20: The 29th ACM International Conference on Information and Knowledge Management, Virtual Event, Ireland, October 19-23, 2020, Mathieu d’Aquin, Stefan Dietze, Claudia Hauff, Edward Curry, and Philippe Cudré-Mauroux (...

    arXiv 2020

  24. [25]

    Yang Li, Kangbo Liu, Ranjan Satapathy, Suhang Wang, and Erik Cambria. 2024. Recent Developments in Recommender Systems: A Survey [Review Article].IEEE Computational Intelligence Magazine19, 2 (2024), 78–95. doi:10.1109/MCI.2024. 3363984

    doi:10.1109/mci.2024 2024

  25. [26]

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

    arXiv 2018

  26. [27]

    Honghao Li, Yiwen Zhang, Yi Zhang, Hanwei Li, Lei Sang, and Jieming Zhu

  27. [28]

    arXiv:2407.13349 [cs.IR] https://arxiv.org/abs/2407.13349

    FCN: Fusing Exponential and Linear Cross Network for Click-Through Rate Prediction. arXiv:2407.13349 [cs.IR] https://arxiv.org/abs/2407.13349

    arXiv

  28. [29]

    Congcong Liu, Yuejiang Li, Jian Zhu, Fei Teng, Xiwei Zhao, Changping Peng, Zhangang Lin, and Jingping Shao. 2022. Position Awareness Modeling with Knowledge Distillation for CTR Prediction. InRecSys ’22: Sixteenth ACM Confer- ence on Recommender Systems, Seattle, W A, USA, September 18 - 23, 2022, Jennifer Golbeck, F. Maxwell Harper, Vanessa Murdock, Mich...

    arXiv 2022

  29. [30]

    Jie Liu, Yinrui Li, Jiankai Sun, Kungang Li, Han Sun, Sihan Wang, Huasen Wu, Siyuan Gao, Paulo Soares, Nan Li, Zhifang Liu, Haoyang Li, Siping Ji, Ling Leng, and Prathibha Deshikachar. 2025. Decoupled Entity Representation Learning for Pinterest Ads Ranking. InRecSys. ACM, 940–944

    2025

  30. [31]

    Mingfu Liang, Xi Liu, Rong Jin, Boyang Liu, Qiuling Suo, Qinghai Zhou, Song Zhou, Laming Chen, Hua Zheng, Zhiyuan Li, Shali Jiang, Jiyan Yang, Xiaozhen Xia, Fan Yang, Yasmine Badr, Ellie Wen, Shuyu Xu, Hansey Chen, Zhengyu Zhang, Jade Nie, Chunzhi Yang, Zhichen Zeng, Weilin Zhang, Xingliang Huang, Qianru Li, Shiquan Wang, Evelyn Lyu, Wenjing Lu, Rui Zhang...

    2025

  31. [32]

    Bo Liu, Ming Ding, Sina Shaham, Wenny Rahayu, Farhad Farokhi, and Zihuai Lin. 2022. When Machine Learning Meets Privacy: A Survey and Outlook.ACM Comput. Surv.54, 2 (2022), 31:1–31:36

    2022

  32. [33]

    Wenyu Ma, Hongbin Xia, and Yuan Liu. 2025. DiffKD: collaborative graph diffusion with knowledge distillation for multimodal recommendation.J. Intell. Inf. Syst.63, 5 (2025), 1487–1510. doi:10.1007/S10844-025-00946-4

    work page doi:10.1007/s10844-025-00946-4 2025

  33. [34]

    Xiao Ma, Hongyu Zhang, Jiangfeng Zeng, Yiqi Duan, and Xuan Wen. 2024. FedKGRec: privacy-preserving federated knowledge graph aware recommender system.Appl. Intell.54, 19 (2024), 9028–9044

    2024

  34. [35]

    Jing Liu, Litao Shang, Yuting Su, Weizhi Nie, Xin Wen, and Anan Liu. 2024. Privacy-preserving Multi-source Cross-domain Recommendation Based on Knowledge Graph.ACM Trans. Multim. Comput. Commun. Appl.20, 5 (2024), 148:1–148:18

    2024

  35. [36]

    Xiaofan Liu, Qinglin Jia, Chuhan Wu, Jingjie Li, Quanyu Dai, Lin Bo, Rui Zhang, and Ruiming Tang. 2023. Task Adaptive Multi-learner Network for Joint CTR and CVR Estimation. InWWW (Companion Volume). ACM, 490–494

    2023

  36. [37]

    Kelong Mao, Jieming Zhu, Liangcai Su, Guohao Cai, Yuru Li, and Zhenhua Dong

  37. [38]

    Peter Müllner. 2023. User Privacy in Recommender Systems. InECIR (3) (Lecture Notes in Computer Science, Vol. 13982). Springer, 456–461

    2023

  38. [39]

    Xiao Ma, Liqin Zhao, Guan Huang, Zhi Wang, Zelin Hu, Xiaoqiang Zhu, and Kun Gai. 2018. Entire Space Multi-Task Model: An Effective Approach for Estimating Post-Click Conversion Rate. InThe 41st International ACM SIGIR Conference on Research & Development in Information Retrieval(Ann Arbor, MI, USA)(SIGIR ’18). Association for Computing Machinery, New York...

    arXiv 2018

  39. [40]

    Kelong Mao, Jieming Zhu, Liangcai Su, Guohao Cai, Yuru Li, and Zhenhua Dong. 2023. FinalMLP: an enhanced two-stream MLP model for CTR pre- diction. InProceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artifi- cial Intelligence and Thirteenth Symposium on Educational Advanc...

    doi:10.1609/aaai.v37i4.25577 2023

  40. [41]

    Yanru Qu, Han Cai, Kan Ren, Weinan Zhang, Yong Yu, Ying Wen, and Jun Wang

  41. [42]

    Omid Rafieian and Hema Yoganarasimhan. 2020. Targeting and Privacy in Mobile Advertising.Marketing Science(Dec. 2020). doi:10.1287/mksc.2020.1235

    arXiv 2020

  42. [43]

    Arpit Rana, Scott Sanner, Mohamed Reda Bouadjenek, Ronald Di Carlantonio, and Gary Farmaner. 2024. User Experience and the Role of Personalization in Critiquing-Based Conversational Recommendation.ACM Trans. Web18, 4 (2024), 43:1–43:21

    2024

  43. [44]

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. 2019. Pytorch: An imperative style, high-performance deep learning library.Advances in neural information processing systems32 (2019)

    2019

  44. [45]

    Liang Qu, Wei Yuan, Ruiqi Zheng, Lizhen Cui, Yuhui Shi, and Hongzhi Yin. 2024. Towards personalized privacy: User-governed data contribution for federated rec- ommendation. InProceedings of the ACM Web Conference 2024. ACM, Singapore, 3910–3918

    2024

  45. [46]

    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 prediction. InProceedings of the 46th international ACM SIGIR conference on research and development in information retrieval. 1376–1385

    2023

  46. [47]

    Product-Based Neural Networks for User Response Prediction. InICDM. IEEE Computer Society, 1149–1154

  47. [48]

    Fangye Wang, Yingxu Wang, Dongsheng Li, Hansu Gu, Tun Lu, Peng Zhang, and Ning Gu. 2022. Enhancing CTR Prediction with Context-Aware Feature Representation Learning. InSIGIR. ACM, 343–352

    2022

  48. [49]

    Fangye Wang, Yingxu Wang, Dongsheng Li, Hansu Gu, Tun Lu, Peng Zhang, and Ning Gu. 2023. CL4CTR: A Contrastive Learning Framework for CTR Prediction. InWSDM. ACM, Singapore, 805–813

    2023

  49. [50]

    Mehdi Sebbar, Corentin Odic, Mathieu Léchine, Aloïs Bissuel, Nicolas Chrysan- thos, Anthony D’Amato, Alexandre Gilotte, Fabian Höring, Sarah Nogueira, and Maxime Vono. 2025. CriteoPrivateAds: A Real-World Bidding Dataset to Design Private Advertising Systems.arXiv preprint arXiv:2502.12103(2025)

    Pith/arXiv arXiv 2025

  50. [51]

    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. 1161–1170

    2019

  51. [52]

    Steve Wang and Will Cukierski. 2014. Click-Through Rate Prediction. Kaggle competition. https://kaggle.com/competitions/avazu-ctr-prediction

    2014

  52. [53]

    Fangye Wang, Hansu Gu, Dongsheng Li, Tun Lu, Peng Zhang, and Ning Gu. 2023. Towards Deeper, Lighter and Interpretable Cross Network for CTR Prediction. InProceedings of the 32nd ACM International Conference on Information and Knowledge Management(Birmingham, United Kingdom)(CIKM ’23). Association for Computing Machinery, New York, NY, USA, 2523–2533. doi:...

    doi:10.1145/3583780 2023

  53. [54]

    Chen Xu, Quan Li, Junfeng Ge, Jinyang Gao, Xiaoyong Yang, Changhua Pei, Fei Sun, Jian Wu, Hanxiao Sun, and Wenwu Ou. 2020. Privileged Features Distillation at Taobao Recommendations. InKDD ’20: The 26th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Virtual Event, CA, USA, August 23-27, 2020, Rajesh Gupta, Yan Liu, Jiliang Tang, and B. Adit...

    arXiv 2020

  54. [55]

    Yanwu Yang and Panyu Zhai. 2022. Click-through rate prediction in online advertising: A literature review.Information Processing & Management59, 2 (2022), 102853. doi:10.1016/j.ipm.2021.102853

    arXiv 2022

  55. [56]

    Ruoxi Wang, Bin Fu, Gang Fu, and Mingliang Wang. 2017. Deep & Cross Network for Ad Click Predictions. InProceedings of the ADKDD’17(Halifax, NS, Canada) (ADKDD’17). Association for Computing Machinery, New York, NY, USA, Article 12, 7 pages. doi:10.1145/3124749.3124754

    arXiv 2017

  56. [57]

    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. InProceedings of the Web Conference 2021(Ljubljana, Slovenia)(WWW ’21). Association for Computing Machinery, New York, NY, USA, 1785–1797. doi:10.1145/3442381.3450078

    arXiv 2021

  57. [58]

    Weinan Zhang, Jiarui Qin, Wei Guo, Ruiming Tang, and Xiuqiang He. 2021. Deep Learning for Click-Through Rate Estimation. InIJCAI. ijcai.org, 4695–4703

    2021

  58. [59]

    Zhiqiang Wang, Qingyun She, and Junlin Zhang. 2021. MaskNet: Introducing Feature-Wise Multiplication to CTR Ranking Models by Instance-Guided Mask. arXiv:2102.07619 [cs.IR] https://arxiv.org/abs/2102.07619

    Pith/arXiv arXiv 2021

  59. [60]

    Jieming Zhu, Quanyu Dai, Liangcai Su, Rong Ma, Jinyang Liu, Guohao Cai, Xi Xiao, and Rui Zhang. 2022. BARS: Towards Open Benchmarking for Recom- mender Systems. InSIGIR ’22: The 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, Madrid, Spain, July 11 - 15, 2022, Enrique Amigó, Pablo Castells, Julio Gonzalo, Ben ...

    arXiv 2022

  60. [61]

    Jieming Zhu, Qinglin Jia, Guohao Cai, Quanyu Dai, Jingjie Li, Zhenhua Dong, Ruiming Tang, and Rui Zhang. 2023. FINAL: Factorized Interaction Layer for CTR Prediction. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval(Taipei, Taiwan)(SIGIR ’23). Association for Computing Machinery, New York, N...

    arXiv 2023

  61. [62]

    Jiangchao Yao, Feng Wang, Kunyang Jia, Bo Han, Jingren Zhou, and Hongxia Yang

  62. [64]

    Huining Yuan, Wenpeng Zhang, Zijie Hao, and Zengde Deng. 2025. Hardness- aware Privileged Features Distillation with Latent Alignment for CVR Prediction. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, V.2, KDD 2025, Toronto ON, Canada, August 3-7, 2025, Luiza Antonie, Jian Pei, Xiaohui Yu, Flavio Chierichetti, Hady...

    arXiv 2025

  63. [66]

    Han Zhu, Junqi Jin, Chang Tan, Fei Pan, Yifan Zeng, Han Li, and Kun Gai. 2017. Optimized Cost per Click in Taobao Display Advertising. InProceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining(Halifax, NS, Canada)(KDD ’17). Association for Computing Machinery, New York, NY, USA, 2191–2200. doi:10.1145/3097983.3098134

    arXiv 2017

  64. [69]

    Jieming Zhu, Jinyang Liu, Shuai Yang, Qi Zhang, and Xiuqiang He. 2021. Open Benchmarking for Click-Through Rate Prediction. InCIKM ’21: The 30th ACM International Conference on Information and Knowledge Management, Virtual Event, Queensland, Australia, November 1 - 5, 2021, Gianluca Demartini, Guido Zuccon, J. Shane Culpepper, Zi Huang, and Hanghang Tong ...

    2021

  65. [2016]

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

    Wide & Deep Learning for Recommender Systems. InProceedings of the 1st Workshop on Deep Learning for Recommender Systems(Boston, MA, USA) (DLRS 2016). Association for Computing Machinery, New York, NY, USA, 7–10. doi:10.1145/2988450.2988454

    arXiv 2016

  66. [2021]

    InProceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining

    Device-cloud collaborative learning for recommendation. InProceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. 3865–3874

  67. [2023]

    arXiv:2304.00902 (Nov

    FinalMLP: An Enhanced Two-Stream MLP Model for CTR Prediction. arXiv:2304.00902 (Nov. 2023). doi:10.48550/arXiv.2304.00902 arXiv:2304.00902 [cs]

    Pith/arXiv arXiv doi:10.48550/arxiv.2304.00902 2023

  68. [2025]

    InRecSys

    Practical Multi-Task Learning for Rare Conversions in Ad Tech. InRecSys. ACM, 1042–1045

  69. [2769]

    doi:10.1145/3459637.3482486

    arXiv