Pith. sign in

REVIEW 4 major objections 6 minor 50 references

Adaptive$^2$: Adaptive Domain Mining for Fine-grained Domain Adaptation Modeling

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read Latent domains mined self-supervised by a VQ-VAE beat hand-crafted domain splits in advertising click-through-rate models, with a 2.3 percent revenue lift online.

desk verdict VQ-VAE domain mining is a promising industrial idea, but the missing randomized-routing control leaves the central claim unproven. read the letter →

arxiv 2412.08198 v2 pith:PBABB7JQ submitted 2024-12-11 cs.LG

classification cs.LG
keywords domainminingadaptationself-supervisedlearningVQ-VAECTRpredictiononlineadvertisingmulti-domainrecommendationsharedandspecificnetwork
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 argues that the biggest bottleneck in multi-domain ad ranking is not the adaptation network but the definition of domains themselves: hand-crafted splits such as ad placement miss fine-grained distribution shifts that matter for click prediction. It proposes Adaptive2, which mines latent domains automatically with a vector-quantized variational autoencoder (VQ-VAE) and routes each sample to a domain-specific network alongside a shared network. On Avazu and Criteo it reports the best AUC among MLP, MMoE, PLE, ADIN, and PEPNet baselines, and shows that under comparable FLOPs the hand-crafted-domain baselines do not beat a single-domain MLP. When deployed in a live-streaming pre-ranking system, the method reports a 2.3% revenue increase and a 2.5% conversion increase. The point, if true, is that domain identity should be learned from the data rather than assumed from business metadata.

What carries the argument

The load-bearing mechanism is the VQ-VAE domain mining module: a vector-quantized variational autoencoder whose codebook acts as a learnable set of latent domains. An encoder projects the concatenated user–item embedding into a continuous latent vector $z_e$, which is mapped to its nearest codebook entry $e_k$ by Euclidean distance; the index $k$ becomes the domain routing signal. The reconstruction objective is trained jointly with the downstream CTR loss, with a stop-gradient on the input so the mining module does not destabilize the feature embeddings. The authors' core assumption is that this self-supervised clustering yields codebooks that correspond to data distributions that are distinct across codes and similar within a code, making the mined index a meaningful domain label. The module plugs into different domain-adaptation backbones (the paper shows it improving both its own shared/specific network and PEPNet), which is why the authors describe it as a general tool for domain identification.

What would settle it

Permutation falsifier: in the trained Adaptive2 on Avazu, keep all weights fixed but route each sample to a randomly permuted domain-specific network (same networks, random assignment of which sample goes to which). If AUC does not fall substantially from the reported 0.7850 toward the MLP baseline of 0.7759, the mined domain index is not actually carrying task-relevant routing information, and the gains would instead come from added capacity or the shared network.

Watch

Extended reading notes

Core claim

Adaptive2 treats domain identification and domain adaptation as two separable tasks. A self-supervised VQ-VAE encoder maps each sample's feature embedding to one of $m$ codebook vectors; the chosen codebook index is the sample's latent domain. That index routes the sample through one of $m$ domain-specific networks, whose output is added to the output of a shared network before the click prediction head. The authors claim that this automatically mined domain structure captures fine-grained distributional differences that manually-selected domain features miss, and that this is why Adaptive2 outperforms baselines both offline (Avazu AUC 0.7850 vs 0.7823 for the best baseline ADIN; Criteo AUC 0.8097 vs 0.8068) and in online deployment. They further claim that previous domain-adaptation methods with hand-crafted domains perform no better than a single-domain MLP when compared at matched FLOPs, which they take as evidence that domain definition itself is the decisive factor.

Load-bearing premise

The whole mechanism rests on the premise that reconstruction-based clustering by the VQ-VAE produces latent domains that correspond to task-relevant distribution shifts; if the codebook groups data by features that have nothing to do with click behavior, routing by the mined domain index would not help beyond random assignment.

Editorial extensions

If this is right

  • Domain definition should be treated as a first-class modeling problem in multi-domain CTR prediction, not a fixed input: hand-crafted domains under matched compute do not beat a single-domain MLP.
  • Self-supervised clustering with vector quantization can serve as a domain-identification module that transfers across network architectures (ADMM and PEPNet both improve when given the mined domain ids).
  • Because the domain signal is learned end-to-end, no expert annotation of domains is needed at training or serving time, which simplifies deployment.
  • Under matched FLOPs and parameter counts, the mined-domain model is more resource-efficient than the larger hand-crafted-domain baselines, which matters for industrial ROI.
  • Online, the mined domains give a 2.3% revenue lift and a 2.5% conversion lift over the previous online SOTA in the live-stream pre-ranking stage.
  • The same VQ-VAE routing idea transfers to other multi-domain ranking problems, such as e-commerce or short-video feeds, where business-defined scene IDs are too coarse; the paper only tests ads, but nothing in the mechanism is ad-specific.
  • The fair-FLOPs result implies that some published multi-domain gains may be artifacts of unequal compute budgets; re-benchmarking earlier methods at matched FLOPs could reorder the leaderboard.
  • A testable refinement is to make the number of latent domains adaptive, growing or pruning codebook entries, rather than a tuned hyperparameter, since the paper leaves $m$ fixed per dataset and acknowledges collapse risk.
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

4 major / 6 minor

Summary. Adaptive^2 proposes to replace hand-crafted domain definitions in multi-domain CTR prediction with an automatically mined partition, using a VQ-VAE trained by reconstruction as a domain mining module and a shared-and-specific network that routes samples to domain-specific layers based on the mined domain index. The paper reports offline improvements over MLP, MMoE, PLE, ADIN, and PEPNet on Avazu and Criteo, shows that several baselines underperform a tuned MLP under approximately matched FLOPs, and reports an online deployment at Kuaishou with +2.376% revenue and +2.522% conversions. The central conceptual claim is that self-supervised reconstruction clustering discovers latent domain structure that hand-crafted domain features miss, and that this structure is responsible for the gains.

Significance. If the central claim were established, the paper would make a useful practical contribution by showing that domain definitions for multi-domain advertising models can be learned automatically rather than hand-crafted, and its computational-fairness comparison is a valuable reminder that model comparisons in this area should control for resource costs. The paper is also refreshingly explicit about its limitations in Appendix D, conceding that the VQ-VAE choice is empirical and lacks theoretical support. However, the load-bearing premise that reconstruction-based VQ-VAE routing identifies task-relevant domains is not demonstrated; the evidence is thin, with small AUC gaps and no reported uncertainty for the main results or the online deployment. The contribution is therefore promising but needs substantially stronger evidence before the claims can be accepted.

major comments (4)
  1. [§4.1, Eq. (10)] The load-bearing premise that VQ-VAE self-supervision 'ensures that data points assigned to different codebooks come from distinct data distributions' is not entailed by the objective. In Eq. (10), the domain-mining loss is purely a reconstruction loss on z = FFN(concat(x_u|x_v)), and Section 4.3 states that a stop gradient is applied to the VQ-VAE input, so no gradient from the task loss L_task (Eq. 17) flows into the encoder, decoder, or codebook selection in Eq. (8). The cluster assignment k is thus optimized to reconstruct the raw feature mix, not to separate click/conversion distributions. The ablation in Table 3 compares mined domains against two hand-picked features but never includes a randomized-routing control (e.g., random codebook assignment, or routing by a fixed random partition with the same architecture). Without such a control, the reported gains could be attributed to the extra capacity of the domain-specific parameters or to more flexible routing rather than to mined domain structure. Please add a random-routing control, and ideally a variant where the routing is trained with a task-gradient path (e.g., Gumbel-softmax), to test the mechanism.
  2. [Tables 2 and 4] The main offline results show AUC improvements over the best baseline of about 0.002-0.003 (Avazu: 0.7850 vs 0.7823 for ADIN; Criteo: 0.8097 vs 0.8068 for PLE). The manuscript gives no error bars, no number of seeds, and no significance test for these differences. In Table 4 the italicized items are claimed to be statistically significant over MLP with P-value<0.01, but the testing procedure is not described (number of runs, paired or unpaired, which model's variance). Please report mean and standard deviation over multiple independent training runs (at least 3-5 seeds) and a proper significance test, for both the main results and the FLOPs-fair comparison. Otherwise the headline improvements are within the range of typical run-to-run noise for CTR models.
  3. [Table 5 (online)] The online experiment reports single raw deltas of +2.376% revenue and +2.522% conversions over 'Previous SOTA', with no confidence intervals, no number of experimental units, and no statistical test. The comparison against ADMM+HD(PageID) (-0.112% revenue, -0.256% conversions) is likewise presented without uncertainty. Given that the experiment ran for 15 days on 10% traffic, daily or bucket-level variance is likely material. Please provide uncertainty estimates (e.g., bootstrap confidence intervals over days or traffic splits) and state explicitly whether the positive and negative deltas are statistically distinguishable from zero and from each other.
  4. [§5.1.4 and Appendix B] The number of latent domains (codebook size m) is selected by increasing it until validation performance stops improving (§5.1.4). This means the clusters are tuned on the downstream objective, and the subsequent interpretation of those clusters as 'domains' with real-world meaning (Appendix B) is therefore circular to some degree: they are by construction clusters that improve validation AUC, not clusters validated against an external notion of domain. The paper should report sensitivity of results to the codebook size m, and, if possible, compare the mined partition against an external domain signal (e.g., ad placement, advertiser sector) to show whether the mined domains correspond to interpretable business segments.
minor comments (6)
  1. [Appendix C] The text refers to 'MES' when discussing the mean-square-error codebook assignment; should be 'MSE' for consistency.
  2. [§4.1] The sentence 'This process ensures that data points assigned to different codebooks come from distinct data distributions' is too strong as a description of what a reconstruction loss guarantees; rephrase to 'aims to encourage' or 'is intended to separate'.
  3. [Figures] Figure 3 and Figures 4-5 are referenced in the text but do not appear in the manuscript body; ensure that all figures are actually included in the submitted version.
  4. [Introduction / Related Work] The claim to be 'the first approach to automatically learn both domain identification and adaptation in online advertising' should be qualified in light of existing work on automatic domain division (e.g., ref [16], D3) and self-supervised clustering methods that have been applied to recommendation; please clarify the precise novelty claim.
  5. [§5.5] The online experiment uses 'ARF loss' but the acronym is never defined in the text; please define it or cite it more prominently.
  6. [Appendix B] For the t-SNE visualization, the citation should refer to van der Maaten and Hinton (2008), not to ref [22] which is about autoregressive image generation.

Circularity Check

1 steps flagged · score 2.0 of 10

Only the appendix t-SNE visualization is circular; the central empirical claims rest on external benchmarks and are not equivalent to the method's inputs.

  1. self definitional [Appendix B, Domain Visualization (Figures 4 and 5)]
    "All nodes in Fig. 4 and Fig. 5 are colored by the domain id. It can be observed that initially, the nodes are distributed throughout the entire space, with data from different domains interwoven. After mining, data within the same domain becomes closely clustered, and significant boundaries emerge between different domains. This further demonstrates that we have mined the implicit domain information present in the data."

    The 'domain id' is defined by Eq. (8) as k = arg min_j ||z_e - e_j||_2, and the 'after mining' plot shows z_e = Encoder(z). Coloring z_e by this argmin guarantees that points with the same domain id are closer to the same codebook vector than to any other codebook vector; the VQ-VAE loss in Eq. (10) also explicitly trains z_e toward the assigned codebook. The observed clustering and boundaries are therefore entailed by the construction and the training objective, so the figure cannot independently demonstrate that 'implicit domain information' has been mined. It restates the cluster assignment rather than validating it against an external notion of domain.

full rationale

The paper's central derivation is not circular. Adaptive2's domain mining module is trained by the VQ-VAE reconstruction loss in Eq. (10) with stop-gradient, and the resulting discrete codebook index is used to route inputs to domain-specific networks. The main evidence for the method is external: offline AUC/LogLoss comparisons on Avazu and Criteo, a FLOPs-fair comparison, ablations against hand-crafted domain features, and an online deployment with revenue and conversion measurements. These benchmarks do not reduce by construction to the method's own definitions. The statement in Section 4.1 that VQ-VAE 'ensures that data points assigned to different codebooks come from distinct data distributions' is an unsupported empirical claim rather than a logical circularity, and the paper itself concedes in Appendix D that the choice of VQ-VAE is empirical and lacks sufficient theoretical support. The one genuinely circular piece of evidence is the appendix t-SNE visualization, where points are colored by the same codebook assignment that is being exhibited as clustering, making the observed separation a tautology. The codebook size is tuned on validation performance, which introduces mild self-referentiality, but the cluster assignments themselves are not trained by the task loss, so this does not make the central claim circular. Overall, the main results are self-contained against external benchmarks; the circularity is confined to a supporting visualization, hence the low score.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim is empirical: VQ-VAE's reconstruction objective is assumed to yield task-relevant domain structure, and the shared/specific architecture is inherited from prior work. No mathematical derivation is attempted, so the ledger records the key empirical assumptions and tuned hyperparameters.

free parameters (3)
  • number of latent domains (codebook size m) = not reported
    Determines the granularity of domain mining and the number of domain-specific networks; selected on validation performance (Section 5.1.4).
  • VQ-VAE commitment cost beta = not reported
    Balances reconstruction and quantization in Eq. (10); no value or tuning range is given.
  • hand-crafted domain features d1 and d2 for ablation = unnamed features
    Used in Table 3 ablation; the authors say they are 'carefully selected features suitable for domain info', which is a hand-chosen choice that affects the comparison.
assumptions (4)
  • domain assumption Reconstruction-based self-supervised clustering yields domains that correspond to meaningful data-distribution shifts.
    Section 4.1 states VQ-VAE 'ensures that data points assigned to different codebooks come from distinct data distributions'; the authors note in Limitations (Appendix D) this is an empirical choice lacking theoretical support.
  • domain assumption The shared and domain-specific architecture from prior work (ADIN/STAR) is a sound base for modeling commonalities and diversities.
    Section 4.2 inherits the architecture from [18,31] without re-deriving it.
  • domain assumption The public datasets, preprocessing from AFN [8], and the evaluation protocol are valid proxies for real advertising multi-domain performance.
    Section 5.1.1 relies on external preprocessing and standard splits.
  • domain assumption Online revenue and conversion deltas over 15 days on 10% traffic are attributable to the model change.
    Section 5.5 reports results without confidence intervals or discussion of confounders.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Adaptive$^2$: Adaptive Domain Mining for Fine-grained Domain Adaptation Modeling." pith.science (2026). https://pith.science/paper/PBABB7JQ

@misc{pith2026241208198,
  author       = {Pith},
  title        = {Pith review of: Adaptive$^2$: Adaptive Domain Mining for Fine-grained Domain Adaptation Modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PBABB7JQ}},
  note         = {Machine review of arXiv:2412.08198}
}
abstract

Advertising systems often face the multi-domain challenge, where data distributions vary significantly across scenarios. Existing domain adaptation methods primarily focus on building domain-adaptive neural networks but often rely on hand-crafted domain information, e.g., advertising placement, which may be sub-optimal. We think that fine-grained "domain" patterns exist that are difficult to hand-craft in online advertisement. Thus, we propose Adaptive$^2$, a novel framework that first learns domains adaptively using a domain mining module by self-supervision and then employs a shared&specific network to model shared and conflicting information. As a practice, we use VQ-VAE as the domain mining module and conduct extensive experiments on public benchmarks. Results show that traditional domain adaptation methods with hand-crafted domains perform no better than single-domain models under fair FLOPS conditions, highlighting the importance of domain definition. In contrast, Adaptive$^2$ outperforms existing approaches, emphasizing the effectiveness of our method and the significance of domain mining. We also deployed Adaptive$^2$ in the live streaming scenario of Kuaishou Advertising System, demonstrating its commercial value and potential for automatic domain identification. To the best of our knowledge, Adaptive$^2$ is the first approach to automatically learn both domain identification and adaptation in online advertising, opening new research directions for this area.

Figures

Figures reproduced from arXiv: 2412.08198 by the authors.

Figure 1
Figure 1. The key tasks of solving multi-domain problems [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The Adaptive2 framework, consists of the domain mining and the adaptive domain modeling module. The domain mining module outputs the domain index of each sample, which is used for routing to the domain-specific networks in the adaptive domain modeling module. Specifically, we employ VQVAE as the domain mining module. domain to enhance the performance of the target domain with limited labeled data. In domain adaptati… view at source ↗
Figure 3
Figure 3. Parameter comparison on two datasets. The hori [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 6
Figure 6. Figure 6: The analysis of different methods for obtaining [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 5
Figure 5. Figure 5: Visualization of the embeddings after VQ-VAE en [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 31 canonical work pages

  1. [1]

    Lu Bai, Abhishek Gupta, and Yew-Soon Ong. 2024. Multi-Task Learning with Multi-Task Optimization. arXiv preprint arXiv:2403.16162 (2024)

  2. [2]

    Mathilde Caron, Piotr Bojanowski, Armand Joulin, and Matthijs Douze. 2018. Deep clustering for unsupervised learning of visual features. In ECCV. 132–149

  3. [3]

    Mathilde Caron, Ishan Misra, Julien Mairal, Priya Goyal, Piotr Bojanowski, and Armand Joulin. 2020. Unsupervised learning of visual features by contrasting cluster assignments. NeurIPS 33 (2020), 9912–9924

  4. [4]

    Rich Caruana. 1997. Multitask learning. Machine learning 28 (1997), 41–75

  5. [5]

    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 SIGKDD. 3795–3804

  6. [6]

    Zhao Chen, Vijay Badrinarayanan, Chen-Yu Lee, and Andrew Rabinovich. 2018. Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks. In ICML. PMLR, 794–803

  7. [7]

    Heng-Tze Cheng, Levent Koc, Jeremiah Harmsen, Tal Shaked, Tushar Chandra, Hrishi Aradhye, Glen Anderson, Greg Corrado, Wei Chai, Mustafa Ispir, et al

  8. [8]

    Weiyu Cheng, Yanyan Shen, and Linpeng Huang. 2020. Adaptive factorization network: Learning adaptive-order feature interactions. In AAAI, Vol. 34. 3609– 3616

Show all 50 references
  1. [9]

    Zhongxiang Fan, Zhaocheng Liu, Jian Liang, Dongying Kong, Han Li, Peng Jiang, Shuang Li, and Kun Gai. 2024. Multi-Epoch learning with Data Augmentation for Deep Click-Through Rate Prediction. arXiv preprint arXiv:2407.01607 (2024)

  2. [10]

    William Fedus, Jeff Dean, and Barret Zoph. 2022. A review of sparse expert models in deep learning. arXiv preprint arXiv:2209.01667 (2022)

  3. [11]

    Yaroslav Ganin and Victor Lempitsky. 2015. Unsupervised domain adaptation by backpropagation. In ICML. PMLR, 1180–1189

  4. [12]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2015. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. In CVPR. 1026–1034

  5. [13]

    Yun He, Xue Feng, Cheng Cheng, Geng Ji, Yunsong Guo, and James Caverlee

  6. [14]

    Yassine Himeur, Somaya Al-Maadeed, Hamza Kheddar, Noor Al-Maadeed, Khalid Abualsaud, Amr Mohamed, and Tamer Khattab. 2023. Video surveillance using deep transfer learning and deep domain adaptation: Towards better generaliza- tion. EAAI 119 (2023), 105698

  7. [15]

    Robert A Jacobs, Michael I Jordan, Steven J Nowlan, and Geoffrey E Hinton. 1991. Adaptive mixtures of local experts. Neural computation 3, 1 (1991), 79–87

  8. [16]

    Pengyue Jia, Yichao Wang, Shanru Lin, Xiaopeng Li, Xiangyu Zhao, Huifeng Guo, and Ruiming Tang. 2024. D3: A Methodological Exploration of Domain Division, Modeling, and Balance in Multi-Domain Recommendations. In AAAI. 8553–8561

  9. [17]

    Biye Jiang, Pengye Zhang, Rihan Chen, Binding Dai, Xinchen Luo, Yin Yang, Guan Wang, Guorui Zhou, Xiaoqiang Zhu, and Kun Gai. 2020. DCAF: A Dy- namic Computation Allocation Framework for Online Serving System. CoRR abs/2006.09684 (2020)

  10. [18]

    Yuchen Jiang, Qi Li, Han Zhu, Jinbei Yu, Jin Li, Ziru Xu, Huihui Dong, and Bo Zheng. 2022. Adaptive domain interest network for multi-domain recommenda- tion. In CIKM. 3212–3221

  11. [20]

    Alex Kendall, Yarin Gal, and Roberto Cipolla. 2018. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. In CVPR. 7482– 7491

  12. [21]

    Kingma and Max Welling

    Diederik P. Kingma and Max Welling. 2013. Auto-Encoding Variational Bayes. CoRR abs/1312.6114 (2013). https://api.semanticscholar.org/CorpusID:216078090

  13. [22]

    Doyup Lee, Chiheon Kim, Saehoon Kim, Minsu Cho, and Wook-Shin Han. 2022. Autoregressive image generation using residual quantization. In CVPR. 11523– 11532

  14. [23]

    Xuejian Li, Ze Wang, Bingqi Zhu, Fei He, Yongkang Wang, and Xingxing Wang

  15. [24]

    Shikun Liu, Edward Johns, and Andrew J Davison. 2019. End-to-end multi-task learning with attention. In CVPR. 1871–1880

  16. [25]

    Ilya Loshchilov and Frank Hutter. 2017. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101 (2017)

  17. [26]

    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 SIGKDD. 1930–1939

  18. [27]

    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. In SIGIR. 1137–1140

  19. [28]

    Shlok Mishra, Anshul Shah, Ankan Bansal, Janit Anjaria, Jonghyun Choi, Ab- hinav Shrivastava, Abhishek Sharma, and David Jacobs. 2020. Learning visual representations for transfer learning by suppressing texture. arXiv preprint arXiv:2011.01901 (2020)

  20. [29]

    Shanlei Mu, Penghui Wei, Wayne Xin Zhao, Shaoguo Liu, Liang Wang, and Bo Zheng. 2023. Hybrid contrastive constraints for multi-scenario Ad ranking. In CIKM. 1857–1866

  21. [30]

    Sebastian Ruder. 2017. An overview of multi-task learning in deep neural net- works. arXiv preprint arXiv:1706.05098 (2017)

  22. [31]

    Xiang-Rong Sheng, Liqin Zhao, Guorui Zhou, Xinyao Ding, Binding Dai, Qiang Luo, Siran Yang, Jingshan Lv, Chi Zhang, Hongbo Deng, et al. 2021. One model to serve all: Star topology adaptive recommender for multi-domain ctr prediction. In CIKM. 4104–4113

  23. [32]

    Zihua Si, Lin Guan, ZhongXiang Sun, Xiaoxue Zang, Jing Lu, Yiqun Hui, Xingchao Cao, Zeyu Yang, Yichen Zheng, Dewei Leng, et al . 2024. TWIN V2: Scaling Ultra-Long User Behavior Sequence Modeling for Enhanced CTR Prediction at Kuaishou. arXiv preprint arXiv:2407.16357 (2024)

  24. [33]

    Vladan Stojnić and Vladimir Risojević. 2018. Analysis of color space quantization in split-brain autoencoder for remote sensing image classification. In NEUREL. IEEE, 1–4

  25. [34]

    Shuo Su, Xiaoshuang Chen, Yao Wang, Yulin Wu, Ziqiang Zhang, Kaiqiao Zhan, Ben Wang, and Kun Gai. 2024. RPAF: A Reinforcement Prediction-Allocation Framework for Cache Allocation in Large-Scale Recommender Systems. In Pro- ceedings of the 18th ACM Conference on Recommender Sys...

  26. [35]

    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 RecSys. 269–278

  27. [36]

    Eric Tzeng, Judy Hoffman, Kate Saenko, and Trevor Darrell. 2017. Adversarial discriminative domain adaptation. In CVPR. 7167–7176

  28. [37]

    Aaron Van Den Oord, Oriol Vinyals, et al. 2017. Neural discrete representation learning. NeurIPS 30 (2017)

  29. [38]

    Yunli Wang, Zhiqiang Wang, Jian Yang, Shiyang Wen, Dongying Kong, Han Li, and Kun Gai. 2024. Adaptive Neural Ranking Framework: Toward Maximized Business Goal for Cascade Ranking Systems. In WWW. ACM, 3798–3809

  30. [39]

    Yunli Wang, Yu Wu, Lili Mou, Zhoujun Li, and Wenhan Chao. 2020. Formality style transfer with shared latent space. In COLING. 2236–2249

  31. [40]

    Yunli Wang, Zixuan Yang, Zhen Zhang, Zhiqiang Wang, Jian Yang, Shiyang Wen, Peng Jiang, and Kun Gai. 2024. Scaling Laws for Online Advertisement Retrieval. arXiv preprint arXiv:2411.13322 (2024)

  32. [41]

    Zirui Wang, Yulia Tsvetkov, Orhan Firat, and Yuan Cao. 2020. Gradient vaccine: Investigating and improving multi-task optimization in massively multilingual models. arXiv preprint arXiv:2010.05874 (2020)

  33. [42]

    Jian Yang, Shuming Ma, Li Dong, Shaohan Huang, Haoyang Huang, Yuwei Yin, Dongdong Zhang, Liqun Yang, Furu Wei, and Zhoujun Li. 2022. Ganlm: Encoder-decoder pre-training with an auxiliary discriminator. arXiv preprint arXiv:2212.10218 (2022)

  34. [43]

    Jian Yang, Shuming Ma, Dongdong Zhang, Shuangzhi Wu, Zhoujun Li, and Ming Zhou. 2020. Alternating language modeling for cross-lingual pre-training. In AAAI, Vol. 34. 9386–9393

  35. [44]

    Xun Yang, Yunli Wang, Cheng Chen, Qing Tan, Chuan Yu, Jian Xu, and Xiaoqiang Zhu. 2021. Computation Resource Allocation Solution in Recommender Systems. CoRR abs/2103.02259 (2021)

  36. [45]

    Çağrı Yeşil and Kaya Turgut. 2024. Star+: A New Multi-Domain Model for CTR Prediction. arXiv preprint arXiv:2406.16568 (2024)

  37. [46]

    Wenqiao Zhang, Zheqi Lv, Hao Zhou, Jia-Wei Liu, Juncheng Li, Mengze Li, Yunfei Li, Dongping Zhang, Yueting Zhuang, and Siliang Tang. 2024. Revisiting the domain shift and sample uncertainty in multi-source active domain transfer. In CVPR. 16751–16761

  38. [47]

    Yiqian Zhang, Yinfu Feng, Wen-Ji Zhou, Yunan Ye, Min Tan, Rong Xiao, Haihong Tang, Jiajun Ding, and Jun Yu. 2024. Multi-Domain Deep Learning from a Multi- View Perspective for Cross-Border E-commerce Search. In AAAI. 9387–9395

  39. [48]

    Lulu Zhao, Fujia Zheng, Weihao Zeng, Keqing He, Ruotong Geng, Huixing Jiang, Wei Wu, and Weiran Xu. 2022. Adpl: Adversarial prompt-based domain adap- tation for dialogue summarization with knowledge disentanglement. In SIGIR. 245–255. A EV ALUATION DETAILS OF FLOPS AND PARAMET...

  40. [2016]

    Wide & deep learning for recommender systems. In DLRS. 7–10

  41. [2022]

    Metabalance: improving multi-task recommendations via adapting gradient magnitudes of auxiliary tasks. In WWW. 2205–2215

  42. [2024]

    In SIGIR

    Deep automated mechanism design for integrating ad auction and allocation in feed. In SIGIR. 1211–1220

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.