Pith. sign in

REVIEW 3 major objections 6 minor 2 cited by

Learning Universal User Representations Leveraging Cross-domain User Intent at Snapchat

T0 review · 3 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read One cross-surface user embedding, added as an auxiliary feature, lifted engagement across four Snapchat recommender systems.

desk verdict Plausible but underverified industrial claim: cross-domain user embeddings help four Snapchat surfaces, but A/B methodology is missing. read the letter →

arxiv 2504.21838 v2 pith:UZVEL3CO submitted 2025-04-30 cs.IR

Clark Mingxuan Ju , Leonardo Neves , Bhuvesh Kumar , Liam Collins , Tong Zhao , Yuwei Qiu , Qing Dou , Yang Zhou
show 6 more authors
Sohail Nizam Rengim Ozturk Yvette Liu Sen Yang Manish Malik Neil Shah
This is my paper · ORCID
classification cs.IR
keywords RecommenderSystemsUniversalUserModelingSequentialRecommendationCross-domainrepresentationlearningEmbedding-basedretrievalA/BtestingMulti-task
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

The paper claims that a single universal user embedding, trained on each user's timestamp-ordered interactions stitched together from Snapchat's various surfaces, can capture cross-domain patterns that per-surface representations miss. Adding this embedding as an auxiliary feature to four production recommender stages reportedly lifted long-form video retrieval open rate by 2.78%, long-form video ranking view-time sum by 19.2%, Lens ranking play time by 1.76%, and notification ranking open rate by 0.87%. The paper also reports offline gains from two architectural refinements, domain-specific encoders and information-bottleneck attention tokens, on next-event retrieval (Recall@20 and NDCG@20). A sympathetic reader would take the central claim to be that one general-purpose representation enriches existing domain-specific representations across an entire platform rather than replacing them.

What carries the argument

The load-bearing object is the stitched, timestamp-sorted, platform-wise event sequence: every user's events from Content, Ads, Lens, and other surfaces are tabularized with null-padded sparse columns, capped at 5,000 events, and cut into 800-event subsequences with a sliding window for training. Each event is encoded as an FFN of its feature embeddings and its domain tag; stacked multi-head self-attention layers then read the sequence, and weighted-sum pooling produces the universal user embedding. This sequence and pooling design is what carries collaborative filtering signals across surfaces into one vector, and that vector is what downstream retrieval and ranking models consume as an auxiliary feature.

What would settle it

Serve the same four production models with the UUM feature removed while holding every other model, feature, and traffic allocation fixed, and check whether the reported lifts persist with confidence intervals; alternatively, train UUM on sequences whose cross-surface order is scrambled so genuine cross-domain alignment is destroyed, and test whether the offline Recall@20 and NDCG@20 gains vanish.

Watch

Extended reading notes

Core claim

On its own terms, the discovery is that cross-domain sequential modeling over stitched user-event logs yields a user representation that is useful in already-mature retrieval and ranking systems. The UUM model maps each event's features plus its domain tag into a shared latent space, applies stacked multi-head self-attention, and pools the resulting token embeddings into a single user vector. Training is next-event prediction with sampled softmax and in-batch negatives, along with auxiliary event-property prediction heads. In production experiments, injecting this vector as an auxiliary feature into four different downstream models produced the reported engagement lifts. The paper further reports that two architectural variants, domain-specific encoders with a shared self-attention layer and information-bottleneck tokens that restrict cross-domain attention flow, improve offline Recall@20 and NDCG@20 compared to the base model.

Load-bearing premise

The reported production gains are attributed to the UUM embeddings, but the paper presents no experiment design, user allocation, duration, or significance tests, so concurrent model or traffic changes could in principle explain the lifts.

Editorial extensions

If this is right

  • A single user embedding trained once can be reused across retrieval and ranking stages and across content types, reducing the need for each surface to train its own cross-domain representation.
  • Cross-domain signals remain valuable even after each surface has its own mature ranker; the reported long-form video ranking lift of 19.2% in view-time sum is the largest such gain in the paper.
  • Because UUM embeddings are appended as auxiliary features rather than replacements, existing domain-specific representations keep their role and the new vector adds complementary information.
  • The offline results suggest that explicitly separating domain-specific encoding and gating cross-domain sharing through information bottlenecks may be the next step for imbalanced cross-domain data.

Reading between the lines

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

  • The authors leave implicit that a platform-wide embedding trained on all surfaces could help cold or sparse users in any single surface, since behavior in other surfaces can stand in for missing history; this could be tested by measuring single-surface cold-start recall with and without UUM.
  • A direct test beyond the paper would compare UUM against simply concatenating per-domain embeddings in the downstream models; the paper reports production gains but not that comparison.
  • The reported 19.2% view-time lift is much larger than the other lifts; if it replicates, it suggests the long-form video ranking stage previously had weak cross-domain signal, and if it does not, the A/B metric set would deserve re-examination.
  • The stitching and null-padding recipe, timestamp-sorted multi-surface events with a 5,000-event cap and 800-length subsequences, is a transferable template for other multi-surface platforms, though the paper does not claim generality.
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

3 major / 6 minor

Summary. The manuscript describes a production universal user modeling (UUM) system at Snapchat. The authors train a Transformer-based user encoder on timestamp-ordered event sequences stitched across multiple in-app surfaces, using a next-event prediction objective with auxiliary property-prediction heads, and then inject the resulting user embedding as an auxiliary feature into downstream retrieval and ranking models. The paper reports positive A/B metrics for four applications (long-form video EBR and L2 ranking, Lens L2 ranking, and Notification L2 ranking) and offline gains for two proposed architecture refinements: domain-specific encoders and information-bottleneck tokens.

Significance. If the reported A/B lifts are reliable, the paper provides valuable industrial evidence that a single cross-domain user representation can improve multiple downstream recommender surfaces, which is a practically important result. The manuscript is also commendable for describing a concrete and relatively simple architecture, for being explicit about data stitching and augmentation choices, and for separating launch results from research directions still under iteration. However, the evidence supporting the central causal claim is incomplete: the A/B results are point estimates without experimental design or uncertainty quantification, and the offline evaluation is based on the model's own training objective. These weaknesses currently limit the paper from being a definitive demonstration of the claimed cross-domain gains.

major comments (3)
  1. [Section 3.2, Tables 1-4] The central causal claim---that adding UUM embeddings 'led to' the reported engagement improvements---is not supported by the evidence presented. The manuscript reports only point estimates (e.g., +2.78% Long-form Video Open Rate, +19.20% Long-form Video View Time Sum) with no randomization unit, traffic allocation, experiment duration, confidence intervals, p-values, or statement about concurrent changes in the downstream rankers. Without these details, the numbers could reflect statistical noise, multiple-comparison selection, or simultaneous feature/model updates. This is load-bearing because the paper's conclusion rests entirely on these attributions. I request an appendix or expanded section specifying the A/B methodology and uncertainty, or, if internal policy prevents disclosure, a clear reframing of these numbers as observed internal metrics rather than validated causal effects.
  2. [Section 3.3.3, Table 5] The offline evaluation assesses the two proposed modeling choices on next-event retrieval, which is the same predictive task used to train the model, and samples 50,000 random negatives per test example without reporting how the candidate pool is constructed or how sensitive the results are to the negative count. This is essentially measuring how well the variants optimize their training objective, and the resulting Recall@20/NDCG@20 differences (e.g., 0.483 vs. 0.519 for Recall@20) may not transfer to downstream retrieval or ranking quality. The table also lacks any variance estimate, so it is unclear whether the reported gaps are meaningful. I recommend adding a downstream or held-out evaluation, and at minimum reporting bootstrap confidence intervals and a sensitivity analysis over the number of negatives.
  3. [Abstract, Section 4, and Tables 1-4] There is an internal inconsistency in the naming of the headline metrics. The abstract states that the EBR launch produced a 2.78% increase in Long-form Video Open Rate, while the conclusion states the same launch produced a 2.78% increase in Story Open Rate. This discrepancy makes it difficult to connect the reported numbers to specific, documented experiments and weakens the traceability of the central claims. Please reconcile the metric names across the abstract, tables, and conclusion, and ensure each table label matches the metric described in the text.
minor comments (6)
  1. [Equation (1), Section 3.1.1] The notation is slightly inconsistent: the text describes FFN(\cdot,\cdot) as a function of two arguments, but Equation (1) uses FFN(Concat(Feature(x_i), d(x_i))) with one argument. Please clarify whether d(x_i) is a domain embedding or one-hot indicator and define it explicitly.
  2. [Section 3.1.2] The training objective uses 'sample softmax with in-batch negatives' but does not report the number of negatives, the loss weighting between the next-event task and the property prediction tasks, or how the property prediction tasks are supervised. These details are needed for reproducibility.
  3. [Figure 3 and Section 3.1.2] The pooling operation is only described as 'weighted summation' but the weights and how they are computed are not specified. Please provide the formal definition or state that the weights are learned and describe their parameterization.
  4. [Section 3, Raw Data Generation] The terms 'high-intent events' and 'low-intent events' are used without an operational definition. Please specify the criteria used to rank or trim events by intent, since this affects the input sequence construction.
  5. [Table 5] The offline evaluation table does not report the number of test examples, the candidate set size, or the exact definition of Recall@20 and NDCG@20 used. Please include these details so the reader can interpret the metric values.
  6. [References] Reference [3] is a placeholder with 'arXiv:YYYY.NNNNN' and reference [20] has an incomplete venue listing. Both need to be completed before publication.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the production A/B lifts are external live benchmarks, independent of the UUM training objective, and the only minor self-citation ([3], IB token mechanism) is not load-bearing for the central claim.

full rationale

The paper's central claim is that UUM embeddings improved engagement in four production systems, supported by A/B metrics in Section 3.2 (Tables 1-4). These are external live benchmarks, not quantities defined by the training loss, so the central result is not equivalent to its inputs by construction. The offline evaluation in Section 3.3.3 (Table 5) reuses the next-event retrieval task, which is close to the training objective, but it compares model variants on held-out next events with random negatives; this is a standard generalization check, not a circular 'prediction' of fitted values, although it does not independently establish cross-domain causal transfer. The only self-citation appearing in a load-bearing role is [3] in Section 3.3.2, where the IB token cross-domain exchange mechanism is attributed to the authors' own unpublished preprint; however, the empirical comparison in Table 5 provides the actual evidence for the IB variant, and [3] is not used to justify the headline production claims. The reported A/B numbers lack experiment design, sample sizes, confidence intervals, and significance testing; this is a correctness and credibility concern, not a circularity one. No equation or parameter is fitted to the reported targets and then renamed as a prediction, so no significant circularity is found.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the validity of the production A/B measurements and on the shared-feature-space assumption for cross-domain sequences. The paper makes several hand-chosen design choices (subsequence length 800, sequence cap 5,000, 50,000 negatives) that affect the reported numbers, and it introduces no new postulated entities. The key unstated inputs are the A/B experiment design and the model hyperparameters.

free parameters (4)
  • Subsequence length = 800
    User sequences are divided into 800-event subsequences as training examples; chosen as a balance between long-range dependencies and compute, 'a compromise supported by empirical results' (§3).
  • Maximum sequence length = 5,000
    Event sequences are capped at 5,000 to prevent a small number of long-sequence users from dominating training (§3).
  • Offline evaluation negative count = 50,000
    Recall@20 and NDCG@20 are computed against 50,000 random negatives per test example (§3.3.3), so the absolute offline numbers depend on this hand-set sampling budget.
  • Model architecture sizes (embedding dimension f, layers L, heads, loss weights)
    Not disclosed; these hand-chosen hyperparameters determine UUM's capacity and affect reported A/B and offline results.
assumptions (4)
  • domain assumption User behaviors across domains are drawn from a sufficiently uniform distribution to be represented in a shared feature space.
    Stated as a critical assumption of the shared encoder in §3.3.1; the base production UUM model rests on it.
  • domain assumption Next-event prediction with sample softmax and in-batch negatives is a valid surrogate objective for learning useful cross-domain user representations.
    The model is trained to maximize next-event retrieval probability; the paper provides no analysis linking this objective to downstream ranking metrics beyond the reported A/B results (§3.1.2).
  • domain assumption The reported A/B and offline metrics are computed on representative, stable populations and are statistically meaningful.
    No experimental setup, significance tests, or confidence intervals are provided for Tables 1-5 (§3.2, §3.3.3).
  • standard math Self-attention, sample softmax, and information bottleneck mechanisms work as described in the cited literature.
    Standard ML machinery taken from [4,11,20,23]; no new mathematical derivations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Universal User Representations Leveraging Cross-domain User Intent at Snapchat." pith.science (2026). https://pith.science/paper/UZVEL3CO

@misc{pith2026250421838,
  author       = {Pith},
  title        = {Pith review of: Learning Universal User Representations Leveraging Cross-domain User Intent at Snapchat},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UZVEL3CO}},
  note         = {Machine review of arXiv:2504.21838}
}
read the original abstract

The development of powerful user representations is a key factor in the success of recommender systems (RecSys). Online platforms employ a range of RecSys techniques to personalize user experience across diverse in-app surfaces. User representations are often learned individually through user's historical interactions within each surface and user representations across different surfaces can be shared post-hoc as auxiliary features or additional retrieval sources. While effective, such schemes cannot directly encode collaborative filtering signals across different surfaces, hindering its capacity to discover complex relationships between user behaviors and preferences across the whole platform. To bridge this gap at Snapchat, we seek to conduct universal user modeling (UUM) across different in-app surfaces, learning general-purpose user representations which encode behaviors across surfaces. Instead of replacing domain-specific representations, UUM representations capture cross-domain trends, enriching existing representations with complementary information. This work discusses our efforts in developing initial UUM versions, practical challenges, technical choices and modeling and research directions with promising offline performance. Following successful A/B testing, UUM representations have been launched in production, powering multiple use cases and demonstrating their value. UUM embedding has been incorporated into (i) Long-form Video embedding-based retrieval, leading to 2.78% increase in Long-form Video Open Rate, (ii) Long-form Video L2 ranking, with 19.2% increase in Long-form Video View Time sum, (iii) Lens L2 ranking, leading to 1.76% increase in Lens play time, and (iv) Notification L2 ranking, with 0.87% increase in Notification Open Rate.

Figures

Figures reproduced from arXiv: 2504.21838 by the authors.

Figure 1
Figure 1. Overview of universal user modeling (UUM). [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Model architecture of UUM. This may involve operations like looking up ID embeddings for categorical features and concatenating the results. 𝑑 (𝑥𝑖) denotes the domain for 𝑥𝑖 , Concat(·) refers to the concatenation operation, and FFN(·, ·) is a function that projects feature vectors from different domains into a shared latent space. With the latent representation H ∈ R 𝑀×𝑓 for 𝑋, we feed H through stacked 𝐿 multi-hea… view at source ↗
Figure 4
Figure 4. Domain-specific attention with IB. User Tower with Domain-specific Encoders Domain￾specific Self-attention Shared Self-attention across Domains … … … Content … Ads … Lens Domain￾specific Self-attention Domain￾specific Self-attention Stitched Sequence Pooling User Representation [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: User tower in UUM with domain-specific encoders. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Beyond Fixed Depths and Widths: Optimizing Textual Decoding Tries in LLM-based Generative Recommendation

    cs.IR 2026-07 conditional novelty 6.0 of 10

    BONSAI constructs variable-depth, low-branching decoding tries for LLM-based generative recommendation and reports 16–22% relative gains over state-of-the-art baselines.

  2. Revisiting Self-attention for Cross-domain Sequential Recommendation

    cs.IR 2025-05 conditional novelty 6.0 of 10

    AutoCDSR improves cross-domain sequential recommendation by adding a Pareto-optimized penalty on cross-domain attention scores to the standard recommendation loss.

Reference graph

Works this paper leans on

29 extracted references · 6 canonical work pages · cited by 2 Pith papers

  1. [3]

    Clark Mingxuan Ju, Leonardo Neves, Bhuvesh Kumar, Liam Collins, Tong Zhao, Yuwei Qiu, Ching Dou, Sohail Nizam, Sen Yang, and Neil Shah. 2025. Revisiting Self-Attention for Cross-Domain Sequential Recommendation.arXiv preprint arXiv:YYYY.NNNNN(2025)

    work page 2025

  2. [1]

    Qixiang Fang, Zhihan Zhou, Francesco Barbieri, Yozen Liu, Leonardo Neves, Dong Nguyen, Daniel Oberski, Maarten Bos, and Ron Dotsch. 2024. General-Purpose User Modeling with Behavioral Logs: A Snapchat Case Study. InProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval. 2431–2436

    2024

  3. [2]

    Carlos A Gomez-Uribe and Neil Hunt. 2015. The netflix recommender system: Algorithms, business value, and innovation.ACM Transactions on Management Information Systems (TMIS)(2015)

    2015

  4. [4]

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

    2018

  5. [5]

    Matthew Kolodner, Mingxuan Ju, Zihao Fan, Tong Zhao, Elham Ghazizadeh, Yan Wu, Neil Shah, and Yozen Liu. 2024. Robust Training Objectives Improve Embedding-based Retrieval in Industrial Recommendation Systems.RobustRecSys Workshop at RecSys(2024)

    2024

  6. [6]

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

    2009

  7. [7]

    Pau Perng-Hwa Kung, Zihao Fan, Tong Zhao, Yozen Liu, Zhixin Lai, Jiahui Shi, Yan Wu, Jun Yu, Neil Shah, and Ganesh Venkataraman. 2024. Improving Embedding-Based Retrieval in Friend Recommendation with ANN Query Expan- sion. InProcs. of SIRIP

    work page 2024

  8. [8]

    Guanyu Lin, Chen Gao, Yu Zheng, Jianxin Chang, Yanan Niu, Yang Song, Kun Gai, Zhiheng Li, Depeng Jin, Yong Li, et al . 2024. Mixed Attention Network for Cross-domain Sequential Recommendation. InProceedings of the 17th ACM International Conference on Web Search and Data Mining. 405–413

    2024

Show all 29 references
  1. [9]

    Muyang Ma, Pengjie Ren, Yujie Lin, Zhumin Chen, Jun Ma, and Maarten de Rijke

  2. [10]

    Lev Muchnik, Sen Pei, Lucas C Parra, Saulo DS Reis, José S Andrade Jr, Shlomo Havlin, and Hernán A Makse. 2013. Origins of power-law degree distribution in the heterogeneity of human activity in social networks.Scientific reports3, 1 (2013), 1783

    2013

  3. [11]

    Arsha Nagrani, Shan Yang, Anurag Arnab, Aren Jansen, Cordelia Schmid, and Chen Sun. 2021. Attention bottlenecks for multimodal fusion.Advances in neural information processing systems34 (2021), 14200–14213

    2021

  4. [12]

    Aditya Pal, Chantat Eksombatchai, Yitong Zhou, Bo Zhao, Charles Rosenberg, and Jure Leskovec. 2020. Pinnersage: Multi-modal user embedding framework for recommendations at pinterest. InProcs. of SIGKDD

    2020

  5. [13]

    Chung Park, Taesan Kim, Taekyoon Choi, Junui Hong, Yelim Yu, Mincheol Cho, Kyunam Lee, Sungil Ryu, Hyungjun Yoon, Minsung Choi, et al. 2023. Cracking the Code of Negative Transfer: A Cooperative Game Theoretic Approach for Cross-Domain Sequential Recommendation. InProceedings ...

    2023

  6. [14]

    Chung Park, Taesan Kim, Hyungjun Yoon, Junui Hong, Yelim Yu, Mincheol Cho, Minsung Choi, and Jaegul Choo. 2024. Pacer and Runner: Cooperative Learning Framework between Single-and Cross-Domain Sequential Recommendation. InProceedings of the 47th International ACM SIGIR Confere...

    2024

  7. [15]

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

  8. [16]

    Aravind Sankar, Yozen Liu, Jun Yu, and Neil Shah. 2021. Graph neural networks for friend ranking in large-scale social platforms. InProceedings of the Web Conference 2021. 2535–2546

    2021

  9. [17]

    J Ben Schafer, Joseph Konstan, and John Riedl. 1999. Recommender systems in e-commerce. InProcs. of ACM conference on Electronic commerce

    1999

  10. [18]

    Jiahui Shi, Vivek Chaurasiya, Yozen Liu, Shubham Vij, Yan Wu, Satya Kanduri, Neil Shah, Peicheng Yu, Nik Srivastava, Lei Shi, et al. 2023. Embedding Based Retrieval in Friend Recommendation. InProceedings of the 46th International ACM SIGIR Conference on Research and Developme...

    2023

  11. [19]

    Xianfeng Tang, Yozen Liu, Xinran He, Suhang Wang, and Neil Shah. 2022. Friend story ranking with edge-contextual local graph convolutions. InProceedings of the Fifteenth ACM International Conference on Web Search and Data Mining. 1007–1015

    2022

  12. [20]

    Yao-Hung Hubert Tsai, Shaojie Bai, Paul Pu Liang, J Zico Kolter, Louis-Philippe Morency, and Ruslan Salakhutdinov. 2019. Multimodal transformer for unaligned multimodal language sequences. InProceedings of the conference. Association for computational linguistics. Meeting, Vol...

    2019

  13. [21]

    Aaron Van den Oord, Sander Dieleman, and Benjamin Schrauwen. 2013. Deep content-based music recommendation. InProcs. of NeurIPS

    2013

  14. [22]

    A Vaswani. 2017. Attention is all you need.Advances in Neural Information Processing Systems(2017)

    2017

  15. [23]

    Jiancan Wu, Xiang Wang, Xingyu Gao, Jiawei Chen, Hongcheng Fu, and Tianyu Qiu. 2024. On the effectiveness of sampled softmax loss for item recommendation. ACM Transactions on Information Systems42, 4 (2024), 1–26

    2024

  16. [24]

    Jiaqi Zhai, Lucy Liao, Xing Liu, Yueming Wang, Rui Li, Xuan Cao, Leon Gao, Zhao- jie Gong, Fangda Gu, Michael He, et al. 2024. Actions speak louder than words: Trillion-parameter sequential transducers for generative recommendations.arXiv preprint arXiv:2402.17152(2024)

    2024 arXiv

  17. [25]

    Feng Zhu, Yan Wang, Chaochao Chen, Jun Zhou, Longfei Li, and Guanfeng Liu

  18. [26]

    Xianghui Zhu, Mengqun Jin, Hengyu Zhang, Chang Meng, Daoxin Zhang, and Xiu Li. 2024. Modeling Domains as Distributions with Uncertainty for Cross- Domain Recommendation. InProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Ret...

    2024

  19. [2009]

    BPR: Bayesian personalized ranking from implicit feedback. InUAI

  20. [2019]

    InProceedings of the 42nd international ACM SIGIR conference on research and development in information retrieval

    𝜋-net: A parallel information-sharing network for shared-account cross- domain sequential recommendations. InProceedings of the 42nd international ACM SIGIR conference on research and development in information retrieval. 685– 694

  21. [2021]

    Cross-domain recommendation: challenges, progress, and prospects.arXiv preprint arXiv:2103.01696(2021)

    2021 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.