Pith. sign in

REVIEW 5 major objections 4 minor 1 cited by

FARM: Frequency-Aware Model for Cross-Domain Live-Streaming Recommendation

T0 review · 5 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read FARM claims that splitting a user's behavior history into low- and high-frequency components lets a live-streaming recommender perceive sparse high-value behaviors such as likes, comments, and gifts, and that this lifts cross-domain…

desk verdict Solid industrial cross-domain recommender with believable small gains, but the 'high-frequency = sparse behaviors' story is asserted, not shown. read the letter →

arxiv 2502.09375 v1 pith:4NHWZ25E submitted 2025-02-13 cs.IR

Xiaodong Li , Ruochen Yang , Shuang Wen , Shen Wang , Yueyang Liu , Guoquan Wang , Weisong Hu , Qiang Luo
show 5 more authors
Jiawei Sheng Tingwen Liu Jiangxia Cao Shuang Yang Zhaojie Liu
This is my paper · ORCID
classification cs.IR
keywords live-streamingrecommendationcross-domainDiscreteFourierTransformfrequency-awaremodelingsparseuserbehaviorsmulti-tasklearningcontrastiveattentionmechanism
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 identifies two data-sparsity problems in live-streaming recommendation: high-value behaviors such as like, comment, and gift are rare compared with clicks, and short-video exposure is about nine times larger than live-streaming exposure, so the live-streaming domain alone under-models user preference. To address this, it proposes FARM, which applies the Discrete Fourier Transform to each domain's historical behavior sequence and recombines low- and high-frequency components with learned per-domain weights, allowing the model to attend to sparse behaviors. It then aligns the short-video and live-streaming preference representations with contrastive learning before fusing them with attention. Across the two deployed apps, offline results over the PLE base model show average gains of 0.42% in AUC, 1.06% in UAUC, and 1.05% in GAUC, and online A/B tests show increases in click, like, follow, and gift count. The paper argues that these gains come from the frequency-aware module perceiving sparse yet valuable high-frequency behaviors.

What carries the argument

The central object is the frequency split of a user behavior sequence via the Discrete Fourier Transform: given a sequence embedding $V$, the spectrum $\bar{x} = Fx$ is partitioned at a cut-off $c$ into low components $Low[V]$ and high components $High[V]$, and the module returns the learned weighted sum $\alpha\,Low[V] + \beta\,High[V]$. The load-bearing identity is the unitary DFT decomposition of a sequence into Fourier basis components, together with the paper's semantic mapping that high-index basis components carry sparse valuable behaviors. This machinery converts sequence position into frequency and is what allows the model to attend to behaviors that occur rarely in the raw sequence.

What would settle it

Compare the high-frequency DFT components of behavior sequences that end in a gift or comment with those that end in a click: if the spectra do not separate, or if ablating only the high-frequency path hurts click prediction as much as gift prediction, then the asserted frequency semantics are not supported and the gains could come from added parameters instead.

Watch

Extended reading notes

Core claim

The central claim is expressed in Eq. (9): $F_{video} = \alpha \, Low[V_{video}] + \beta \, High[V_{video}]$ and $F_{live} = \gamma \, Low[V_{live}] + \delta \, High[V_{live}]$, where the $Low$ and $High$ operators reconstruct a sequence from the low-index and high-index parts of its Discrete Fourier Transform spectrum. The paper asserts that low-frequency information reflects dense behaviors such as click and long-view, while high-frequency information represents sparse valuable behaviors such as like, comment, and gift; the trainable sigmoid weights $\alpha, \beta, \gamma, \delta$ let the model upweight the high-frequency part. After reweighting, a multi-head target-attention using the candidate live-streaming author as query extracts the behavior signals and feeds the multi-task prediction head. The supporting evidence is that removing either frequency module degrades all six prediction tasks, that a case study shows a low-active user's predicted like and gift probabilities rise from about 0.4 to above 0.88 when the high-frequency weights are high, and that online A/B tests show gains in like, comment, follow, and gift count. In the paper's account, this is what it means for the model to perceive sparse user preference.

Load-bearing premise

The load-bearing premise is that the frequency index in the Discrete Fourier Transform of a user's behavior sequence corresponds to behavior rarity, so the high-frequency components of the history carry the sparse behaviors like, comment, and gift.

Editorial extensions

If this is right

  • In both deployed apps, FARM's offline AUC, UAUC, and GAUC beat the base model across all six prediction targets, with the largest relative gains on long-view, comment, and gift metrics.
  • Removing the cross-domain preference fuse module causes the largest degradation among the ablations, so the transfer of short-video preference into the live-streaming domain is load-bearing for the result.
  • The case study and the learned high-frequency weights indicate that the frequency module specifically repairs prediction of sparse like and gift behaviors for low-active users.
  • Online A/B testing shows increases in click, like, follow, and gift count, implying that frequency-aware perception of sparse behaviors converts into engagement and revenue metrics.

Reading between the lines

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

  • An implication the paper leaves implicit is that the same low/high frequency split could be applied to other multi-behavior domains, such as purchase or bookmark, where the sparse behavior is the one the platform most wants to predict.
  • Because the DFT is applied along the sequence-position axis, the 'high frequency equals rare behavior' mapping is not guaranteed by the transform itself; a direct test would compare the frequency spectra of sequences ending in a gift with those ending in a click.
  • The contrastive alignment runs only from video to live, so a bidirectional or live-to-video variant could be tested to see whether the trade-off between the two domains changes.
  • The learned frequency weights could serve as a per-user interpretability signal, flagging users whose sparse behaviors carry the most predictive weight.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 4 minor

Summary. The paper proposes FARM, a cross-domain live-streaming recommendation model for Kuaishou. FARM applies a Discrete Fourier Transform (DFT) to user behavior sequences in both short-video and live-streaming domains, splits each sequence into low- and high-frequency components, and claims that low-frequency components reflect dense behaviors (click, long-view) while high-frequency components represent sparse valuable behaviors (like, comment, gift). It then aligns user preferences across domains with a contrastive loss and fuses them with attention mechanisms before feeding a multi-task model (PLE/MMoE). Offline experiments on two Kuaishou applications compare FARM against a PLE base and four ablations; online A/B testing reports gains in interaction and core metrics. The central claim is that the frequency-aware module enables perception of sparse behaviors, leading to the reported improvements.

Significance. If the frequency-to-rarity mapping were validated, FARM would offer a novel angle on sparse-behavior modeling in industrial live-streaming recommendation. The paper's strengths include a large-scale real deployment with online A/B testing, a clear modular architecture, and ablation evidence that each module contributes to offline metrics. The main scientific claim, however, rests on an asserted rather than derived association between DFT frequency and behavior rarity, and the current experiments do not provide a direct test of that association. The empirical gains over the chosen base are small and are not compared against published state-of-the-art cross-domain or frequency-based recommenders. The significance of the work is therefore currently limited by the lack of support for its central mechanism, though the engineering contribution is credible.

major comments (5)
  1. [Section 2.1, Eq. (9)] The statement that low-frequency DFT components reflect dense behaviors and high-frequency components represent sparse behaviors is asserted without derivation, and it is in tension with standard DFT properties. A DFT coefficient's index corresponds to the rate of variation along the sequence positions, not to the rarity of a behavior type. A single isolated like or gift is a localized impulse whose spectrum is broadband, so its energy is not concentrated in the high-frequency band; conversely, alternating dense behaviors can produce high-frequency variation. This mapping is load-bearing because Eq. (9) builds the entire intra-domain frequency-aware module on it. Please provide a direct empirical test: for example, compute the distribution of high-frequency energy contributed by positions labeled as like/comment/gift versus other positions, or train the model with only the high-frequency branch and report like/comment/gift prediction metrics. Without such evidence, the stated mechanism remains a post-hoc interpretation rather than a validated design.
  2. [Tables 1-2, Section 4.1.2] The offline evaluation compares FARM only against the PLE base model. Since the paper claims superiority over prior live-streaming and cross-domain methods, the evaluation should include at least one or two representative published baselines from the related work cited by the authors, such as Moment&Cross [3], LCN [14], eLiveRec [40], or a frequency-based sequential model like FMLP-Rec [43] or BSARec [28]. Additionally, all reported numbers are point estimates with no standard deviations or significance tests. The improvements are small (e.g., +0.18% AUC for Effective-view on Kuaishou), so statistical reliability matters. Please report multiple seeds or confidence intervals and clarify whether the 'Improve' column is percentage-point difference or relative improvement.
  3. [Section 4.3 (RQ2)] The frequency-aware analysis is a case study of a single user and a single candidate author, with two predicted probabilities and two learned weights. This is anecdotal evidence. The large values of β and δ only show that the learnable gates are active; they do not show that the high-frequency band specifically encodes like/comment/gift behaviors, since a gated residual branch could equally well be amplifying useful signal for independent reasons. Please provide a population-level study: for instance, slice users by the density of their sparse behaviors and show that the high-frequency component contributes more for low-density users, or correlate the model's like/gift prediction improvements with the proportion of high-frequency energy in the corresponding behavior sequence positions.
  4. [Section 4.4, Eq. (14)] Figure 3 reports that FARM has higher cosine similarity between h_video and h_live than the w/o C-PA variant, and this is presented as evidence that the preference align module works. However, Eq. (14) directly optimizes exactly this cosine similarity through the contrastive loss, so the observed increase is partly by construction. This analysis is self-confirmatory and should be framed as a sanity check, not as independent evidence of effectiveness. The stronger evidence is the degradation of w/o C-PA in Tables 1-2; please rely on that and, if possible, add a downstream metric such as online click or gift rate for the align module.
  5. [Section 4.2, Tables 1-2] The offline results are reported to three or four decimal places with no variance, despite the statement that 'we report the average performance of all models on testing set over hours.' If this averaging is over hourly test sets, the standard deviation across hours should be available and should be reported. Given the very small absolute improvements (e.g., +0.18% AUC for Effective-view on Kuaishou), the lack of uncertainty makes it impossible to assess whether the differences are meaningful. Please add standard deviations or statistical tests for the main comparisons in Tables 1 and 2.
minor comments (4)
  1. [Section 3.2, Eq. (10)] The text says 'γ and δ control the strength of high-pass components,' but γ already denotes the live-domain low-pass weight in Eq. (9); this should be β and δ.
  2. [Abstract and Section 3.3] The phrase 'a serious of tailor-designed attention mechanisms' appears multiple times; it should be 'a series of'.
  3. [Section 2.1, Eq. (1)] The definition of the DFT matrix and basis vectors is underspecified: the matrix F is defined via rows f_j, but the dimension N and the indexing convention are not stated precisely, and writing f_j in R^N is inconsistent with complex exponentials. Please clarify with a standard definition.
  4. [Section 4.1.1] The data-collection description is terse. The phrase 'first-only strategy [3]' and the 30s sliding window need a brief explanation (e.g., what is a 'segment sample' and how negatives are defined) for reproducibility.

Circularity Check

2 steps flagged · score 6.0 of 10

Two internal steps reduce to their own inputs: the §2.1 stipulation that high-frequency information 'represents' sparse behaviors makes Eq. (9)'s mechanism claim definitional, and §4.4's alignment evidence reports the cosine similarity that Eq. (14) directly optimizes. The held-out AUC/GAUC and online A/B gains remain non-circular.

  1. self definitional [Section 2.1 (Fourier Transform) and Section 3.2, Eq. (9)]
    "In our FARM, the low-frequency information reflects user's dense behaviors (e.g., click and long-view), while the high-frequency information represents user's sparse behaviors (e.g., like, comment and gift). ... we introduce the intra-domain frequency-aware module to enable our model to perceive user's sparse yet valuable behaviors (i.e., high-frequency information), supported by Discrete Fourier Transform (DFT)."

    The mapping is stipulated, not derived. A DFT index over a user's ordered sequence indexes how rapidly embeddings change along the sequence, not how rare a behavior type is. Because 'sparse behaviors' are equated with 'high-frequency information' by definition, the module's use of High[V] in Eq. (9) makes 'perceiving sparse behaviors' true by construction. The Table 3 case study then only shows that the learned weights beta and delta are large; it cannot independently confirm that the high-frequency band carries like/comment/gift. The ablation shows the branch helps, but any semantic interpretation of the branch is equally consistent with that result.

  2. fitted input called prediction [Section 3.3.1 Eq. (14) and Section 4.4 (Preference Align Analysis, Figure 3)]
    "we perform contrastive loss function to enforce the model to narrow the gap between user preference representations for positive pairs within a given batch... we randomly sample 10000 users and calculate their cosine similarity (ranging from -1 to 1) of the preference representation h_video in the short-video domain and h_live in the live-streaming domain..."

    Eq. (14) is exactly a contrastive loss over sim(h_live_i, h_video_i) for positive pairs within the batch, so maximizing it is the training objective of the align module. Figure 3's observation that FARM, which is trained with this loss, has higher cosine similarity than w/o C-PA, which is not, is guaranteed by construction. It confirms only that the optimizer minimized the loss, not that user preferences are aligned in any independently measured sense. The held-out AUC/GAUC improvements do not isolate this alignment mechanism.

full rationale

FARM's headline offline and online results are measured against held-out user logs and A/B metrics, so the main performance claims are not circular, and no load-bearing result depends on a self-citation chain; the DFT references are external. Two internal steps do reduce to their own inputs. First, the paper's mechanism claim that the module 'perceives user's sparse yet valuable behaviors' is made true by stipulation: Section 2.1 declares high-frequency information to represent sparse behaviors, and Section 3.2 restates sparse behaviors as 'i.e., high-frequency information'. Since DFT frequency over sequence positions is not the same property as behavior-type rarity, this identification is an assumption, and Eq. (9)'s weighted high-pass branch cannot by itself demonstrate perception of like/comment/gift. Second, the RQ3 verification in Section 4.4 measures the very cosine similarity that the contrastive loss in Eq. (14) directly optimizes, so higher similarity in FARM than in w/o C-PA is forced by the training objective. The ablations and case study do not independently test the frequency-band semantics: they show the branch contributes and that its learned weights are large, which is consistent with a learned residual branch. These issues make parts of the argument self-confirmatory, but they do not undermine the held-out empirical gains.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The model's central mechanism rests on the assumed semantic mapping between DFT frequency components and behavior rarity, plus standard assumptions about unbiased labels and cross-domain preference similarity. Hyperparameters c and lambda are fitted and not reported in final form.

free parameters (2)
  • cut-off frequency c = not reported; searched over {1, 3, 5, 7, 9}
    Determines the boundary between low- and high-frequency components; the central claim about high-frequency sparse behaviors depends on this choice, and the final value is not stated.
  • contrastive loss weight lambda = not reported; searched over {0.1, 0.3, 0.5, 0.7, 0.9}
    Controls the strength of the cross-domain alignment loss; the trade-off between alignment and task loss is tuned but the selected value is not disclosed.
assumptions (5)
  • ad hoc to paper High-frequency DFT components of the behavior-sequence embeddings correspond to sparse valuable behaviors (like, comment, gift).
    No derivation given; used in Sections 2.1 and 3.2 to motivate Eq. (9). A DFT frequency index over sequence positions does not by itself encode behavior rarity.
  • standard math The DFT of a real-valued sequence can be split into low- and high-frequency components that are real-valued and reconstruct the sequence as in Eq. (9).
    Section 2.1 defines Low[x] and High[x] using only the first c (or last N-c) complex Fourier basis vectors; without conjugate symmetry these outputs are not guaranteed to be real, yet they are used as real vectors.
  • domain assumption User preferences are similar across the short-video and live-streaming domains for the same user, justifying positive pairs in Eq. (14).
    Stated as intuition in Section 3.3.1; required for the contrastive loss to be meaningful.
  • domain assumption The first-only labeling strategy and exit-based negatives give unbiased supervision for all six targets.
    Described in Section 4.1.1; any bias in the labels would propagate to all reported gains.
  • domain assumption PLE/MMoE is an adequate base model for comparing multi-task recommendation quality.
    Chosen in Section 4.1.2 as the sole comparison base; the absolute gains are relative to this choice and may not hold against stronger published baselines.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FARM: Frequency-Aware Model for Cross-Domain Live-Streaming Recommendation." pith.science (2026). https://pith.science/paper/4NHWZ25E

@misc{pith2026250209375,
  author       = {Pith},
  title        = {Pith review of: FARM: Frequency-Aware Model for Cross-Domain Live-Streaming Recommendation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4NHWZ25E}},
  note         = {Machine review of arXiv:2502.09375}
}
read the original abstract

Live-streaming services have attracted widespread popularity due to their real-time interactivity and entertainment value. Users can engage with live-streaming authors by participating in live chats, posting likes, or sending virtual gifts to convey their preferences and support. However, the live-streaming services faces serious data-sparsity problem, which can be attributed to the following two points: (1) User's valuable behaviors are usually sparse, e.g., like, comment and gift, which are easily overlooked by the model, making it difficult to describe user's personalized preference. (2) The main exposure content on our platform is short-video, which is 9 times higher than the exposed live-streaming, leading to the inability of live-streaming content to fully model user preference. To this end, we propose a Frequency-Aware Model for Cross-Domain Live-Streaming Recommendation, termed as FARM. Specifically, we first present the intra-domain frequency aware module to enable our model to perceive user's sparse yet valuable behaviors, i.e., high-frequency information, supported by the Discrete Fourier Transform (DFT). To transfer user preference across the short-video and live-streaming domains, we propose a novel preference align before fuse strategy, which consists of two parts: the cross-domain preference align module to align user preference in both domains with contrastive learning, and the cross-domain preference fuse module to further fuse user preference in both domains using a serious of tailor-designed attention mechanisms. Extensive offline experiments and online A/B testing on Kuaishou live-streaming services demonstrate the effectiveness and superiority of FARM. Our FARM has been deployed in online live-streaming services and currently serves hundreds of millions of users on Kuaishou.

Figures

Figures reproduced from arXiv: 2502.09375 by the authors.

Figure 1
Figure 1. An illustration of (a) relative percentages of dif [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall framework of FARM. The intra-domain frequency-aware module enable the model to perceive user’s [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Visualization of the fused user preference. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figures from the paper (1 more)
Figure 3
Figure 3. Figure 3: A statistical result of the cosine similarity with respect to (a) w/o C-PA variant, and (b) FARM. make accurate prediction for these sparse yet valuable behaviors (i.e., 𝑦ˆ 𝑙𝑡𝑟 𝑡 = 0.41, 𝑦ˆ 𝑔𝑡𝑟 𝑡 = 0.36). We can also draw similar observa￾tions from the perspective of t…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Towards Generalizable Safety in Crowd Navigation via Conformal Uncertainty Handling

    cs.RO 2025-08 unverdicted novelty 5.0 of 10

    A crowd navigation method augmenting reinforcement learning with conformal uncertainty estimates is claimed to cut collisions under distribution shift, but the manuscript body is an unrelated live streaming dataset paper.

Reference graph

Works this paper leans on

45 extracted references · 24 canonical work pages · cited by 1 Pith paper

  1. [3]

    Jiangxia Cao, Shen Wang, Yue Li, Shenghui Wang, Jian Tang, Shiyao Wang, Shuang Yang, Zhaojie Liu, and Guorui Zhou. 2024. Moment&Cross: Next- Generation Real-Time Cross-Domain CTR Prediction for Live-Streaming Recom- mendation at Kuaishou. arXiv preprint arXiv:2408.05709 (2024)

    arXiv 2024

  2. [14]

    Ruijie Hou, Zhaoyang Yang, Yu Ming, Hongyu Lu, Zhuobin Zheng, Yu Chen, Qinsong Zeng, and Ming Chen. 2024. Cross-Domain LifeLong Sequential Model- ing for Online Click-Through Rate Prediction. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 5116–5125

    2024

  3. [40]

    Yixin Zhang, Yong Liu, Hao Xiong, Yi Liu, Fuqiang Yu, Wei He, Yonghui Xu, Lizhen Cui, and Chunyan Miao. 2023. Cross-domain disentangled learning for e-commerce live streaming recommendation. In 2023 IEEE 39th International Conference on Data Engineering (ICDE) . IEEE, 2955–2968

    2023

  4. [43]

    Kun Zhou, Hui Yu, Wayne Xin Zhao, and Ji-Rong Wen. 2022. Filter-enhanced MLP is all you need for sequential recommendation. In Proceedings of the ACM web conference 2022. 2388–2399

    work page 2022

  5. [28]

    Yehjin Shin, Jeongwhan Choi, Hyowon Wi, and Noseong Park. 2024. An atten- tive inductive bias for sequential recommendation beyond the self-attention. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 38. 8984–8992

    work page 2024

  6. [1]

    Zhicheng An, Zhexu Gu, Li Yu, Ke Tu, Zhengwei Wu, Binbin Hu, Zhiqiang Zhang, Lihong Gu, and Jinjie Gu. 2024. DDCDR: A Disentangle-based Distillation Framework for Cross-Domain Recommendation. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 4764–4773

    2024

  7. [2]

    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)

    work page 2022

  8. [4]

    Yue Cao, Xiaojiang Zhou, Jiaqi Feng, Peihao Huang, Yao Xiao, Dayao Chen, and Sheng Chen. 2022. Sampling is all you need on modeling long-term user behaviors for CTR prediction. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management . 2974–2983

    2022

Show all 45 references
  1. [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 Conferen...

    2023

  2. [6]

    Jianxin Chang, Chenbin Zhang, Yiqun Hui, Dewei Leng, Yanan Niu, Yang Song, and Kun Gai. 2023. Pepnet: Parameter and embedding personalized network for infusing with personalized prior information. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data...

    2023

  3. [7]

    Gaode Chen, Ruina Sun, Yuezihan Jiang, Jiangxia Cao, Qi Zhang, Jingjian Lin, Han Li, Kun Gai, and Xinghua Zhang. 2024. A Multi-modal Modeling Framework for Cold-start Short-video Recommendation. In Proceedings of the 18th ACM Conference on Recommender Systems . 391–400

    2024

  4. [8]

    Qiwei Chen, Changhua Pei, Shanshan Lv, Chao Li, Junfeng Ge, and Wenwu Ou

  5. [9]

    Xu Chen, Zida Cheng, Jiangchao Yao, Chen Ju, Weilin Huang, Jinsong Lan, Xiaoyi Zeng, and Shuai Xiao. 2024. Enhancing cross-domain click-through rate prediction via explicit feature augmentation. In Companion Proceedings of the ACM on Web Conference 2024. 423–432

    2024

  6. [10]

    Jiaxin Deng, Dong Shen, Shiyao Wang, Xiangyu Wu, Fan Yang, Guorui Zhou, and Gaofeng Meng. 2023. ContentCTR: Frame-level Live Streaming Click-Through Rate Prediction with Multimodal Transformer. arXiv preprint arXiv:2306.14392 (2023)

    2023 arXiv

  7. [11]

    Jiaxin Deng, Shiyao Wang, Dong Shen, Liqin Zhao, Fan Yang, Guorui Zhou, and Gaofeng Meng. 2024. A Multimodal Transformer for Live Streaming Highlight Prediction. In 2024 IEEE International Conference on Multimedia and Expo (ICME)

    2024

  8. [12]

    Jiaxin Deng, Shiyao Wang, Yuchen Wang, Jiansong Qi, Liqin Zhao, Guorui Zhou, and Gaofeng Meng. 2024. MMBee: Live Streaming Gift-Sending Recommenda- tions via Multi-Modal Fusion and Behaviour Expansion. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and ...

    2024

  9. [13]

    Yongqiang Han, Hao Wang, Kefan Wang, Likang Wu, Zhi Li, Wei Guo, Yong Liu, Defu Lian, and Enhong Chen. 2024. END4Rec: Efficient Noise-Decoupling for Multi-Behavior Sequential Recommendation. arXiv preprint arXiv:2403.17603 (2024)

    2024 arXiv

  10. [15]

    Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic opti- mization. arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  11. [16]

    Chenyi Lei, Yong Liu, Lingzi Zhang, Guoxin Wang, Haihong Tang, Houqiang Li, and Chunyan Miao. 2021. Semi: A sequential multi-modal information transfer network for e-commerce micro-video recommendations. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery &...

    2021

  12. [17]

    Danwei Li, Zhengyu Zhang, Siyang Yuan, Mingze Gao, Weilin Zhang, Chaofei Yang, Xi Liu, and Jiyan Yang. 2023. AdaTT: Adaptive Task-to-Task Fusion Network for Multitask Learning in Recommendations. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data ...

    2023

  13. [18]

    Junnan Li, Ramprasaath Selvaraju, Akhilesh Gotmare, Shafiq Joty, Caiming Xiong, and Steven Chu Hong Hoi. 2021. Align before fuse: Vision and language repre- sentation learning with momentum distillation. Advances in neural information processing systems 34 (2021), 9694–9705

    2021

  14. [19]

    Xiaodong Li, Jiawei Sheng, Jiangxia Cao, Wenyuan Zhang, Quangang Li, and Tingwen Liu. 2024. CDRNP: Cross-Domain Recommendation to Cold-Start Users via Neural Process. In Proceedings of the 17th ACM International Conference on Web Search and Data Mining. 378–386

    2024

  15. [20]

    Xiaodong Li, Hengzhu Tang, Jiawei Sheng, Xinghua Zhang, Li Gao, Suqi Cheng, Dawei Yin, and Tingwen Liu. 2025. Exploring Preference-Guided Diffusion Model for Cross-Domain Recommendation. arXiv preprint arXiv:2501.11671 (2025)

    2025 arXiv

  16. [21]

    Jiahao Liu, Dongsheng Li, Hansu Gu, Tun Lu, Peng Zhang, Li Shang, and Ning Gu. 2023. Personalized graph signal processing for collaborative filtering. In Proceedings of the ACM Web Conference 2023 . 1264–1272

    2023

  17. [22]

    Xinchen Luo, Jiangxia Cao, Tianyu Sun, Jinkai Yu, Rui Huang, Wei Yuan, Hezheng Lin, Yichen Zheng, Shiyao Wang, Qigen Hu, et al . 2024. QARM: Quantita- tive Alignment Multi-Modal Recommendation at Kuaishou. arXiv preprint arXiv:2411.11739 (2024)

    2024 arXiv

  18. [23]

    Jiaqi Ma, Zhe Zhao, Xinyang Yi, Jilin Chen, Lichan Hong, and Ed H Chi. 2018. Modeling task relationships in multi-task learning with multi-gate mixture-of- experts. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining . 1930–1939

    2018

  19. [24]

    Gonzalo Mateos, Santiago Segarra, Antonio G Marques, and Alejandro Ribeiro

  20. [25]

    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

  21. [26]

    Jérémie Rappaz, Julian McAuley, and Karl Aberer. 2021. Recommendation on live-streaming platforms: Dynamic availability and repeat consumption. In Pro- ceedings of the 15th ACM Conference on Recommender Systems . 390–399

    2021

  22. [27]

    Yifei Shen, Yongji Wu, Yao Zhang, Caihua Shan, Jun Zhang, B Khaled Letaief, and Dongsheng Li. 2021. How powerful is graph convolution for recommendation?. In Proceedings of the 30th ACM international conference on information & knowledge management. 1619–1629

    2021

  23. [29]

    Guoqiang Sun, Yibin Shen, Sijin Zhou, Xiang Chen, Hongyan Liu, Chunming Wu, Chenyi Lei, Xianhui Wei, and Fei Fang. 2023. Self-supervised interest transfer network via prototypical contrastive learning for recommendation. InProceedings of the AAAI Conference on Artificial Intel...

    2023

  24. [30]

    Hongyan Tang, Junning Liu, Ming Zhao, and Xudong Gong. 2020. Progressive layered extraction (ple): A novel multi-task learning (mtl) model for personalized recommendations. In Proceedings of the 14th ACM Conference on Recommender Systems. 269–278

    2020

  25. [31]

    Shisong Tang, Qing Li, Dingmin Wang, Ci Gao, Wentao Xiao, Dan Zhao, Yong Jiang, Qian Ma, and Aoyang Zhang. 2023. Counterfactual video recommendation for duration debiasing. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 4894–4903

    2023

  26. [32]

    Charles Van Loan. 1992. Computational frameworks for the fast Fourier transform . SIAM

    1992

  27. [33]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems 30 (2017)

    2017

  28. [34]

    Chenyang Wang, Weizhi Ma, Min Zhang, Chong Chen, Yiqun Liu, and Shaoping Ma. 2020. Toward dynamic user intention: Temporal evolutionary effects of item relations in sequential recommendation. ACM Transactions on Information Systems (TOIS) 39, 2 (2020), 1–33

    2020

  29. [35]

    Ruize Wang, Hui Xu, Ying Cheng, Qi He, Xing Zhou, Rui Feng, Wei Xu, Lei Huang, and Jie Jiang. 2024. ADSNet: Cross-Domain LTV Prediction with an Adaptive Siamese Network in Advertising. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 5872–5881

    2024

  30. [36]

    Xiyuan Wang and Muhan Zhang. 2022. How powerful are spectral graph neural networks. In International conference on machine learning . PMLR, 23341–23362

    2022

  31. [37]

    Jiafeng Xia, Dongsheng Li, Hansu Gu, Tun Lu, Peng Zhang, Li Shang, and Ning Gu. 2024. Hierarchical Graph Signal Processing for Collaborative Filtering. In Proceedings of the ACM on Web Conference 2024 . 3229–3240

    2024

  32. [38]

    Kun Yi, Qi Zhang, Wei Fan, Hui He, Liang Hu, Pengyang Wang, Ning An, Long- bing Cao, and Zhendong Niu. 2024. FourierGNN: Rethinking multivariate time series forecasting from a pure graph perspective. Advances in Neural Information Processing Systems 36 (2024)

    2024

  33. [39]

    Shuai Zhang, Hongyan Liu, Jun He, Sanpu Han, and Xiaoyong Du. 2021. Deep sequential model for anchor recommendation on live streaming platforms. Big Data Mining and Analytics 4, 3 (2021), 173–182

    2021

  34. [41]

    Yongfeng Zhang, Min Zhang, Yi Zhang, Guokun Lai, Yiqun Liu, Honghui Zhang, and Shaoping Ma. 2015. Daily-aware personalized recommendation based on feature-level time series analysis. In Proceedings of the 24th international confer- ence on world wide web . 1373–1383

    2015

  35. [42]

    Guorui Zhou, Na Mou, Ying Fan, Qi Pi, Weijie Bian, Chang Zhou, Xiaoqiang Zhu, and Kun Gai. 2019. Deep interest evolution network for click-through rate prediction. In Proceedings of the AAAI conference on artificial intelligence

    2019

  36. [2019]

    IEEE Signal Processing Magazine 36, 3 (2019), 16–43

    Connecting the dots: Identifying network structure via graph signal pro- cessing. IEEE Signal Processing Magazine 36, 3 (2019), 16–43

    2019

  37. [2021]

    arXiv preprint arXiv:2108.04468 (2021)

    End-to-end user behavior retrieval in click-through rateprediction model. arXiv preprint arXiv:2108.04468 (2021)

    2021 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.