Pith. sign in

REVIEW 4 major objections 5 minor 36 references

Why Not Together? A Multiple-Round Recommender System for Queries and Items

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read MAGUS claims that recommending both artificial queries and items in a multi-round loop, initialized by any offline recommender and updated by label propagation over a word graph, lets users find a target item in fewer rounds than…

desk verdict The framework is clean and the task is new, but the simulated user answers exactly along the relation MAGUS propagates, so the reported gains are internal consistency, not real-user evidence. read the letter →

arxiv 2412.10787 v1 pith:RSTDH4UE submitted 2024-12-14 cs.IR

Jiarui Jin , Xianyu Chen , Weinan Zhang , Yong Yu , Jun Wang This is my paper · ORCID
classification cs.IR
keywords multi-roundrecommendationqueryitemrelationalgraphlabelpropagationconversationalrecommenderuserfeedbackcategoricalwordcombinations
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

MAGUS tries to establish that queries and items should be recommended together in a multi-round loop rather than treated as separate tasks. Its central claim is that any offline-tuned recommender can be wrapped in a graph-based framework that guesses user interests with artificial queries (word combinations) and items each round, and that this integrated loop lets users find a preferred item in fewer rounds than item-only or conversational baselines. If true, the practical payoff is large: existing recommender platforms could add query-plus-item interaction without retraining their core models, and users with vague preferences could iterate to a target item faster. The paper reports consistent gains across 12 base recommenders and three real-world e-commerce datasets.

What carries the argument

The load-bearing object is the relational graph G=(E,R), where E contains individual words and word combinations that correspond to queries or items, and R is split into mutual-improvement edges R+ (one node is part of another), mutual-inhibition edges R− (nodes from distinct items that compete), and mutual-independence edges R⊥ (compatible but non-overlapping). The mechanism that carries the argument is label propagation: initial item scores from the offline recommender spread along R+ and R− to score all word nodes, and after each user response the feedback is propagated back along the same edges, so the next recommendation reacts to the current session rather than only to history. Edge weights are either fixed (Equation 3) or learned from item embeddings (Equation 14), with the learned variant called MAGUS+.

What would settle it

In the paper's simulator, replace the deterministic containment-based user response with a noisy response model (e.g., Yes with probability 0.8 when query words appear in a target item, and 0.2 otherwise) or with queries sampled from real search logs; if SA@K and RA@K gains over item-only baselines disappear, the containment assumption is the load-bearing premise.

Watch

Extended reading notes

Core claim

On its own terms, the paper argues that queries and items are not two recommendation tasks but two granularities of the same user-interest representation: both are combinations of categorical words. MAGUS builds a relational graph whose nodes are individual words and word combinations, connects nodes by mutual-improvement, mutual-inhibition, and mutual-independence edges, initializes item-node scores with any offline recommender, and propagates user feedback through the graph each round. The formal objective is to minimize the round K at which the recommended action lands in the target item set. The paper's headline finding is that this query-plus-item loop significantly improves round-wise and session-wise accuracy over item-only and conversational baselines on all tested datasets.

Load-bearing premise

The load-bearing premise is that when a user says Yes to a suggested query, every word in that query is contained in an item they would accept, which is exactly the containment rule the graph and the score updates rely on.

Editorial extensions

If this is right

  • Any single-round recommender can be turned into a multi-round query-and-item recommender without redesigning the base model, since MAGUS only needs its item scores as initialization.
  • Users should reach a target item in fewer rounds when queries are recommended alongside items, because query feedback constrains the word graph before an item is hit.
  • More interaction rounds help more than longer recommendation lists, since each round lets the label propagation incorporate fresh feedback.
  • Learning edge weights from item embeddings (MAGUS+) adds a consistent gain on graph-based base recommenders.
  • The framework extends to ambiguous user feedback (Not Care responses) and still beats the conversational baselines, though absolute scores drop.

Reading between the lines

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

  • Because MAGUS treats queries as generated word combinations, its gains depend on the quality of the word vocabulary; building Q from real search logs instead of enumerating all attribute combinations is a natural stress test the paper does not run.
  • The containment-response user model turns the problem into a structured twenty-questions game; framing the query-selection rule as information gain over the graph could yield an even cheaper query policy than pure score ranking.
  • The same graph and propagation loop could be reused for conversational product search or voice assistants, where users answer attribute-style questions; the paper only sketches the LLM integration as a query-decoration step.
  • If real users answer queries more noisily than the simulator, the safe prediction is that the benefit persists only while item feedback remains available each round; a mixed policy that balances query and item suggestions would be the testable fix.
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 / 5 minor

Summary. The paper proposes MAGUS, a multiple-round recommendation framework that augments any offline-trained recommender by jointly scoring and recommending artificial queries and items. Queries and items are represented as nodes in a relational graph over words and word combinations; the base recommender initializes item scores, which are propagated over mutual-improvement/inhibition edges, and real-time feedback is incorporated via label propagation. The stated objective (Definition 2.1) is to minimize the number of rounds needed to surface an item in a target set. Experiments on three e-commerce datasets with 12 base recommenders report significant gains in round-wise accuracy (RA@3) and session-wise accuracy (SA@3, SA@5) over item-only and conversational baselines.

Significance. If the central efficiency claim were established, MAGUS would be a practically attractive plug-in: it is compatible with a wide range of base recommenders, has bounded online overhead, includes an offline weight-learning variant, and comes with a deployment-oriented architecture. The paper also provides a formal graph-based formulation, a complexity analysis, and a code link, which are strengths. However, the current evidence is generated by a simulator whose affirmative-response rule is identical to the containment relation encoded in the graph and used in score propagation; the reported gains therefore mostly validate internal consistency between the simulator's oracle and MAGUS's update rule. External validity for real user behavior remains to be demonstrated.

major comments (4)
  1. [Appendix D.1, Definition D.1; Section 3.1; Eq. (9)] The simulated user in Appendix D.1 answers 'Yes' to a query exactly when the query is an element of the word combination of a target item. This is the same containment relation that defines the R+ edge class in Definition 3.1 and that Eq. (9) uses to raise the scores of all items containing an affirmed query when weights are set by Eq. (3). Consequently, the reported SA@3/SA@5 and RA@3 improvements are largely a test of how well MAGUS walks the subset lattice defined by the oracle, not of whether real users' yes/no answers can be exploited in this way. Appendix E.1's 'Not Care' response does not remove this problem, because affirmative answers still use the subset rule. A concrete remedy would be an evaluation with a user model whose yes/no response depends on item-level utility or on real logged query-item click data, with the containment relation used only by the recommender and not by the oracle; the authors should show that gains persist when the oracle and the propagation rule are not aligned.
  2. [Appendix C.2; Eq. (5)] The query set Q is synthesized by enumerating all word combinations over item feature words; it is not drawn from real user search logs, and the optional historical-query set H_q of Eq. (5) plays no role in the experiments. Thus the system is evaluated on artificial queries constructed from the same vocabulary used to build the graph. This makes it difficult to conclude that MAGUS improves efficiency for realistic user-issued queries; the authors should evaluate on a test collection containing real queries, or at minimum provide a sensitivity analysis over query-generation distributions.
  3. [Section 4.2, paragraph 1; Table 1] The first finding in Section 4.2 compares SAC (single round, items only) with RA@3/SA@3/SA@5 (multiple rounds, queries and items). This comparison cannot be attributed to the addition of queries, because any multiple-round protocol with feedback has more chances to hit a target. The valid evidence for the contribution is the comparison with the RNN-augmented multiple-round baselines (X) and conversational baselines, but those comparisons inherit the same oracle-related concern as the first major comment.
  4. [Table 1; Appendix C.5] No error bars, standard deviations, or numbers of random sessions are reported for RA@3/SA@3/SA@5, and the * significance markers are not accompanied by a description of how the significance test was constructed or how many repeated runs were used. Since the simulated user is deterministic given the sampled session, the authors should report variance across session draws and across random seeds of the base recommenders.
minor comments (5)
  1. [Section 3.1] In the paragraph defining the relational graph, 'MAGUS commerces by extracting' should read 'MAGUS commences by extracting'.
  2. [Abstract] 'efficiency, with which users can identify' should be 'efficiency with which users can identify'.
  3. [Section 4.4] 'We also derive into a complexity study' should read 'We also report a complexity study'.
  4. [Figures 3 and 4] The axis labels in Figures 3 and 4 are not legible in the submitted version; please provide clean versions.
  5. [Appendix C.4, Figure 5 caption] The phrase 'in the masonry layout' appears to be an artifact and should be clarified or removed.

Circularity Check

1 steps flagged · score 6.0 of 10

MAGUS's efficiency gains are measured against a simulated user whose Yes/No rule is the same containment relation that defines R+ and drives Eq. (9) propagation; the evaluation oracle is the method's own relation.

  1. self definitional [Section 3.1 (R+ definition); Eq. (9); Appendix D.1 (Definition D.1); Appendix C.2]
    "R+ delineates connections between pairs of nodes in situations where one node is subsumed within or constitutes a component of the other node. ... Upon querying query q, the user should respond with Yes if q is one of the elements within the word combination of the target items, or No otherwise."

    The only evidence for the headline 'efficiency' claim is a simulator whose ground truth for a positive query answer is exactly the containment relation defining R+ (Section 3.1). When the user agent says Yes to q, the target item is by definition among the nodes containing q's words; Eq. (9) then raises b_y_v' for precisely those nodes via ←R+ propagation. RA@K/SA@K therefore measure how quickly MAGUS's own subset-lattice propagation reaches the target, not whether artificial queries carry real user interest. The paper concedes 'users always have a clear picture of their interests ... may not hold in practice,' and the Not Care variant still treats containment as the truth condition for affirmative answers.

full rationale

The paper's algorithmic derivation—graph construction, score propagation, and feature-propagation weight learning—is self-contained and not circular; it does not rest on a self-citation chain, and references [6] and [14] are methodological aids rather than load-bearing evidence for the central claim. The circularity is confined to the evaluation protocol for the multi-round efficiency claim. MAGUS's update rule Eq. (9) boosts items containing an affirmed query's words using the R+ containment relation, and the simulated user in Definition D.1 is defined to answer Yes to exactly those queries; the simulator's preference model is the same relation MAGUS exploits. Thus the reported SA@K and RA@K improvements over item-only and conversational baselines measure how well the system navigates a subset lattice whose oracle is the method's own relation. This is a genuine reduction-by-construction for the user-efficiency claim, though the framework itself has independent technical content and the baseline comparisons are internally consistent. For real users the claim remains untested rather than disproven; the ambiguity/Not Care experiments soften but do not remove the definitional alignment, since affirmative answers still follow the containment rule. A score of 6 reflects that the headline evaluation reduces to the method's own R+ relation while the surrounding system is not itself a tautology.

Assumptions & free parameters 2 free parameters · 4 assumptions · 1 invented entities

The central claim rests primarily on the simulated user's response rule (a domain assumption), the decomposability of queries and items into feature words, the sufficiency of three edge types, and a specific session-construction protocol. Fitted numbers include standard training hyperparameters and, for MAGUS+, feature-propagation edge weights trained with log-loss.

free parameters (2)
  • Feature propagation weights W1, W2 and node embeddings (MAGUS+) = trained by log-loss, Eq. (13)
    For the MAGUS+ variant, Eqs. (11)-(14) learn edge weights from H+ and H- with the same labels used to train the base recommender; these weights affect the reported MAGUS+ results.
  • Training hyperparameters (embedding dim, batch size, L2 weight, dropout, learning rate) = 64, 1000, 4e-4, 0.5, 1e-2 to 1e-5
    Stated in Appendix C.5 and chosen by hand; they are standard settings rather than the central scientific contribution, but the ledger lists them for completeness.
assumptions (4)
  • domain assumption The simulated user answers 'Yes' to a query exactly when the query's word set is a subset of a target item's word set.
    Definition D.1 in Appendix D.1 fixes this response rule; all RA@K and SA@K results depend on this oracle, so if real users answer differently, the measured gains change.
  • domain assumption Every query and item can be faithfully represented as a combination of words extracted from item categorical features.
    Definition 3.1 and Appendix C.2 build Q by enumerating word combinations; the paper does not use real user search logs, so this decomposition is an unverified modeling choice.
  • domain assumption The three relation types R+, R-, and R_perp capture the dependence among queries and items that matters for multi-round recommendation.
    Section 3.1 defines these edge types as a complete partition of node pairs; Eqs. (4) and (9) use them to encode mutual improvement, inhibition, and independence; the paper offers anecdotal examples, not validation that this taxonomy covers real user behavior.
  • domain assumption A test session can be formed by randomly selecting 30 items with at least one positively interacted item, and the target set for that session is the set of positively interacted items among the 30.
    Section 4.1 states this session construction; the evaluation metrics SA@K and RA@K are computed against this synthetic target set, and the distribution of this sampling is not specified.
invented entities (1)
  • Artificial query set Q (all word combinations over item feature words)
    purpose: Provides abstract pre-item actions so the system can ask about user interests before committing to items
    The paper constructs Q by enumerating word combinations (Appendix C.2) and explicitly calls the resulting nodes 'artificial queries' (Section 1); no real user query logs are used, so there is no external evidence that users would formulate these queries.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Why Not Together? A Multiple-Round Recommender System for Queries and Items." pith.science (2026). https://pith.science/paper/RSTDH4UE

@misc{pith2026241210787,
  author       = {Pith},
  title        = {Pith review of: Why Not Together? A Multiple-Round Recommender System for Queries and Items},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RSTDH4UE}},
  note         = {Machine review of arXiv:2412.10787}
}
read the original abstract

A fundamental technique of recommender systems involves modeling user preferences, where queries and items are widely used as symbolic representations of user interests. Queries delineate user needs at an abstract level, providing a high-level description, whereas items operate on a more specific and concrete level, representing the granular facets of user preference. While practical, both query and item recommendations encounter the challenge of sparse user feedback. To this end, we propose a novel approach named Multiple-round Auto Guess-and-Update System (MAGUS) that capitalizes on the synergies between both types, allowing us to leverage both query and item information to form user interests. This integrated system introduces a recursive framework that could be applied to any recommendation method to exploit queries and items in historical interactions and to provide recommendations for both queries and items in each interaction round. Empirical results from testing 12 different recommendation methods demonstrate that integrating queries into item recommendations via MAGUS significantly enhances the efficiency, with which users can identify their preferred items during multiple-round interactions.

Figures

Figures reproduced from arXiv: 2412.10787 by the authors.

Figure 1
Figure 1. An illustrated example of MAGUS recommending both queries (e.g., query B) and items (e.g., items D, E, and F), as shown in (b). This framework effectively leverages both queries (e.g., query A) and items (e.g., items A, B, and C) present in the browsing logs, serving to alleviate the data sparsity issue of user feedback, as shown in (a). and updates. However, this recursive framework presents a chal￾lenge in terms o… view at source ↗
Figure 2
Figure 2. An illustrated example of our relational graph organizing individual words and certain combinations of words (as [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Performance comparisons of MAGUS with different maximum numbers of rounds (i.e.,𝐾MAX) in terms of RA@𝐾MAX and SA@𝐾MAX with the FM recommendation base. detailed description of our simulator and the integration of the large language models are provided in Appendix D.1, D.2. Evaluation Metrics. In our evaluation, we focus on two primary types of metrics in effectively identifying items that match user needs within a se… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Performance comparisons of MAGUS using differ￾ent lengths of the recommendation list in terms of RA@3 and SA@3 with the FM recommendation base. terms of RA@3 but struggle with SA@3 and SA@5, likely because high popularity often occurs in queries, leading to an overemph…
Figure 5
Figure 5. Figure 5: An illustrated example of extending a single-round [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Comparisons of training and inference time of [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: An illustrated case of MAGUS providing one recom￾mendation to user 2742 at each round. The number depicted over each node is its score regarding the user. The popularity level of each node is computed according to its frequency in the user’s browsing log. This comparis…
Figure 9
Figure 9. Figure 9: An illustrated example of exploratory search is [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 20 canonical work pages

  1. [1]

    Ziv Bar-Yossef and Naama Kraus. 2011. Context-sensitive query auto-completion. In WWW

    work page 2011

  2. [2]

    Keping Bi, Qingyao Ai, Yongfeng Zhang, and W Bruce Croft. 2019. Conversa- tional product search based on negative feedback. In Proceedings of the 28th acm international conference on information and knowledge management . 359–368

    work page 2019

  3. [3]

    Dan Busbridge, Dane Sherburn, Pietro Cavallo, and Nils Y Hammerla. 2019. Relational graph attention networks. arXiv preprint arXiv:1904.05811 (2019)

    arXiv 2019

  4. [4]

    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

  5. [5]

    Mostafa Dehghani, Sascha Rothe, Enrique Alfonseca, and Pascal Fleury. 2017. Learning to attend, copy, and generate for session-based query suggestion. In CIKM

    work page 2017

  6. [6]

    Kounianhua Du, Weinan Zhang, Ruiwen Zhou, Yangkun Wang, Xilong Zhao, Jiarui Jin, Quan Gan, Zheng Zhang, and David P Wipf. 2022. Learning enhanced representation for tabular data via neighborhood propagation.Advances in Neural Information Processing Systems 35 (2022), 16373–16384

    work page 2022

  7. [7]

    Chongming Gao, Wenqiang Lei, Xiangnan He, Maarten de Rijke, and Tat-Seng Chua. 2021. Advances and challenges in conversational recommender systems: A survey. AI Open 2 (2021), 100–126

    2021

  8. [8]

    Xavier Glorot and Yoshua Bengio. 2010. Understanding the difficulty of training deep feedforward neural networks. In Proceedings of the thirteenth international conference on artificial intelligence and statistics . JMLR Workshop and Conference Proceedings, 249–256

    2010

Show all 36 references
  1. [9]

    Huifeng Guo, Ruiming Tang, Yunming Ye, Zhenguo Li, and Xiuqiang He. 2017. DeepFM: a factorization-machine based neural network for CTR prediction.IJCAI (2017)

    2017

  2. [10]

    Qingyu Guo, Fuzhen Zhuang, Chuan Qin, Hengshu Zhu, Xing Xie, Hui Xiong, and Qing He. 2020. A survey on knowledge graph-based recommender systems. IEEE Transactions on Knowledge and Data Engineering 34, 8 (2020), 3549–3568

    2020

  3. [11]

    Balázs Hidasi, Alexandros Karatzoglou, Linas Baltrunas, and Domonkos Tikk

  4. [12]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory.Neural computation (1997)

    1997

  5. [13]

    Dietmar Jannach, Ahtsham Manzoor, Wanling Cai, and Li Chen. 2021. A survey on conversational recommender systems. ACM Computing Surveys (CSUR) 54, 5 (2021), 1–36

    2021

  6. [14]

    Jiarui Jin, Xianyu Chen, Weinan Zhang, Junjie Huang, Ziming Feng, and Yong Yu. 2022. Learn over Past, Evolve for Future: Search-based Time-aware Recom- mendation with Sequential Behavior Data. In WWW

    2022

  7. [15]

    Wenqiang Lei, Xiangnan He, Yisong Miao, Qingyun Wu, Richang Hong, Min- Yen Kan, and Tat-Seng Chua. 2020. Estimation-action-reflection: Towards deep interaction between conversational and recommender systems. In Proceedings of the 13th International Conference on Web Search and...

    2020

  8. [16]

    Wenqiang Lei, Gangyi Zhang, Xiangnan He, Yisong Miao, Xiang Wang, Liang Chen, and Tat-Seng Chua. 2020. Interactive path reasoning on graph for conver- sational recommendation. In Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining ....

    2020

  9. [17]

    Raymond Li, Samira Ebrahimi Kahou, Hannes Schulz, Vincent Michalski, Laurent Charlin, and Chris Pal. 2018. Towards deep conversational recommendations. Advances in neural information processing systems 31 (2018)

    2018

  10. [18]

    Tie-Yan Liu et al. 2009. Learning to rank for information retrieval. Foundations and Trends® in Information Retrieval 3, 3 (2009), 225–331

    2009

  11. [19]

    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 KDD

    2018

  12. [20]

    Sabrina J Mielke, Zaid Alyafeai, Elizabeth Salesky, Colin Raffel, Manan Dey, Matthias Gallé, Arun Raja, Chenglei Si, Wilson Y Lee, Benoît Sagot, et al. 2021. Between words and characters: a brief history of open-vocabulary modeling and tokenization in nlp. arXiv preprint arXiv...

    2021 arXiv

  13. [21]

    Jessie Ooi, Xiuqin Ma, Hongwu Qin, and Siau Chuin Liew. 2015. A survey of query expansion, query suggestion and query refinement techniques. In 2015 4th International Conference on Software Engineering and Computer Systems (ICSECS) . IEEE, 112–117

    2015

  14. [22]

    Yanru Qu, Han Cai, Kan Ren, Weinan Zhang, Yong Yu, Ying Wen, and Jun Wang

  15. [23]

    Steffen Rendle. 2010. Factorization machines. In ICDM

    2010

  16. [24]

    Paul Resnick and Hal R Varian. 1997. Recommender systems. Commun. ACM 40, 3 (1997), 56–58

    1997

  17. [25]

    Michael Schlichtkrull, Thomas N Kipf, Peter Bloem, Rianne van den Berg, Ivan Titov, and Max Welling. 2018. Modeling relational data with graph convolutional networks. In European semantic web conference

    2018

  18. [26]

    Alessandro Sordoni, Yoshua Bengio, Hossein Vahabi, Christina Lioma, Jakob Grue Simonsen, and Jian-Yun Nie. 2015. A hierarchical recurrent encoder-decoder for generative context-aware query suggestion. In CIKM

    2015

  19. [27]

    Yueming Sun and Yi Zhang. 2018. Conversational recommender system. InSIGIR

    2018

  20. [28]

    Xiang Wang, Xiangnan He, Yixin Cao, Meng Liu, and Tat-Seng Chua. 2019. Kgat: Knowledge graph attention network for recommendation. In Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining . 950–958

    2019

  21. [29]

    Xiang Wang, Xiangnan He, Meng Wang, Fuli Feng, and Tat-Seng Chua. 2019. Neural graph collaborative filtering. In SIGIR

    2019

  22. [30]

    Xiang Wang, Tinglin Huang, Dingxian Wang, Yancheng Yuan, Zhenguang Liu, Xiangnan He, and Tat-Seng Chua. 2021. Learning intents behind interactions with knowledge graph for recommendation. In WWW

    2021

  23. [31]

    Xiaoying Zhang, Hong Xie, Hang Li, and John CS Lui. 2020. Conversational contextual bandit: Algorithm and application. InProceedings of the web conference

    2020

  24. [32]

    Qinkai Zheng, Houyi Li, Peng Zhang, Zhixiong Yang, Guowei Zhang, Xintan Zeng, and Yongchao Liu. 2021. GIPA: General Information Propagation Algorithm for Graph Learning. arXiv preprint arXiv:2105.06035 (2021)

    2021 arXiv

  25. [33]

    iPhone” items, if the user clearly expresses a preference for other items such as “MacBook

    Guorui Zhou, Xiaoqiang Zhu, Chenru Song, Ying Fan, Han Zhu, Xiao Ma, Yanghui Yan, Junqi Jin, Han Li, and Kun Gai. 2018. Deep interest network for click-through rate prediction. In KDD. KDD’25, August X - Y, 2025, Toronto, Canada Jiarui Jin et al. A A SUMMARY OF NOTATIONS Table...

    2018

  26. [36]

    ‘text”’"

    The maximum number of rounds (denoted as𝐾MAX) is set in the evaluation metrics (see Section 4.1 for details). D SIMULATION DESIGN D.1 Description of Conversational Agent Our simulator introduces a conversational agent built upon the rec- ommendation model to interact with a hu...

    2025

  27. [2015]

    ICLR (2015)

    Session-based recommendations with recurrent neural networks. ICLR (2015)

    2015

  28. [2016]

    Product-based neural networks for user response prediction. In ICDM

Pith tools

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