pith. sign in

arxiv: 2510.27157 · v2 · submitted 2025-10-31 · 💻 cs.IR

A Survey on Generative Recommendation: Data, Model, and Tasks

Min Hou , Le Wu , Yuxin Liao , Yonghui Yang , Zhen Zhang , Yu Wang , Changlong Zheng , Han Wu
show 1 more author
Richang Hong
This is my paper

Pith reviewed 2026-05-18 03:44 UTC · model grok-4.3

classification 💻 cs.IR
keywords generative recommendationlarge language modelsdiffusion modelsrecommender systemsdata augmentationmodel alignmentconversational recommendationpersonalized generation
0
0 comments X

The pith

Generative recommendation reframes user-item matching as a generation task instead of scoring.

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

This survey organizes recent work showing how large language models and diffusion models change recommender systems from ranking candidates to producing outputs directly. The authors group the approaches around three operational stages that cover preparing data, aligning and training models, and defining what the system should generate at runtime. They point out five concrete benefits that follow from this shift, such as pulling in outside knowledge and following scaling patterns as models grow. The survey also flags practical obstacles in testing these systems and running them efficiently. Overall the work maps a path toward recommendation systems that can converse, explain choices, and create new content tailored to each user.

Core claim

The paper states that generative recommendation reconceptualizes the core matching problem as a generation task rather than discriminative scoring. It supplies a unified tripartite framework across data, model, and task dimensions and decomposes the literature into the stages of data augmentation and unification, model alignment and training, and task formulation and execution. At each stage the authors catalog techniques such as knowledge-infused augmentation, agent-based simulation, LLM alignment methods, and new task formats that support conversational interaction, explainable reasoning, and personalized content generation. They identify five resulting advantages: world knowledge, natural

What carries the argument

The tripartite framework of data augmentation and unification, model alignment and training, and task formulation and execution that organizes LLM-based methods, large recommendation models, and diffusion approaches.

Load-bearing premise

The existing literature on generative recommendation can be fully and cleanly decomposed into the stages of data augmentation and unification, model alignment and training, and task formulation and execution without major omissions or overlaps.

What would settle it

A later review that identifies a sizable set of generative recommendation papers whose methods do not fit into the proposed data-model-task stages or that fail to demonstrate the five listed advantages.

Figures

Figures reproduced from arXiv: 2510.27157 by Changlong Zheng, Han Wu, Le Wu, Min Hou, Richang Hong, Yonghui Yang, Yu Wang, Yuxin Liao, Zhen Zhang.

Figure 1
Figure 1. Figure 1: Discriminative Recommendation and Generative Recommendation innovation, where generative inference and probabilistic reasoning became integral parts of recommendation archi￾tectures. More recently, task-level generation has emerged, extending generative capabilities to high-level applications such as personalized content creation, multi-modal recom￾mendation, and explainable reasoning. Together, these thre… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of this survey. interaction-driven generative models; the use of LLM and textual data for natural language recommendation; and the integration of multimodal models for generating and pro￾cessing images/videos in RS. Liu et al. [106] explored the advancements in multimodal pretraining, adaptation, and generation techniques, as well as their applications to recom￾mender systems. Li et al. [80] revie… view at source ↗
Figure 3
Figure 3. Figure 3: Taxonomy of research on generative recommendation. proposing directions for advancing this emerging field. Fi￾nally, Section 7 concludes the survey by summarizing the contributions of generative models to RSs. 2. Preliminary and Background In this section, we begin with a foundational overview of traditional discriminative and generative recommendation models. We then explore the potential benefits that ge… view at source ↗
Figure 4
Figure 4. Figure 4: Outline of key techniques in LLM-empowered data generation. Behavior Augmentation. Data sparsity and cold-start challenges are key constraints limiting recommendation sys￾tems, and LLMs offer opportunities to address these chal￾lenges. Through appropriate prompting, LLMs can under￾stand user behavior and generate context for content of inter￾est to users. Thus, we can leverage the reasoning and general￾iza… view at source ↗
Figure 5
Figure 5. Figure 5: LLM empowered data unification 3.2.3. Multi-Modal Data Unification Recommendation involves multiple modalities like text, images, and behavior logs. Traditional methods fuse these separately, but large vision-language models (LVLMs) now enable unified multimodal representations, improving accu￾racy and interpretability. UniMP [184] and MQL4GRec [217] unify multimodal inputs into shared semantic spaces, boo… view at source ↗
Figure 6
Figure 6. Figure 6: The paradigms of aligning LLMs for recommendation. Inspired by the figure in [214]. (2) Large Recommendation Models, and (3) Diffusion￾Based Generative Recommendation. 4.1. LLM-Based Generative Recommendation With the rapid progress of LLMs, applying them to rec￾ommendation has become a major research line. LLM-based generative recommendation leverages pretrained LLMs to produce personalized suggestions vi… view at source ↗
Figure 7
Figure 7. Figure 7: Illustration of two research directions of large recommendation models. (a) The Architecture of LRMs, and (b) End-to￾End Recommendation [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Illustration of two types of diffusion-based generative recommendation. (a) Augmented Data Generation, and (b) Target Item Generation. item to explore the underlying distribution of item space, and generate recommended items directly, thereby eliminating the need for negative sampling and enabling exploration of the entire item space. DiffRIS[129]uses both local and global implicit features from user histo… view at source ↗
Figure 9
Figure 9. Figure 9: Illustration of traditional discriminative recommendation and generative recommendation assistant. that restrict a candidate set, generative models are capable of generating recommendations directly. In summary, large generative models bring stronger capabilities to the model side. On the task side, traditional RSs were often designed to perform a single task, such as CTR prediction or rat￾ing prediction. … view at source ↗
read the original abstract

Recommender systems serve as foundational infrastructure in modern information ecosystems, helping users navigate digital content and discover items aligned with their preferences. At their core, recommender systems address a fundamental problem: matching users with items. Over the past decades, the field has experienced successive paradigm shifts, from collaborative filtering and matrix factorization in the machine learning era to neural architectures in the deep learning era. Recently, the emergence of generative models, especially large language models (LLMs) and diffusion models, have sparked a new paradigm: generative recommendation, which reconceptualizes recommendation as a generation task rather than discriminative scoring. This survey provides a comprehensive examination through a unified tripartite framework spanning data, model, and task dimensions. Rather than simply categorizing works, we systematically decompose approaches into operational stages-data augmentation and unification, model alignment and training, task formulation and execution. At the data level, generative models enable knowledge-infused augmentation and agent-based simulation while unifying heterogeneous signals. At the model level, we taxonomize LLM-based methods, large recommendation models, and diffusion approaches, analyzing their alignment mechanisms and innovations. At the task level, we illuminate new capabilities including conversational interaction, explainable reasoning, and personalized content generation. We identify five key advantages: world knowledge integration, natural language understanding, reasoning capabilities, scaling laws, and creative generation. We critically examine challenges in benchmark design, model robustness, and deployment efficiency, while charting a roadmap toward intelligent recommendation assistants that fundamentally reshape human-information interaction.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. This survey paper examines generative recommendation as an emerging paradigm that reconceptualizes recommendation as a generation task using models such as LLMs and diffusion models, rather than traditional discriminative scoring. It organizes the literature via a unified tripartite framework that decomposes approaches into operational stages of data augmentation and unification, model alignment and training (covering LLM-based methods, large recommendation models, and diffusion approaches), and task formulation and execution (including conversational interaction, explainable reasoning, and personalized content generation). The paper identifies five key advantages—world knowledge integration, natural language understanding, reasoning capabilities, scaling laws, and creative generation—while critically discussing challenges in benchmark design, model robustness, and deployment efficiency, and outlining a roadmap toward intelligent recommendation assistants.

Significance. If the tripartite framework proves comprehensive without major omissions or forced categorizations, the survey could provide a valuable organizing lens for an emerging subfield, helping researchers map the shift from neural recommender systems to generative ones. By taxonomizing methods across data, model, and task dimensions and explicitly naming advantages and open challenges, it may accelerate identification of research gaps in areas like agent-based simulation and scaling laws for recommendation.

major comments (2)
  1. [Abstract / tripartite framework] Abstract and framework description: the central claim that the literature can be systematically decomposed into data augmentation/unification, model alignment/training, and task formulation/execution without major overlaps or omissions is load-bearing for the survey's utility; the manuscript should add an explicit discussion (perhaps in a dedicated taxonomy subsection) of boundary cases, such as works that span data unification and task execution, to demonstrate the framework's robustness.
  2. [Advantages discussion] Five key advantages section: the advantages (world knowledge integration, natural language understanding, reasoning capabilities, scaling laws, creative generation) are presented as distinguishing features of the paradigm shift, but each should be tied to at least two concrete cited works with brief evidence of the claimed benefit to avoid appearing as high-level assertions.
minor comments (3)
  1. [Model level] Add a summary table in the model-level section that cross-references the three model categories (LLM-based, large recommendation models, diffusion) against alignment mechanisms and representative papers for improved readability.
  2. [Challenges] The challenges section on benchmark design would benefit from citing specific existing benchmarks in generative recommendation and explicitly noting which ones fail to evaluate the claimed advantages such as creative generation.
  3. [Introduction / Conclusion] Ensure consistent use of terminology (e.g., 'generative recommendation' vs. 'generative models for recommendation') throughout the introduction and conclusion to prevent minor reader confusion.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and positive overall assessment of our survey on generative recommendation. The suggestions regarding the tripartite framework and the advantages section are helpful for strengthening the manuscript's clarity and rigor. We address each major comment below and will incorporate the revisions in the next version.

read point-by-point responses

  1. Referee: [Abstract / tripartite framework] Abstract and framework description: the central claim that the literature can be systematically decomposed into data augmentation/unification, model alignment/training, and task formulation/execution without major overlaps or omissions is load-bearing for the survey's utility; the manuscript should add an explicit discussion (perhaps in a dedicated taxonomy subsection) of boundary cases, such as works that span data unification and task execution, to demonstrate the framework's robustness.

    Authors: We agree that an explicit discussion of boundary cases would better demonstrate the framework's robustness and address potential overlaps or ambiguities. In the revised manuscript, we will add a dedicated subsection (or expanded paragraph) within the taxonomy discussion that analyzes boundary cases, including examples of works spanning data unification and task execution. This will explain how such works are accommodated in the tripartite structure, any necessary clarifications, and why the decomposition remains systematic without major omissions or forced categorizations. revision: yes

  2. Referee: [Advantages discussion] Five key advantages section: the advantages (world knowledge integration, natural language understanding, reasoning capabilities, scaling laws, creative generation) are presented as distinguishing features of the paradigm shift, but each should be tied to at least two concrete cited works with brief evidence of the claimed benefit to avoid appearing as high-level assertions.

    Authors: We appreciate this observation. While the advantages are drawn from patterns across the surveyed literature, we acknowledge that grounding them with specific citations would make the claims more concrete and less high-level. In the revision, we will expand the five key advantages section to tie each advantage to at least two concrete cited works, including brief evidence of the claimed benefit drawn from those works (e.g., empirical results or qualitative demonstrations in the original papers). This will substantiate the discussion without altering the overall structure or identified advantages. revision: yes

Circularity Check

0 steps flagged

No significant circularity: descriptive survey with no derivations or fitted predictions

full rationale

This is a literature survey paper that organizes prior work on generative recommendation into a tripartite framework (data augmentation/unification, model alignment/training, task formulation/execution) without presenting any original mathematical derivations, equations, predictions, or parameter-fitting procedures. All claims about advantages and paradigms rest on citations to external prior literature rather than self-referential reductions or self-citation chains that bear the central load. The structure is an organizing lens for an emerging field and introduces no self-definitional, fitted-input, or ansatz-smuggling circularities.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The survey rests on standard domain assumptions about recommender systems and generative models without introducing free parameters or new entities; the framework is presented as a useful lens rather than a derived necessity.

axioms (1)
  • domain assumption Recommender systems address a fundamental problem of matching users with items and have undergone paradigm shifts from collaborative filtering to neural architectures.
    Invoked in the opening of the abstract as the historical and conceptual foundation for introducing generative recommendation.

pith-pipeline@v0.9.0 · 5816 in / 1238 out tokens · 40419 ms · 2026-05-18T03:44:32.545785+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 7 Pith papers

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

  1. On the Equivalence Between Auto-Regressive Next Token Prediction and Full-Item-Vocabulary Maximum Likelihood Estimation in Generative Recommendation--A Short Note

    cs.IR 2026-04 accept novelty 7.0

    Auto-regressive next-token prediction is strictly equivalent to full-vocabulary maximum likelihood estimation in generative recommendation under bijective item-to-token-sequence mapping.

  2. Tencent Advertising Algorithm Challenge 2025: All-Modality Generative Recommendation

    cs.IR 2026-04 accept novelty 7.0

    Releases TencentGR-1M and TencentGR-10M datasets with baselines for all-modality generative recommendation in advertising, including weighted evaluation for conversions.

  3. Differentiable Semantic ID for Generative Recommendation

    cs.IR 2026-01 unverdicted novelty 7.0

    DIGER makes semantic IDs in generative recommendation differentiable via Gumbel noise and decay schedules, yielding consistent gains on public datasets by aligning indexing and recommendation losses.

  4. A Cascaded Generative Approach for e-Commerce Recommendations

    cs.AI 2026-05 unverdicted novelty 6.0

    A cascaded generative merchandising framework with placement theme generation, constrained keyword generation, and teacher-student fine-tuning achieves a 2.7% lift in cart adds per page view over a strong baseline in ...

  5. RoTE: Coarse-to-Fine Multi-Level Rotary Time Embedding for Sequential Recommendation

    cs.IR 2026-04 unverdicted novelty 6.0

    RoTE is a multi-level rotary time embedding module that explicitly models time spans in sequential recommendation and improves NDCG@5 by up to 20.11% when added to standard backbones on public benchmarks.

  6. How Personal Characteristics Shape User Exploration of Diverse Movie Recommendations with a LLM-Based Multi-Agent System

    cs.HC 2026-04 unverdicted novelty 5.0

    A multi-agent LLM recommender boosts perceived novelty and diversity in movie suggestions, with effects shaped by user conscientiousness, extraversion, GenAI experience, and skepticism.

  7. A Cascaded Generative Approach for e-Commerce Recommendations

    cs.AI 2026-05 unverdicted novelty 4.0

    A cascaded generative system for e-commerce recommendations using theme and keyword generation with teacher-student fine-tuning achieves a 2.7% lift in cart adds per page view.

Reference graph

Works this paper leans on

242 extracted references · 242 canonical work pages · cited by 6 Pith papers · 4 internal anchors

  1. [1]

    Ao, X., Wang, X., Luo, L., Qiao, Y., He, Q., Xie, X., 2021. Pens: A dataset and generic framework for personalized news headline generation, in: Proceedings of the 59th Annual Meeting of the As- sociation for Computational Linguistics and the 11th International JointConferenceonNaturalLanguageProcessing(Volume1:Long Papers), pp. 82–92

    work page 2021

  2. [2]

    Bai, T., Huang, L., Yu, Y., Yang, C., Hou, C., Zhao, Z., Shi, C.,

    work page

  3. [3]

    ACM Transactions on Information Systems

    Efficientmulti-taskprompttuningforrecommendation. ACM Transactions on Information Systems

    work page

  4. [4]

    Abi-stepgroundingparadigmforlarge language models in recommendation systems

    Bao, K., Zhang, J., Wang, W., Zhang, Y., Yang, Z., Luo, Y., Chen, C.,Feng,F.,Tian,Q.,2025. Abi-stepgroundingparadigmforlarge language models in recommendation systems. ACM Transactions on Recommender Systems 3, 1–27

    work page 2025

  5. [5]

    Tallrec: An effective and efficient tuning framework to align large language model with recommendation, in: Proceedings of the 17th ACM Conference on Recommender Systems, pp

    Bao, K., Zhang, J., Zhang, Y., Wang, W., Feng, F., He, X., 2023. Tallrec: An effective and efficient tuning framework to align large language model with recommendation, in: Proceedings of the 17th ACM Conference on Recommender Systems, pp. 1007–1014

    work page 2023

  6. [6]

    Simuser:Simulatinguserbehavior with large language models for recommender system evaluation

    Bougie,N.,Watanabe,N.,2025. Simuser:Simulatinguserbehavior with large language models for recommender system evaluation. arXiv preprint arXiv:2504.12722

    work page arXiv 2025

  7. [7]

    Generating user-engaging news headlines, in: Rogers, A., Boyd-Graber, J., Okazaki, N

    Cai, P., Song, K., Cho, S., Wang, H., Wang, X., Yu, H., Liu, F., Yu, D., 2023. Generating user-engaging news headlines, in: Rogers, A., Boyd-Graber, J., Okazaki, N. (Eds.), Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Association for Computational Linguis- tics,Toronto,Canada.pp.3265–3280...

    work page doi:10.18653/v1/2023.acl-long.183 2023

  8. [8]

    Cai,Z.,Wang,S.,Chu,V.W.,Naseem,U.,Wang,Y.,Chen,F.,2025. Unleashing the potential of diffusion models towards diversified sequential recommendations, in: Proceedings of the 48th Interna- tional ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 1476–1486

    work page 2025

  9. [9]

    Llms generate structurally realistic social networks but overestimate political homophily, in: Proceedings of the International AAAI Conference on Web and Social Media, pp

    Chang, S., Chaszczewicz, A., Wang, E., Josifovska, M., Pierson, E., Leskovec, J., 2025. Llms generate structurally realistic social networks but overestimate political homophily, in: Proceedings of the International AAAI Conference on Web and Social Media, pp. 341–371

    work page 2025

  10. [10]

    Hllm: Enhancing sequential recom- mendations via hierarchical large language models for item and user modeling.arXiv preprint arXiv:2409.12740, 2024

    Chen, J., Chi, L., Peng, B., Yuan, Z., 2024a. Hllm: Enhancing sequential recommendations via hierarchical large language models for item and user modeling. arXiv preprint arXiv:2409.12740

    work page arXiv

  11. [11]

    Chen, J., He, J., Li, H., Wang, S., Cao, Y., Wei, K., Yang, Z., Ji, Y., 2025a. Hierarchical intent-guided optimization with pluggable llm- driven semantics for session-based recommendation, in: Proceed- ings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 1655–1665

    work page

  12. [12]

    Chen, J., Xu, Y., Jiang, Y., 2025b. Unlocking the power of diffu- sion models in sequential recommendation: A simple and effective Min Hou et al.:Preprint submitted to ElsevierPage 22 of 30 A Survey on Generative Recommendation approach,in:Proceedingsofthe31stACMSIGKDDConferenceon Knowledge Discovery and Data Mining, pp. 155–166

    work page

  13. [13]

    Chen,J.,Yang,X.,Yang,C.,Bao,J.,Guo,Z.,Li,Y.,Shi,C.,2025c. Corona: A coarse-to-fine framework for graph-based recommenda- tion with large language models, in: Proceedings of the 48th Inter- national ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 2048–2058

    work page 2048

  14. [14]

    Cp-rec: Contextual prompting for conversational recommender systems

    Chen, K., Sun, S., 2023. Cp-rec: Contextual prompting for conversational recommender systems. Proceedings of the AAAI Conference on Artificial Intelligence 37, 12635– 12643. URL:https://ojs.aaai.org/index.php/AAAI/article/view/ 26487, doi:10.1609/aaai.v37i11.26487

    work page doi:10.1609/aaai.v37i11.26487 2023

  15. [15]

    Enhancing id-based recommendation with large language models

    Chen,L.,Gao,C.,Du,X.,Luo,H.,Jin,D.,Li,Y.,Wang,M.,2025d. Enhancing id-based recommendation with large language models. ACM Transactions on Information Systems 43, 1–30

    work page

  16. [16]

    Enhancing item tokenization for generative recommendation through self-improvement

    Chen, R., Ju, M., Bui, N., Antypas, D., Cai, S., Wu, X., Neves, L., Wang, Z., Shah, N., Zhao, T., 2024b. Enhancing item tokenization for generative recommendation through self-improvement. arXiv preprint arXiv:2412.17171

    work page arXiv

  17. [17]

    Cheng, W., Qin, Z., Wu, Z., Zhou, P., Huang, T., 2025. Large languagemodelsenhancedhyperbolicspacerecommendersystems, in: Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 1944– 1953

    work page 2025

  18. [18]

    Leveraging large language models for pre-trained recommender systems

    Chu, Z., Hao, H., Ouyang, X., Wang, S., Wang, Y., Shen, Y., Gu, J., Cui, Q., Li, L., Xue, S., Zhang, J.Y., Li, S., 2023. Leveraging large language models for pre-trained recommender systems. URL: https://arxiv.org/abs/2308.10837,arXiv:2308.10837

    work page arXiv 2023

  19. [19]

    Suber: An rl environment with simulated human behavior for recommender systems

    Corecco, N., Piatti, G., Lanzendörfer, L.A., Fan, F.X., Wattenhofer, R., 2024. Suber: An rl environment with simulated human behavior for recommender systems. arXiv preprint arXiv:2406.01631

    work page arXiv 2024

  20. [20]

    Distillation matters: empowering sequential recommenders to match the performance of large language models, in: Proceedings of the 18th ACM Conference on Recommender Systems, pp

    Cui, Y., Liu, F., Wang, P., Wang, B., Tang, H., Wan, Y., Wang, J., Chen, J., 2024. Distillation matters: empowering sequential recommenders to match the performance of large language models, in: Proceedings of the 18th ACM Conference on Recommender Systems, pp. 507–517

    work page 2024

  21. [21]

    M6-rec: Gen- erative pretrained language models are open-ended recommender systems

    Cui, Z., Ma, J., Zhou, C., Zhou, J., Yang, H., 2022. M6-rec: Gen- erative pretrained language models are open-ended recommender systems. arXiv preprint arXiv:2205.08084

    work page arXiv 2022

  22. [22]

    Dao, H., Deng, Y., Le, D.D., Liao, L., 2024. Broadening the view:Demonstration-augmentedpromptlearningforconversational recommendation, in: Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, Association for Computing Machinery, New York, NY, USA. p. 785–795. URL:https://doi.org/10.1145/3626772....

    work page doi:10.1145/3626772.3657755 2024

  23. [23]

    Areviewofmodernrecommendersystemsusinggenerativemodels (gen-recsys),in:Proceedingsofthe30thACMSIGKDDconference on Knowledge Discovery and Data Mining, pp

    Deldjoo, Y., He, Z., McAuley, J., Korikov, A., Sanner, S., Ramisa, A., Vidal, R., Sathiamoorthy, M., Kasirzadeh, A., Milano, S., 2024. Areviewofmodernrecommendersystemsusinggenerativemodels (gen-recsys),in:Proceedingsofthe30thACMSIGKDDconference on Knowledge Discovery and Data Mining, pp. 6448–6458

    work page 2024

  24. [24]

    OneRec: Unifying Retrieve and Rank with Generative Recommender and Iterative Preference Alignment

    Deng, J., Wang, S., Cai, K., Ren, L., Hu, Q., Ding, W., Luo, Q., Zhou, G., 2025. Onerec: Unifying retrieve and rank with generative recommender and iterative preference alignment. arXiv preprint arXiv:2502.18965

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  25. [25]

    Federated recommender system based on diffusion augmentation and guided denoising

    Di, Y., Shi, H., Wang, X., Ma, R., Liu, Y., 2025. Federated recommender system based on diffusion augmentation and guided denoising. ACM Transactions on Information Systems 43, 1–36

    work page 2025

  26. [26]

    ACM Transactions on Information Systems 43, 1–26

    Dong,Z.,Hu,L.,Chen,J.,Wang,Z.,Wu,F.,2025.Comprehendthen predict:Promptinglargelanguagemodelsforrecommendationwith semantic and collaborative data. ACM Transactions on Information Systems 43, 1–26

    work page 2025

  27. [27]

    Enhancing job recommendation through llm- basedgenerativeadversarialnetworks,in:ProceedingsoftheAAAI conference on artificial intelligence, pp

    Du, Y., Luo, D., Yan, R., Wang, X., Liu, H., Zhu, H., Song, Y., Zhang, J., 2024. Enhancing job recommendation through llm- basedgenerativeadversarialnetworks,in:ProceedingsoftheAAAI conference on artificial intelligence, pp. 8363–8371

    work page 2024

  28. [28]

    Lusifer: Llm-based user simulated feedback environment for online recommender systems

    Ebrat, D., Paradalis, E., Rueda, L., 2024. Lusifer: Llm-based user simulated feedback environment for online recommender systems. arXiv preprint arXiv:2405.13362

    work page arXiv 2024

  29. [31]

    Reason4rec: Large language models for recommendation with deliberative user preference alignment

    Fang, Y., Wang, W., Zhang, Y., Zhu, F., Wang, Q., Feng, F., He, X., 2025c. Reason4rec: Large language models for recommendation with deliberative user preference alignment. URL:https://arxiv. org/abs/2502.02061,arXiv:2502.02061

    work page arXiv

  30. [32]

    A unified framework for multi-domain ctr prediction via large language models

    Fu, Z., Li, X., Wu, C., Wang, Y., Dong, K., Zhao, X., Zhao, M., Guo, H., Tang, R., 2025. A unified framework for multi-domain ctr prediction via large language models. ACM Transactions on Information Systems

    work page 2025

  31. [33]

    5075–5084

    Gao,C.,Chen,R.,Yuan,S.,Huang,K.,Yu,Y.,He,X.,2025a.Sprec: Self-play to debias llm-based recommendation, in: Proceedings of the ACM on Web Conference 2025, pp. 5075–5084

    work page 2025

  32. [34]

    Gao,C.,Gao,M.,Fan,C.,Yuan,S.,Shi,W.,He,X.,2025b. Process- supervised llm recommenders via flow-guided tuning, in: Proceed- ings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 1934–1943

    work page 1934

  33. [35]

    S$^3$: Social-network Simulation System with Large Language Model-Empowered Agents

    Gao, C., Lan, X., Lu, Z., Mao, J., Piao, J., Wang, H., Jin, D., Li, Y., 2023a. S3: Social-network simulation system with large language model-empowered agents. arXiv preprint arXiv:2307.14984

    work page internal anchor Pith review Pith/arXiv arXiv

  34. [36]

    Llm4rerank: Llm-based auto-reranking framework for recommendations, in: Proceedings of the ACM on Web Conference 2025, pp

    Gao, J., Chen, B., Zhao, X., Liu, W., Li, X., Wang, Y., Wang, W., Guo, H., Tang, R., 2025c. Llm4rerank: Llm-based auto-reranking framework for recommendations, in: Proceedings of the ACM on Web Conference 2025, pp. 228–239

    work page 2025

  35. [38]

    Chat- rec: Towards interactive and explainable llms-augmented recommender system,

    Gao,Y.,Sheng,T.,Xiang,Y.,Xiong,Y.,Wang,H.,Zhang,J.,2023c. Chat-rec: Towards interactive and explainable llms-augmented rec- ommender system. arXiv preprint arXiv:2303.14524

    work page arXiv

  36. [39]

    Geng, S., Liu, S., Fu, Z., Ge, Y., Zhang, Y., 2022. Recommenda- tion as language processing (rlp): A unified pretrain, personalized prompt & predict paradigm (p5), in: Proceedings of the 16th ACM Conference on Recommender Systems, pp. 299–315

    work page 2022

  37. [40]

    Shilling attacks against recommender systems: a comprehensive survey

    Gunes, I., Kaleli, C., Bilge, A., Polat, H., 2014. Shilling attacks against recommender systems: a comprehensive survey. Artificial Intelligence Review 42, 767–799

    work page 2014

  38. [41]

    doi: 10.1038/s41586-025-09422-z

    Guo, D., Yang, D., Zhang, H., Song, J., Zhang, R., Xu, R., Zhu, Q., Ma, S., Wang, P., Liang, W., et al., 2025a. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learn- ing. Nature645,633–638. URL:https://www.nature.com/articles/ s41586-025-09422-z, doi:10.1038/s41586-025-09422-z

    work page doi:10.1038/s41586-025-09422-z

  39. [42]

    Request-only optimization for recommendation systems

    Guo, L., Li, W., Liao, L., Cheng, H., Zhang, R., Shi, Y., Wang, Y., Huang, Y., Zhai, K., Wang, P., Shi, T., Cao, X., Wang, S., Cai, R., Gong, Z., Vichare, O., Jian, R., Gao, L., Deng, S., Liu, X., Zhang, X., Li, F., Xie, W., Wen, B., Li, R., Fang, L., Liu, X., Zhai, J., 2025b. Request-only optimization for recommendation systems. URL:https://arxiv.org/abs...

    work page arXiv

  40. [43]

    Semantic-enhanced co-attention prompt learning for non- overlapping cross-domain recommendation

    Guo, L., Song, C., Guo, F., Han, X., Chang, X., Zhu, L., 2025c. Semantic-enhanced co-attention prompt learning for non- overlapping cross-domain recommendation. ACM Transactions on Information Systems

    work page

  41. [44]

    Onesug: The unified end-to-end generative framework for e-commerce query suggestion

    Guo, X., Chen, B., Wang, S., Yang, Y., Lei, C., Ding, Y., Li, H., 2025d. Onesug: The unified end-to-end generative framework for e-commerce query suggestion. arXiv preprint arXiv:2506.06913

    work page arXiv

  42. [45]

    Han, R., Li, Q., Jiang, H., Li, R., Zhao, Y., Li, X., Lin, W., 2024. Enhancing ctr prediction through sequential recommendation pre- training: Introducing the srp4ctr framework, in: Proceedings of the 33rdACMInternationalConferenceonInformationandKnowledge Management, pp. 3777–3781. Min Hou et al.:Preprint submitted to ElsevierPage 23 of 30 A Survey on Ge...

    work page 2024

  43. [47]

    Mtgr: Industrial-scale generative recommendation framework in meituan

    Han, R., Yin, B., Chen, S., Jiang, H., Jiang, F., Li, X., Ma, C., Huang, M., Li, X., Jing, C., et al., 2025b. Mtgr: Industrial-scale generative recommendation framework in meituan. arXiv preprint arXiv:2505.18654

    work page arXiv

  44. [48]

    Zero-shotrecom- mendations with pre-trained large language models for multimodal nudging, in: 2023 IEEE International Conference on Data Mining Workshops (ICDMW), IEEE

    Harrison,R.M.,Dereventsov,A.,Bibin,A.,2023. Zero-shotrecom- mendations with pre-trained large language models for multimodal nudging, in: 2023 IEEE International Conference on Data Mining Workshops (ICDMW), IEEE. pp. 1535–1542

    work page 2023

  45. [49]

    Llm2rec: Largelanguagemodelsarepowerfulembeddingmodelsforsequen- tial recommendation, in: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp

    He, Y., Liu, X., Zhang, A., Ma, Y., Chua, T.S., 2025. Llm2rec: Largelanguagemodelsarepowerfulembeddingmodelsforsequen- tial recommendation, in: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 896– 907

    work page 2025

  46. [50]

    He, Z., Xie, Z., Jha, R., Steck, H., Liang, D., Feng, Y., Majumder, B.P., Kallus, N., Mcauley, J., 2023. Large language models as zero-shotconversationalrecommenders,in:Proceedingsofthe32nd ACM International Conference on Information and Knowledge Management, Association for Computing Machinery, New York, NY, USA. p. 720–730. URL:https://doi.org/10.1145/3...

    work page doi:10.1145/3583780 2023

  47. [51]

    Query-aware sequential recommendation, in: Proceedings of the 31st ACM International Conference on Information & Knowledge Management, pp

    He, Z., Zhao, H., Wang, Z., Lin, Z., Kale, A., Mcauley, J., 2022. Query-aware sequential recommendation, in: Proceedings of the 31st ACM International Conference on Information & Knowledge Management, pp. 4019–4023

    work page 2022

  48. [52]

    Molorec:Ageneralizableandefficientframework for llm-based recommendation

    Hou, M., Bai, C., Wu, L., Liu, H., Zhang, K., Zhang, K., Hong, R., Wang,M.,2025a. Molorec:Ageneralizableandefficientframework for llm-based recommendation. arXiv preprint arXiv:2502.08271

    work page arXiv

  49. [53]

    Generating long semantic ids in parallel for recommendation, in: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp

    Hou,Y.,Li,J.,Shin,A.,Jeon,J.,Santhanam,A.,Shao,W.,Hassani, K., Yao, N., McAuley, J., 2025b. Generating long semantic ids in parallel for recommendation, in: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 956–966

    work page

  50. [54]

    Actionpiece: Contextually tok- enizing action sequences for generative recommendation, in: Forty- second International Conference on Machine Learning

    Hou, Y., Ni, J., He, Z., Sachdeva, N., Kang, W.C., Chi, E.H., McAuley, J., Cheng, D.Z., 2025c. Actionpiece: Contextually tok- enizing action sequences for generative recommendation, in: Forty- second International Conference on Machine Learning

    work page

  51. [55]

    Large language models are zero-shot rankers for recommender systems, in: European Conference on Information Retrieval, Springer

    Hou, Y., Zhang, J., Lin, Z., Lu, H., Xie, R., McAuley, J., Zhao, W.X., 2024. Large language models are zero-shot rankers for recommender systems, in: European Conference on Information Retrieval, Springer. pp. 364–381

    work page 2024

  52. [56]

    Up5: Unbiased foun- dation model for fairness-aware recommendation

    Hua,W.,Ge,Y.,Xu,S.,Ji,J.,Zhang,Y.,2023a.Up5:Unbiasedfoun- dation model for fairness-aware recommendation. arXiv preprint arXiv:2305.12090

    work page arXiv

  53. [57]

    Hua, W., Xu, S., Ge, Y., Zhang, Y., 2023b. How to index item ids for recommendation foundation models, in: Proceedings of the Annual International ACM SIGIR Conference on Research and Development in Information Retrieval in the Asia Pacific Region, Association for Computing Machinery, New York, NY, USA. p. 195–204. URL:https://doi.org/10.1145/3624918.3625...

    work page doi:10.1145/3624918.3625339

  54. [58]

    Large language model simulator for cold-startrecommendation,in:ProceedingsoftheEighteenthACM InternationalConferenceonWebSearchandDataMining,pp.261– 270

    Huang, F., Bei, Y., Yang, Z., Jiang, J., Chen, H., Shen, Q., Wang, S., Karray, F., Yu, P.S., 2025a. Large language model simulator for cold-startrecommendation,in:ProceedingsoftheEighteenthACM InternationalConferenceonWebSearchandDataMining,pp.261– 270

    work page

  55. [59]

    Large language model simulator for cold-startrecommendation,in:ProceedingsoftheEighteenthACM InternationalConferenceonWebSearchandDataMining,pp.261– 270

    Huang, F., Bei, Y., Yang, Z., Jiang, J., Chen, H., Shen, Q., Wang, S., Karray, F., Yu, P.S., 2025b. Large language model simulator for cold-startrecommendation,in:ProceedingsoftheEighteenthACM InternationalConferenceonWebSearchandDataMining,pp.261– 270

    work page

  56. [60]

    How to mitigate information loss in knowledge graphs for graphrag: Leveraging triple context restoration and query-driven feedback

    Huang, M., Bu, C., He, Y., Wu, X., 2025c. How to mitigate information loss in knowledge graphs for graphrag: Leveraging triple context restoration and query-driven feedback. arXiv preprint arXiv:2501.15378

    work page arXiv

  57. [61]

    Towards large-scale generative ranking.arXiv preprint arXiv:2505.04180, 2025

    Huang,Y.,Chen,Y.,Cao,X.,Yang,R.,Qi,M.,Zhu,Y.,Han,Q.,Liu, Y., Liu, Z., Yao, X., et al., 2025d. Towards large-scale generative ranking. arXiv preprint arXiv:2505.04180

    work page arXiv

  58. [62]

    Flow matching for denoised social recommendation, in: Forty-second International Conference on Machine Learning

    Huang, Y., Liang, K., Dong, Z., Qu, X., Tianxiang, W., Han, Y., Xu, J., Zhou, B., Wang, Y., 2025e. Flow matching for denoised social recommendation, in: Forty-second International Conference on Machine Learning

    work page

  59. [63]

    Learn:Knowledgeadaptationfrom large language model to recommendation for practical industrial application, in: Proceedings of the AAAI Conference on Artificial Intelligence, pp

    Jia, J., Wang, Y., Li, Y., Chen, H., Bai, X., Liu, Z., Liang, J., Chen, Q.,Li,H.,Jiang,P.,etal.,2025. Learn:Knowledgeadaptationfrom large language model to recommendation for practical industrial application, in: Proceedings of the AAAI Conference on Artificial Intelligence, pp. 11861–11869

    work page 2025

  60. [64]

    Large language model as universal retriever in industrial-scale recommender system

    Jiang, J., Huang, Y., Liu, B., Kong, X., Li, X., Xu, Z., Zhu, H., Xu, J., Zheng, B., 2025a. Large language model as universal retriever in industrial-scale recommender system. arXiv preprint arXiv:2502.03041

    work page arXiv

  61. [65]

    Item-sidefairnessoflargelanguagemodel-basedrecommen- dation system, in: Proceedings of the ACM Web Conference 2024, pp

    Jiang, M., Bao, K., Zhang, J., Wang, W., Yang, Z., Feng, F., He, X., 2024a. Item-sidefairnessoflargelanguagemodel-basedrecommen- dation system, in: Proceedings of the ACM Web Conference 2024, pp. 4717–4726

    work page 2024

  62. [66]

    Reclm: Recommendation instruction tuning, in: Proceedings of the 63rdAnnualMeetingoftheAssociationforComputationalLinguis- tics (Volume 1: Long Papers), p

    Jiang, Y., Yang, Y., Xia, L., Luo, D., Lin, K., Huang, C., 2025b. Reclm: Recommendation instruction tuning, in: Proceedings of the 63rdAnnualMeetingoftheAssociationforComputationalLinguis- tics (Volume 1: Long Papers), p. 15443–15459

    work page

  63. [67]

    Casevo: A cognitive agents and social evolution simulator

    Jiang, Z., Shi, Y., Li, M., Xiao, H., Qin, Y., Wei, Q., Wang, Y., Zhang, Y., 2024b. Casevo: A cognitive agents and social evolution simulator. arXiv preprint arXiv:2412.19498

    work page arXiv

  64. [68]

    arXiv preprint arXiv:2305.06474 , year=

    Kang,W.C.,Ni,J.,Mehta,N.,Sathiamoorthy,M.,Hong,L.,Chi,E., Cheng,D.Z.,2023. Dollmsunderstanduserpreferences?evaluating llms on user rating prediction. arXiv preprint arXiv:2305.06474

    work page arXiv 2023

  65. [69]

    Kim, J., Kim, H., Cho, H., Kang, S., Chang, B., Yeo, J., Lee, D., 2025a. Review-driven personalized preference reasoning with large language models for recommendation, in: Proceedings of the 48th International ACM SIGIR Conference on Research and Develop- ment in Information Retrieval, pp. 1697–1706

    work page

  66. [70]

    1395–1406

    Kim,S.,Kang,H.,Choi,S.,Kim,D.,Yang,M.,Park,C.,2024.Large language models meet collaborative filtering: An efficient all-round llm-based recommender system, in: Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 1395–1406

    work page 2024

  67. [71]

    Lost in sequence: Do large language models understand sequential recommendation? arXiv preprint arXiv:2502.13909

    Kim, S., Kang, H., Kim, K., Kim, J., Kim, D., Yang, M., Oh, K., McAuley, J., Park, C., 2025b. Lost in sequence: Do large language models understand sequential recommendation? arXiv preprint arXiv:2502.13909

    work page arXiv

  68. [72]

    Large language models are zero-shot reasoners, in: Proceedings of the36thInternationalConferenceonNeuralInformationProcessing Systems, Curran Associates Inc., Red Hook, NY, USA

    Kojima, T., Gu, S.S., Reid, M., Matsuo, Y., Iwasawa, Y., 2022. Large language models are zero-shot reasoners, in: Proceedings of the36thInternationalConferenceonNeuralInformationProcessing Systems, Curran Associates Inc., Red Hook, NY, USA

    work page 2022

  69. [73]

    Customizing language models with instance-wise lora for sequential recommendation

    Kong, X., Wu, J., Zhang, A., Sheng, L., Lin, H., Wang, X., He, X., 2024. Customizing language models with instance-wise lora for sequential recommendation. Advances in Neural Information Processing Systems 37, 113072–113095

    work page 2024

  70. [74]

    Matrix factorization tech- niques for recommender systems

    Koren, Y., Bell, R., Volinsky, C., 2009. Matrix factorization tech- niques for recommender systems. Computer 42, 30–37

    work page 2009

  71. [75]

    Uncertainty quan- tification and decomposition for llm-based recommendation, in: Proceedings of the ACMon Web Conference 2025, pp

    Kweon, W., Jang, S., Kang, S., Yu, H., 2025. Uncertainty quan- tification and decomposition for llm-based recommendation, in: Proceedings of the ACMon Web Conference 2025, pp. 4889–4901

    work page 2025

  72. [76]

    Li,H.,Shen,D.,Wang,C.,Liu,Y.,Gu,J.,2025a. Canllmsenhance fairnessinrecommendationsystems?adataaugmentationapproach, in:Proceedingsofthe48thInternationalACMSIGIRConferenceon Research and Development in Information Retrieval, pp. 570–580

    work page

  73. [77]

    Li, J., Li, Y., Shen, X., Zhang, C., Qi, G., Sheng, B., 2025b. Open-world attribute mining for e-commerce products with multi- modalself-correctioninstructiontuning,in:Proceedingsofthe63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 1702–1714

    work page

  74. [78]

    Li, J., Wang, S., Zhang, Q., Yu, S., Chen, F., 2025c. Generating with fairness: A modality-diffused counterfactual framework for Min Hou et al.:Preprint submitted to ElsevierPage 24 of 30 A Survey on Generative Recommendation incomplete multimodal recommendations, in: Proceedings of the ACM on Web Conference 2025, pp. 2787–2798

    work page 2025

  75. [79]

    Li, L., Zhang, Y., Chen, L., 2020. Generate neural template expla- nations for recommendation, in: Proceedings of the 29th ACM In- ternational Conference on Information & Knowledge Management, Association for Computing Machinery, New York, NY, USA. p. 755–764. URL:https://doi.org/10.1145/3340531.3411992, doi:10. 1145/3340531.3411992

    work page doi:10.1145/3340531.3411992 2020

  76. [80]

    Personalized transformer for explainable recommendation, in: ACL

    Li, L., Zhang, Y., Chen, L., 2021. Personalized transformer for explainable recommendation, in: ACL

    work page 2021

  77. [81]

    Large language models forgenerativerecommendation:Asurveyandvisionarydiscussions

    Li, L., Zhang, Y., Liu, D., Chen, L., 2023a. Large language models forgenerativerecommendation:Asurveyandvisionarydiscussions. arXiv preprint arXiv:2309.01157

    work page arXiv

  78. [82]

    Li, W., Huang, R., Zhao, H., Liu, C., Zheng, K., Liu, Q., Mou, N., Zhou, G., Lian, D., Song, Y., et al., 2025d. Dimerec: a unified framework for enhanced sequential recommendation via generative diffusion models, in: Proceedings of the Eighteenth ACM Interna- tional Conference on Web Search and Data Mining, pp. 726–734

    work page

  79. [83]

    E4srec: An elegant effective efficient extensible solution of large lan- guage models for sequential recommendation

    Li, X., Chen, C., Zhao, X., Zhang, Y., Xing, C., 2023b. E4srec: An elegant effective efficient extensible solution of large lan- guage models for sequential recommendation. arXiv preprint arXiv:2312.02443

    work page arXiv

  80. [84]

    Exploring preference-guided diffusion model for cross-domain recommendation, in: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp

    Li, X., Tang, H., Sheng, J., Zhang, X., Gao, L., Cheng, S., Yin, D., Liu, T., 2025e. Exploring preference-guided diffusion model for cross-domain recommendation, in: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 719–728

    work page

Showing first 80 references.