Pre-train and Fine-tune: Recommenders as Large Models

Chenghao Chen; Hao Feng; Jia Jia; Jie Zhang; Jin Liu; Ning Hu; Yu Yang; Zhenhao Jiang

REVIEW 5 major objections 6 minor 64 references

Pre-train and Fine-tune: Recommenders as Large Models

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Fine-tuning a frozen recommender with a small plug-in module can beat zero-shot serving across domains.

desk verdict The IAK adapter shows real gains on a billion-scale platform, but the information-bottleneck theory that supposedly justifies it does not survive contact with the appendix. read the letter →

arxiv 2501.14268 v1 pith:DZOFVOTK submitted 2025-01-24 cs.IR cs.AI

Zhenhao Jiang , Chenghao Chen , Hao Feng , Yu Yang , Jin Liu , Jie Zhang , Jia Jia , Ning Hu This is my paper · ORCID

classification cs.IRcs.AI

keywords recommendationsysteminformationbottleneckfine-tuningmulti-domainlearningpre-trainedrecommenderInformation-AwareAdaptiveKernelCTRpredictionandCTCVR

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

Recommenders in industry are huge, expensive to retrain, and serve users whose tastes shift across scenes, regions, and times. The paper proposes to treat a trained recommender as a large pre-trained model and adapt it per domain by fine-tuning only a small plugin module, the Information-Aware Adaptive Kernel (IAK), while freezing the base model. It gives an information-bottleneck account of fine-tuning, splitting it into knowledge compression and knowledge matching, and derives a practical objective from an upper bound. On 11 offline datasets from a food platform, IAK improves the pre-trained model's CTR-AUC by 0.26% to 0.66% and CTCVR-AUC by 0.11% to 0.49% on average across domain groups, and online A/B tests report order-rate gains of 0.46% to 1.14%. If the claim holds, multi-domain personalization becomes a cheap add-on to existing industrial recommenders rather than a full retraining project.

What carries the argument

The load-bearing mechanism is the Information-Aware Adaptive Kernel (IAK), an MLP encoder-decoder inserted into the frozen pre-trained recommender. It is trained with objective (19): minimize the label cross-entropy $H(\hat{Y}|Y)$ plus the KL divergence $D_{\mathrm{KL}}(w\|w_0)$ between the trained adapter weights and their Gaussian initialization. That KL term is the practical stand-in for the information-bottleneck compression term, while the encoder performs knowledge compression and the decoder performs knowledge matching with the downstream task data.

What would settle it

Train an IAK on one domain and record pairs of weight-space KL distance $D_{\mathrm{KL}}(w\|w_0)$ and the task-relevant mutual information of the encoder output across checkpoints; if the two are not positively correlated, Assumption 1 fails and the information-bottleneck justification for objective (19) does not carry, leaving the module's gains to be explained by other effects such as regularization.

Watch

Extended reading notes

Core claim

The paper's central claim is that fine-tuning a pre-trained recommender can be understood as solving an information-bottleneck problem, and that a small encoder-decoder module trained on objective (19) approximates that solution. IAK freezes the billion-parameter base model and trains only the adapter, compressing general business knowledge and then matching it to the downstream domain. In experiments, OLR+IAK outperforms zero-shot OLR by on average +0.26% to +0.66% in CTR-AUC and +0.11% to +0.49% in CTCVR-AUC across multi-scene, multi-period, and multi-region tasks, and online A/B tests show order-rate gains of 0.46% to 1.14%. The authors also state that IAK is a universal module that can be applied to any recommendation model, and that the version deployed to the homepage of a billion-scale online food platform has run for several months and produced considerable business profit.

Load-bearing premise

The load-bearing premise is that the distance between the trained adapter weights and their initialization is positively correlated with the amount of information those weights contain, which is the assumption that converts the information-bottleneck bound into the practical training objective.

Editorial extensions

If this is right

One frozen industrial recommender can serve many scenes, regions, and time periods by training only small per-domain IAK modules, avoiding the cost of retraining the full model.
The information-bottleneck decomposition of fine-tuning into knowledge compression and knowledge matching gives future adapter designs an interpretable objective to optimize.
Because the base model stays frozen, new domains can be added by training and deploying additional lightweight adapters, which can run in parallel and be domain-activated at serving time.
The method transfers to standard multi-task baselines as well as the proprietary recommender, supporting the claim that IAK is model-agnostic.
The reported online order-rate improvements imply that fine-tuning adapters can produce business-visible gains in real traffic, not just offline metric changes.

Reading between the lines

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

An extension the authors do not draw: the empirical gains may come from the KL penalty acting as a regularizer on a small adapter rather than from genuine information compression, and a direct comparison with other regularizers would separate those explanations.
The paper's weighted-sample experiment for overlapping users and items suggests that sharing data across related domains during fine-tuning can be formalized as a multi-task objective; the authors only demonstrate a heuristic version.
A testable extension is to use the dynamic batch-aware learning rate from equation (21) with automatic grouping of similar tasks, since the authors note that tasks of similar magnitude train better together.
If Assumption 1 fails, the information-bottleneck interpretation of objective (19) would no longer carry, but the IAK module could still survive as a practical adapter; the theory and the module are separable claims.

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.

Desk Editor's Note

read the letter

Here's my take on the IAK paper. The empirical core is real: an encoder-decoder adapter with a KL penalty to a Gaussian prior, applied to a frozen billion-parameter recommender, improves CTR/CTCVR across domains and lifts online order rate in A/B tests on a large food platform. That is a practical result worth knowing, and the deployment lessons (parallel inference, dynamic batch training) are genuinely useful. If you work on industrial fine-tuning for recommenders, this is a reasonable starting point.

But the theory is not just thin; it's broken in a load-bearing way. The derivation from Eq. (12) to Eq. (19) drops p(Ĝ|G) by treating it as a probability ≤1. That is false for continuous densities, and even in a discrete reading it throws away the dependence on the pre-trained knowledge G. The resulting expression is no longer an upper bound on I(Ĝ;G). Then Assumption 1, which asserts a positive correlation between parameter-space KL and information content, is doing all the heavy lifting to convert the IB objective into the practical loss. The paper gives no evidence for it, and it's doing exactly the kind of work that should be derived, not assumed. The stress-test note is accurate: the connection between the IB bound and D_KL(w||w0) is asserted, not established. The authors also call Theorem 1 the "information bottleneck for fine-tuning," but it is just the standard IB objective with variables renamed. So the interpretability contribution collapses.

On the empirical side, there are also soft spots. The offline tables report no error bars or significance tests, and Scene2 CTR actually drops (which the text mentions but never addresses). The single-domain online A/B test was run on Region4, which had the largest offline gain, so cherry-picking is a real concern. The baselines are classic multi-task models—Shared Bottom, ESMM, MMoE—not modern multi-domain methods like STAR or ADIN, so the comparison against the state of the art is weaker than claimed.

What does the paper do well? The A/B tests and the sheer scale are credible, and the case study in Figure 4 is a nice sanity check. The discussion of pseudo cold start and user/item overlap is thoughtful and shows the authors understand the deployment constraints. The control tests on baselines give some evidence that IAK is model-agnostic.

Who is this for? Practitioners who want a low-cost adapter for fine-tuning a large recommender and are willing to ignore the theory. It deserves a serious referee because of the industrial evidence, but the paper needs major revision: either remove the theory entirely and present IAK as an empirical adapter, or fix the mathematics and validate Assumption 1. As written, I would not cite it for the theory, and I would not send it to a reading group for anything other than a cautionary tale about over-claiming. Send it to review, but expect the theoretical contribution to be dismantled.

Referee Report

5 major / 6 minor

Summary. The manuscript proposes IAK, a plug-and-play encoder-decoder module for fine-tuning large pre-trained recommender systems on downstream domains. It claims a theoretical justification based on an information-bottleneck upper bound, formulates fine-tuning as two phases of knowledge compression and knowledge matching, and reports offline gains of OLR+IAK over zero-shot OLR as well as online A/B test lifts, with deployment at a large-scale food platform. The central theoretical derivation contains several invalid mathematical steps, and the empirical evaluation is reported without statistical support.

Significance. If the theory were sound, the paper would offer a principled, model-agnostic fine-tuning method for industrial recommenders, which would be a useful contribution given the high cost of retraining large models. The empirical work is substantial: 11 proprietary datasets, comparisons with strong baselines, cross-topic tests, and live A/B tests with business metrics. The two practical issues identified (pseudo cold start and user/item overlap) and the deployment lessons are also of interest to practitioners. However, the theoretical foundation is not sound: the upper-bound derivation relies on invalid inequalities, the final objective is connected to the information bottleneck only through an unverified assumption, and the offline and online results are given without error bars, confidence intervals, or significance tests. The paper does not provide code or public data, so the industrial results cannot be independently verified. The contribution is therefore best characterized as an empirically motivated adapter with a post-hoc information-theoretic narrative, not a validated theory of fine-tuning.

major comments (5)

[Section 4.3, Eq. (12) and Appendix A.1, Eq. (24)] The step from the third to the fourth line of Eq. (24) drops the density p(Ĝ,G) from the integrand, replacing it by 1. This is invalid for continuous densities, which are not bounded by 1, and, even for discrete probabilities, p·log(a) ≤ log(a) holds only when log(a) ≥ 0, which is not guaranteed here. Consequently, the expression in Eq. (12) is not established as an upper bound on I(Ĝ;G), and minimizing it is not equivalent to minimizing the information-bottleneck objective. In addition, the jump from Eq. (8) to Eq. (10) is not derived: H(Ĝ|T) is a conditional entropy of the representation, not the cross-entropy H(Ŷ|Y) of predictions.
[Section 5.3, Eq. (25) and Appendix A.2] The replacement p(Ĝ_i|G_i) ≤ 1 in Eq. (25) is invalid for conditional densities and, more importantly, discards the dependence of the bound on the pre-trained knowledge G. Dropping p(Ĝ_i|G_i) removes precisely the term that couples the bound to the pre-trained model, so the resulting expression is not a valid upper bound on the IB objective. Furthermore, the step from (β/N)Σ log(1/r(Ĝ)) to -β log r(Ĝ) is arithmetically incorrect unless log r(Ĝ) is constant across samples; the sample average is silently dropped. Equation (14) also commits a type error by writing Ĝ ≤ H(Ŷ|Y), since Ĝ is a random variable, not a scalar objective.
[Section 5.3, Assumption 1 and Eq. (19)] The substitution of D_KL(w∥w0) for -β log r(Ĝ) is justified only by Assumption 1, which states that the distance between the trained and original parameter distributions is positively correlated with the information contained in the parameter. The manuscript provides no evidence for this correlation and reports no experiment that measures it. Even if such a correlation existed, D_KL(w∥w0) is a distance between distributions of encoder weights, not between distributions of the representation Ĝ, so it does not operationalize the information-bottleneck term I(Ĝ;G). The central claim that IAK is derived from the essence of fine-tuning therefore rests on an unsupported heuristic.
[Section 4.2, Theorems 2 and 3] Theorems 2 and 3 are not derived from Theorem 1; they are new optimization objectives stated without proof. The paper does not show that the knowledge-compression and knowledge-matching phases follow from the information-bottleneck formulation, nor how the IAK encoder-decoder training stage explicitly approximates these two phases. The two-phase interpretation is an informal narrative rather than a theorem, yet it is presented as part of the theoretical contribution.
[Sections 7.2-7.7] The empirical claims are not supported by statistical analysis. Offline results in Tables 2 and 3 are reported as point estimates without error bars, standard deviations, or significance tests, even though the reported improvements are small (e.g., average CTR-AUC +0.26% for multi-scene). The online A/B tests in Section 7.7 report average lifts over 7-30 days but give no confidence intervals or p-values, so the observed differences cannot be distinguished from noise. The text also states in Section 7.2 that OLR+IAK outperforms ZS-OLR on all datasets except Scene2 in CTR-AUC, which is inconsistent with the subsequent characterization of the gains as 'huge improvement' and with the universality claim in Section 7.3.

minor comments (6)

[Section 4.3, Eq. (9)] Equation (9) defines the cross-entropy between p and q but labels it with H(p(X), q(X)); this definition is not used consistently and is confusing.
[Section 4.3 and Appendix A.1] The coefficient β appears in Eq. (12) without derivation and is introduced only at the end of Appendix A.1; the notation should be made consistent.
[Section 5.4.2, Eq. (21)] In Eq. (21), W is described as 'the reciprocal of the gradient for IAKs' without a precise definition, and the behavior of the SoftMax operation over N_B is not explained.
[Section 7.3, Figure 1] Figure 1 is a heatmap without numeric labels, so the claimed consistent improvements of IAK on the baseline models cannot be verified from the figure alone.
[References] Several entries in the reference list do not appear to be cited in the body of the paper, including [8], [10], [19], [20], [22], and [57]; the citation list should be checked.
[End matter] The line 'Received 20 February 2007; revised 12 March 2009; accepted 5 June 2009' appears to be a template artifact and should be removed.

Circularity Check

2 steps flagged · score 6.0 of 10

The 'information bottleneck for fine-tuning' is the standard IB objective with renamed variables, and the practical IAK objective (19) is reached only via Assumption 1, which injects the KL regularizer the method uses; the empirical IAK results remain an independent test.

renaming known result [Section 4.1, Theorem 1, Eq. (5)]
"Theorem 1 (Information Bottleneck for Fine-tuning). Given the general knowledge in the pre-trained model 𝐺 and the knowledge in the given downstream task 𝑇 , the optimization objective of Information Bottleneck for Fine-tuning is to find the minimum sufficient representation ˆ𝐺: ˆ𝐺 = arg min−𝐼( ˆ𝐺;𝑇)+ 𝛽𝐼( ˆ𝐺;𝐺). (5)"

Eq. (5) is exactly Eq. (1) from Section 3 with X→G, Y→T, and X̂→Ĝ. The paper presents this variable re-labeling as a new theorem ('we first propose the theory of the information bottleneck for fine-tuning'), but no fine-tuning-specific derivation or content is added. The claim that fine-tuning should be understood via IB is a choice of framing, and Theorem 1 is the known IB objective restated; the subsequent 'two phases' are then read back into this renamed objective. This is the 'renaming a known result' pattern rather than a derived prediction.
self definitional [Section 5.3, Eqs. (15)-(19) and Assumption 1]
"Based on the above assumptions, the objective can be simplified as follows. More details are given in Appendix B. arg min𝐻( ˆ𝑌|𝑌)+ 𝐷𝐾𝐿(𝑤∥𝑤0). (19) Here, we make the following assumption. Assumption 1. The distance between the distribution of the trained parameter and that of the original parameter is positively correlated with the information contained in the parameter."

The upper bound in (12) still contains p(Ĝ|G), the term that couples the representation to the pre-trained knowledge. Appendix A.2 drops it, leaving -β log r(Ĝ). The final objective (19) replaces this with D_KL(w∥w0), a distance on encoder parameters. The sole bridge is Assumption 1, which asserts a positive correlation between exactly that parameter-space distance and the information content. Since the KL penalty is the very regularizer IAK is built on, Assumption 1 does not derive the objective from IB; it stipulates that the desired term stands in for the IB compression term. No experiment checks the correlation, so the link is definitional/assumed, not established.

full rationale

The empirical results (RQ2, RQ3, RQ7) compare OLR+IAK against ZS-OLR and standard baselines on held-out offline logs and online A/B tests; those comparisons are an independent test of the IAK module, so the practical/performance claim is not circular. I found no load-bearing self-citation: the authors' prior papers [10], [19], [20] are cited as background and do not supply the argument. The circularity is in the theoretical derivation. Theorem 1 (Eq. 5) reproduces the standard IB objective (Eq. 1) with variables renamed (X→G, Y→T, X̂→Ĝ) and is offered as 'the theory of information bottleneck for fine-tuning'; that is pattern 6 (renaming a known result). More importantly, the route from the IB upper bound to the actual IAK objective is by construction: Eq. (15) still has -β log r(Ĝ), while Eq. (19) uses D_KL(w∥w0). Appendix A.2 first removes p(Ĝ|G) from the bound, and then the replacement of -β log r(Ĝ) by D_KL(w∥w0) is justified only by Assumption 1, which asserts the positive correlation between this exact parameter-space KL and 'the information contained in the parameter.' This is a self-definitional step: the regularizer that IAK minimizes is assumed to be the IB compression term rather than derived from it, and no experiment measures the asserted correlation. The invalid inequality log p(Ĝ|G) ≤ 0 used in A.2 is a correctness issue rather than a circularity. Overall, the central theoretical claim partially reduces to its own inputs; empirical evaluation keeps the score at 6 rather than 8-10.

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

The theoretical derivation rests on the assumption that fine-tuning can be cast as an information bottleneck with G and T as random variables, plus Assumption 1 connecting KL divergence of parameters to information content, neither of which is derived or empirically validated. The practical objective (19) is a standard cross-entropy plus KL regularizer.

free parameters (3)

KL weight β = 1 (loss weights set to 1 in experiments)
The IB objective (5) includes β, but the final objective (19) is used with all loss weights set to 1, with no sensitivity analysis.
Encoding dimension d_e = 50
Chosen from the parameter sensitivity test (Fig. 2); performance saturates as d_e increases, and 50 is recommended.
IAK MLP widths = [1024, 512, 256]
Appendix C sets all DNNs to three layers of [1024,512,256], presumably including IAK, without justification.

assumptions (3)

domain assumption Fine-tuning a pre-trained recommender can be modeled as the information bottleneck objective (5) with general knowledge G and downstream task T.
Asserted in Theorem 1, Section 4.1, without formal justification that G and T are random variables with a well-defined joint distribution.
ad hoc to paper Assumption 1: D_KL between trained and initial encoder parameters is positively correlated with the information contained in the parameters.
Stated in Section 5.3; it is the bridge that turns the IB upper bound into the practical KL objective, and no evidence is given.
ad hoc to paper The encoder parameters follow a Gaussian distribution whose KL to the standard normal prior approximates -log r(Ĝ).
Section 5.3 and Appendix B; the variational approximation r(Ĝ) is replaced by a Gaussian over weights without justification.

invented entities (1)

IAK (Information-Aware Adaptive Kernel) module independent evidence
purpose: Fine-tuning adapter that compresses general knowledge and matches domain-specific knowledge for large frozen recommenders
The paper provides offline AUC tables and online A/B results for IAK, though without error bars or disclosed code.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Pre-train and Fine-tune: Recommenders as Large Models." pith.science (2026). https://pith.science/paper/DZOFVOTK

@misc{pith2026250114268,
  author       = {Pith},
  title        = {Pith review of: Pre-train and Fine-tune: Recommenders as Large Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DZOFVOTK}},
  note         = {Machine review of arXiv:2501.14268}
}

read the original abstract

In reality, users have different interests in different periods, regions, scenes, etc. Such changes in interest are so drastic that they are difficult to be captured by recommenders. Existing multi-domain learning can alleviate this problem. However, the structure of the industrial recommendation system is complex, the amount of data is huge, and the training cost is extremely high, so it is difficult to modify the structure of the industrial recommender and re-train it. To fill this gap, we consider recommenders as large pre-trained models and fine-tune them. We first propose the theory of the information bottleneck for fine-tuning and present an explanation for the fine-tuning technique in recommenders. To tailor for recommendation, we design an information-aware adaptive kernel (IAK) technique to fine-tune the pre-trained recommender. Specifically, we define fine-tuning as two phases: knowledge compression and knowledge matching and let the training stage of IAK explicitly approximate these two phases. Our proposed approach designed from the essence of fine-tuning is well interpretable. Extensive online and offline experiments show the superiority of our proposed method. Besides, we also share unique and important lessons we learned when deploying the method in a large-scale online platform. We also present the potential issues of fine-tuning techniques in recommendation systems and the corresponding solutions. The recommender with IAK technique has been deployed on the homepage of a billion-scale online food platform for several months and has yielded considerable profits in our business.

Figures

Figures reproduced from arXiv: 2501.14268 by the authors.

**Figure 1.** visualize the results of the control test. We can see that IAK fine-tuning technique consistently improves the performance of (a) Results of control test in terms of CTR-AUC. (b) Results of control test in terms of CTCVR-AUC [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗

**Figure 2.** Results of parameter sensitivity test. more and more slowly. Even though an increase in time span can further improve model performance, it is not reasonable to increase the time span all the time. Besides, training data over a week-long span is recommended for fine-tuning. To evaluate the user/item overlapping issue, we try not to completely isolate samples from different domains. For Scene1, we use Scene1 samples… view at source ↗

**Figure 4.** Some hot items in Region4, the vertical axis repre [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

Figures from the paper (1 more)

**Figure 3.** Figure 3: Results of exploration study. 7.7 RQ7: Online A/B Test To evaluate the performance of OLR+IAK in large-scale online recommender, we conducted three online A/B tests (one single IAK test and two multiple IAKs tests) by deploying it to the recommendation scenario on th…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 37 canonical work pages

[1]

Sajad Ahmadian, Nima Joorabloo, Mahdi Jalili, and Milad Ahmadian. 2022. Al- leviating data sparsity problem in time-aware recommender systems using a reliable rating profile enrichment approach. Expert Systems with Applications 187 (2022), 115849

work page 2022
[2]

Robert B Ash. 2012. Information theory. Courier Corporation

work page 2012
[3]

Som S Biswas. 2023. Potential use of chat gpt in global warming. Annals of biomedical engineering 51, 6 (2023), 1126–1127

work page 2023
[4]

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

1997
[5]

Jiawei Chen, Hande Dong, Xiang Wang, Fuli Feng, Meng Wang, and Xiangnan He. 2023. Bias and debias in recommender system: A survey and future directions. ACM Transactions on Information Systems 41, 3 (2023), 1–39

2023
[6]

Qiwei Chen, Changhua Pei, Shanshan Lv, Chao Li, Junfeng Ge, and Wenwu Ou
[7]

Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. 2017. Deep reinforcement learning from human preferences. Advances in neural information processing systems 30 (2017)

2017
[8]

Alexandre Défossez, Léon Bottou, Francis Bach, and Nicolas Usunier. 2020. A simple convergence proof of adam and adagrad. arXiv preprint arXiv:2003.02395 (2020)

arXiv 2020

Show all 64 references

[9]

Yihe Dong, Jean-Baptiste Cordonnier, and Andreas Loukas. 2021. Attention is not all you need: Pure attention loses rank doubly exponentially with depth. In International Conference on Machine Learning . PMLR, 2793–2803

2021
[10]

Boya Du, Shaochuan Lin, Jiong Gao, Xiyu Ji, Mengya Wang, Taotao Zhou, Hengxu He, Jia Jia, and Ning Hu. 2023. BASM: A Bottom-up Adaptive Spatiotemporal Model for Online Food Ordering Service. In 2023 IEEE 39th International Confer- ence on Data Engineering (ICDE) . IEEE, 3549–3562

2023
[11]

Ziv Goldfeld and Yury Polyanskiy. 2020. The information bottleneck problem and its applications in machine learning. IEEE Journal on Selected Areas in Information Theory 1, 1 (2020), 19–38

2020
[12]

Elizabeth Gómez, Ludovico Boratto, and Maria Salamó. 2022. Provider fairness across continents in collaborative recommender systems. Information Processing & Management 59, 1 (2022), 102719

2022
[13]

Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2014. Generative adversarial nets. Advances in neural information processing systems 27 (2014)

2014
[14]

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

2021
[15]

Junyou He, Guibao Mei, Feng Xing, Xiaorui Yang, Yongjun Bao, and Weipeng Yan
[16]

Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition . 770–778

2016
[17]

Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685 (2021)

2021 arXiv
[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 Proceedings of the 31st ACM International Conference on Information & Knowledge Management. 3212–3221

2022
[19]

Zhenhao Jiang, Biao Zeng, Hao Feng, Jin Liu, Jicong Fan, Jie Zhang, Jia Jia, Ning Hu, Xingyu Chen, and Xuguang Lan. 2023. ESMC: Entire Space Multi-Task Model for Post-Click Conversion Rate via Parameter Constraint. arXiv preprint arXiv:2307.09193 (2023)

2023 arXiv
[20]

Zhenhao Jiang, Biao Zeng, Hao Feng, Jin Liu, Jie Zhang, Jia Jia, and Ning Hu
[21]

Xuan Nhat Lam, Thuc Vu, Trong Duc Le, and Anh Duc Duong. 2008. Addressing cold-start problem in recommendation systems. In Proceedings of the 2nd inter- national conference on Ubiquitous information management and communication . 208–211

2008
[22]

Hao Li, Charless Fowlkes, Hao Yang, Onkar Dabeer, Zhuowen Tu, and Stefano Soatto. 2023. Guided recommendation for model fine-tuning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 3633–3642

2023
[23]

Roger Zhe Li, Julián Urbano, and Alan Hanjalic. 2023. Mitigating Mainstream Bias in Recommendation via Cost-sensitive Learning. In Proceedings of the 2023 ACM SIGIR International Conference on Theory of Information Retrieval . 135–142

2023
[24]

Yakun Li, Lei Hou, Dongmei Li, and Juanzi Li. 2023. HKGCL: Hierarchical graph contrastive learning for multi-domain recommendation over knowledge graph. Expert Systems with Applications 233 (2023), 120963

2023
[25]

Blerina Lika, Kostas Kolomvatsos, and Stathes Hadjiefthymiades. 2014. Facing the cold start problem in recommender systems. Expert systems with applications 41, 4 (2014), 2065–2073

2014
[26]

Haochen Liu, Xiangyu Zhao, Chong Wang, Xiaobing Liu, and Jiliang Tang. 2020. Automated embedding size search in deep recommender systems. In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. 2307–2316

2020
[27]

Junling Liu, Chao Liu, Renjie Lv, Kang Zhou, and Yan Zhang. 2023. Is chatgpt a good recommender? a preliminary study. arXiv preprint arXiv:2304.10149 (2023)

2023 arXiv
[28]

Jie Lu, Dianshuang Wu, Mingsong Mao, Wei Wang, and Guangquan Zhang
[29]

Linyuan Lü, Matúš Medo, Chi Ho Yeung, Yi-Cheng Zhang, Zi-Ke Zhang, and Tao Zhou. 2012. Recommender systems. Physics reports 519, 1 (2012), 1–49

2012
[30]

Linhao Luo, Yumeng Li, Buyu Gao, Shuai Tang, Sinan Wang, Jiancheng Li, Tanchao Zhu, Jiancai Liu, Zhao Li, and Shirui Pan. 2022. MAMDR: a model agnostic learning method for multi-domain recommendation. arXiv preprint arXiv:2202.12524 (2022)

2022 arXiv
[31]

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
[32]

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 The 41st International ACM SIGIR Conference on Research & Development in Information Retrie...

2018
[33]

Jarana Manotumruksa and Emine Yilmaz. 2020. Sequential-based adversarial optimisation for personalised top-n item recommendation. In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. 2045–2048

2020
[34]

Alan Tan Wei Min, Ramon Sagarna, Abhishek Gupta, Yew-Soon Ong, and Chi Keong Goh. 2017. Knowledge transfer through machine learning in air- craft design. IEEE Computational Intelligence Magazine 12, 4 (2017), 48–60

2017
[35]

Wentao Ning, Xiao Yan, Weiwen Liu, Reynold Cheng, Rui Zhang, and Bo Tang
[36]

Pyeonghwa Oh and Byung-Yeon Hwang. 2017. Real-time Spatial Recommen- dation System based on Sentiment Analysis of Twitter. Journal of Society for e-Business Studies 21, 3 (2017)

2017
[37]

Fulan Qian, Kaili Qin, Hai Chen, Jie Chen, Shu Zhao, and Yanping Zhang. 2023. Adaptive social recommendation combined with the multi-domain influence. Information Systems 113 (2023), 102145

2023
[38]

Andrew M Saxe, Yamini Bansal, Joel Dapello, Madhu Advani, Artemy Kolchinsky, Brendan D Tracey, and David D Cox. 2019. On the information bottleneck theory of deep learning. Journal of Statistical Mechanics: Theory and Experiment 2019, 12 (2019), 124020

2019
[39]

Ohad Shamir, Sivan Sabato, and Naftali Tishby. 2010. Learning and generalization with the information bottleneck. Theoretical Computer Science 411, 29-30 (2010), 2696–2711

2010
[40]

arXiv preprint arXiv:2308.05508 (2023)

Multi-domain Recommendation with Embedding Disentangling and Do- main Alignment. arXiv preprint arXiv:2308.05508 (2023)

2023 arXiv
[41]

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 Proceedings of the 30th ACM Internationa...

2021
[42]

Jagendra Singh, Mohammad Sajid, Chandra Shekhar Yadav, Shashank Sheshar Singh, and Manthan Saini. 2022. A Novel Deep Neural-based Music Recom- mendation Method considering User and Song Data. In 2022 6th International Conference on Trends in Electronics and Informatics (ICOEI)...

2022
[43]

Yueming Sun and Yi Zhang. 2018. Conversational recommender system. In The 41st international acm sigir conference on research & development in information retrieval. 235–244

2018
[44]

Shulong Tan, Meifang Li, Weijie Zhao, Yandan Zheng, Xin Pei, and Ping Li. 2021. Multi-Task and Multi-Scene Unified Ranking Model for Online Advertising. In 2021 IEEE International Conference on Big Data (Big Data) . IEEE, 2046–2051

2021
[45]

Qijie Shen, Wanjie Tao, Jing Zhang, Hong Wen, Zulong Chen, and Quan Lu. 2021. Sar-net: a scenario-aware ranking network for personalized fair recommendation in hundreds of travel scenarios. In Proceedings of the 30th ACM International Conference on Information & Knowledge Mana...

2021
[46]

Genevieve N Thompson, Carole A Estabrooks, and Lesley F Degner. 2006. Clari- fying the concepts in knowledge transfer: a literature review.Journal of advanced nursing 53, 6 (2006), 691–701

2006
[47]

Naftali Tishby, Fernando C Pereira, and William Bialek. 2000. The information bottleneck method. arXiv preprint physics/0004057 (2000)

2000 arXiv
[48]

Naftali Tishby and Noga Zaslavsky. 2015. Deep learning and the information bottleneck principle. In 2015 ieee information theory workshop (itw) . IEEE, 1–5

2015
[49]

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
[50]

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 WWW Companion ’25, April 28-May 2, 2025, Sydney...

2020
[51]

Tailin Wu, Ian Fischer, Isaac L Chuang, and Max Tegmark. 2020. Learnability for the information bottleneck. In Uncertainty in Artificial Intelligence . PMLR, 1050–1060

2020
[52]

Tailin Wu, Hongyu Ren, Pan Li, and Jure Leskovec. 2020. Graph information bottleneck. Advances in Neural Information Processing Systems 33 (2020), 20437– 20448

2020
[53]

Yingyi Zhang, Xianneng Li, Yahe Yu, Jian Tang, Huanfang Deng, Junya Lu, Yeyin Zhang, Qiancheng Jiang, Yunsen Xian, Liqian Yu, et al. 2023. Meta-Generator Enhanced Multi-Domain Recommendation. InCompanion Proceedings of the ACM Web Conference 2023. 485–489

2023
[54]

Yu Zhang and Qiang Yang. 2018. An overview of multi-task learning. National Science Review 5, 1 (2018), 30–43

2018
[55]

Hong Wen, Jing Zhang, Yuan Wang, Fuyu Lv, Wentian Bao, Quan Lin, and Keping Yang. 2020. Entire space multi-task modeling via post-click behavior decomposi- tion for conversion rate prediction. In Proceedings of the 43rd International ACM SIGIR conference on research and develo...

2020
[56]

Zihan Zhang, Xiaoming Jin, Lianghao Li, Guiguang Ding, and Qiang Yang. 2016. Multi-domain active learning for recommendation. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 30

2016
[57]

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 , Vol. 33. 5941–5948

2019
[58]

Yuanhang Zhou, Kun Zhou, Wayne Xin Zhao, Cheng Wang, Peng Jiang, and He Hu. 2022. C 2-CRS: Coarse-to-Fine Contrastive Learning for Conversational Rec- ommender System. In Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining . 1488–1496

2022
[59]

Ziwei Zhu and James Caverlee. 2022. Fighting mainstream bias in recommender systems via local fine tuning. In Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining . 1497–1506. A Mathematical Derivation A.1 Mathematical Derivation in Section 4...

2022
[60]

Yu Zhang and Qiang Yang. 2021. A survey on multi-task learning. IEEE Transac- tions on Knowledge and Data Engineering 34, 12 (2021), 5586–5609

2021
[2015]

Decision support systems 74 (2015), 12–32

Recommender system application developments: a survey. Decision support systems 74 (2015), 12–32

2015
[2020]

arXiv preprint arXiv:2011.11938 (2020)

Dadnn: multi-scene ctr prediction via domain-aware deep neural network. arXiv preprint arXiv:2011.11938 (2020)

2020 arXiv
[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
[2023]

arXiv preprint arXiv:2308.03855 (2023)

Mobile Supply: The Last Piece of Jigsaw of Recommender System. arXiv preprint arXiv:2308.03855 (2023)

2023 arXiv

Pith tools

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

[1] [1]

Sajad Ahmadian, Nima Joorabloo, Mahdi Jalili, and Milad Ahmadian. 2022. Al- leviating data sparsity problem in time-aware recommender systems using a reliable rating profile enrichment approach. Expert Systems with Applications 187 (2022), 115849

work page 2022

[2] [2]

Robert B Ash. 2012. Information theory. Courier Corporation

work page 2012

[3] [3]

Som S Biswas. 2023. Potential use of chat gpt in global warming. Annals of biomedical engineering 51, 6 (2023), 1126–1127

work page 2023

[4] [4]

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

1997

[5] [5]

Jiawei Chen, Hande Dong, Xiang Wang, Fuli Feng, Meng Wang, and Xiangnan He. 2023. Bias and debias in recommender system: A survey and future directions. ACM Transactions on Information Systems 41, 3 (2023), 1–39

2023

[6] [6]

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

[7] [7]

Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. 2017. Deep reinforcement learning from human preferences. Advances in neural information processing systems 30 (2017)

2017

[8] [8]

Alexandre Défossez, Léon Bottou, Francis Bach, and Nicolas Usunier. 2020. A simple convergence proof of adam and adagrad. arXiv preprint arXiv:2003.02395 (2020)

arXiv 2020