pith. sign in

arxiv: 2605.17994 · v1 · pith:VCC5DMASnew · submitted 2026-05-18 · 💻 cs.IR · cs.AI

Towards Sustainable Growth: A Multi-Value-Aware Retrieval Framework for E-Commerce Search

Yifan Wang , Yixuan Wang , YiDan Liang , Qiang Liu , Fei Xiao This is my paper

Pith reviewed 2026-05-20 00:44 UTC · model grok-4.3

classification 💻 cs.IR cs.AI
keywords e-commerce searchnew item growthgenerative retrievalcounterfactual inferencelong-term value predictionmulti-value optimizationcold-start itemsMatthew effect
0
0 comments X p. Extension
pith:VCC5DMAS Add to your LaTeX paper What is a Pith Number? 
\usepackage{pith}
\pithnumber{VCC5DMAS}

Prints a linked pith:VCC5DMAS badge after your title and writes the identifier into PDF metadata. Compiles on arXiv with no extra files. Learn more

The pith

A new retrieval framework for e-commerce search uses counterfactual long-term value estimates to promote new items while lifting overall sales.

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

The paper presents GrowthGR, a retrieval framework that counters the tendency of search systems to favor already popular items by estimating each item's future contribution to platform value. It introduces an ItemLTV module that applies counterfactual inference to isolate the lasting transaction impact of a single user interaction with a new item. A MultiGR module then incorporates these estimates into a generative retrieval model trained under a multi-value policy that aligns with the different stages of an online search system. When deployed on Taobao, the approach delivered measurable gains in both new-item and total gross merchandise volume, showing that explicit balancing of short-term conversion and long-term growth is feasible at production scale.

Core claim

The authors establish that a generative retrieval architecture, when augmented with structured cascade signals and trained via Multi-Value-Aware Policy Optimization, can jointly optimize for immediate transactional value and the long-term growth potential predicted by the ItemLTV counterfactual module, producing a 5.3 percent increase in new-item GMV and a 0.3 percent increase in overall search GMV upon production deployment.

What carries the argument

The Item Long-term Transaction Value Prediction (ItemLTV) module, which quantifies long-term value increment from one interaction through counterfactual inference, paired with the Multi-Value-Aware Generative Retrieval (MultiGR) module that applies Multi-Value-Aware Policy Optimization on semantic-ID samples to balance short-term and long-term objectives.

If this is right

  • Search systems can increase exposure for cold-start items without sacrificing short-term conversion metrics.
  • Training objectives that incorporate cascade-stage signals and long-term value estimates improve alignment with multi-stage business outcomes.
  • Generative retrieval models become capable of explicitly trading off immediate revenue against ecosystem growth when supplied with structured multi-value labels.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same counterfactual-plus-generative approach could be tested on other large platforms that currently exhibit strong popularity bias in their ranking layers.
  • Extending the framework to include additional downstream metrics such as repeat purchase rate or seller retention would test whether the balancing effect scales to more value dimensions.
  • If the observed GMV gains persist over longer time windows, the method may offer a practical lever for reducing the Matthew effect across entire item catalogs.

Load-bearing premise

Counterfactual inference in the ItemLTV module can isolate the true long-term value added by one user interaction without being distorted by other user behaviors or platform changes.

What would settle it

A controlled online experiment that disables the ItemLTV estimates or replaces them with random scores and measures whether the lift in new-item GMV disappears while overall GMV remains unchanged or declines.

Figures

Figures reproduced from arXiv: 2605.17994 by Fei Xiao, Qiang Liu, YiDan Liang, Yifan Wang, Yixuan Wang.

Figure 1
Figure 1. Figure 1: The Cold-start Dilemma: Immediate Conversion vs. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The overview of the proposed GrowthGR framework. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Performance improvement on all-net labels across [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: The framework of the decoding strategy. advertisement, mainstream, and the specialized new item stream. For the new item stream, the MultiGR model performs inference in parallel with other pre-processing tasks (such as query parsing) and asynchronously writes the retrieved candidates into a Redis cache. When the request reaches the Match & Rank Engine, it fetches the generative results from Redis. Due to t… view at source ↗
Figure 6
Figure 6. Figure 6: Average Uplift Score Across Different Online Days. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
read the original abstract

New item growth is critical for maintaining a healthy ecosystem in large-scale e-commerce platforms. However, existing systems tend to prioritize presenting users with already popular items, a phenomenon often referred to as the "Matthew effect". In the context of search retrieval, current cold-start models suffer from the misalignment between training objectives and online business metrics, and they lack effective mechanisms to measure an item's growth potential. In this paper, we propose a Multi-Value-Aware retrieval framework tailored for e-commerce search, designed to better align with the cascaded online values across different stages of the search system while balancing immediate conversion and long-term item growth. Our framework GrowthGR consists of two key components: an Item Long-term Transaction Value Prediction (ItemLTV) module and a Multi-Value-Aware Generative Retrieval (MultiGR) module. First, in the ItemLTV module, we employ counterfactual inference to quantify the long-term value increment attributable to a single user interaction. Second, in the MultiGR module, building upon a semantic-ID-based generative retrieval architecture, we leverage structured samples with the search cascade signals and adopt a Multi-Value-Aware Policy Optimization (MoPO) training paradigm to align with multi-stage online values, while explicitly balancing short-term transactional value and long-term growth potential estimated by ItemLTV. We successfully deployed GrowthGR on Taobao's production platform, achieving a substantial 5.3% lift in new item GMV while delivering a non-trivial 0.3% gain in overall search GMV. Extensive online analysis and A/B testing demonstrate its positive impact on the overall ecosystem value.

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 / 1 minor

Summary. The paper proposes GrowthGR, a Multi-Value-Aware retrieval framework for e-commerce search consisting of an ItemLTV module that applies counterfactual inference to estimate long-term transaction value increments from single user interactions and a MultiGR module that extends semantic-ID generative retrieval with a Multi-Value-Aware Policy Optimization (MoPO) objective. The MoPO paradigm incorporates search cascade signals to balance short-term transactional value against long-term growth potential. The central claim is successful production deployment on Taobao yielding a 5.3% lift in new-item GMV and a 0.3% gain in overall search GMV, supported by online A/B testing and ecosystem analysis.

Significance. If the counterfactual estimates prove robust and the A/B results are replicable under proper controls, the framework offers a concrete mechanism to counteract the Matthew effect in retrieval by explicitly trading off immediate conversion against sustainable item growth. The integration of structured cascade signals into a generative retrieval policy is a practical contribution for large-scale platforms seeking multi-stage value alignment.

major comments (2)
  1. [Abstract] Abstract: the reported production A/B test lifts (5.3% new-item GMV, 0.3% overall GMV) constitute the central empirical claim yet supply no information on baselines, statistical tests, data splits, or confounds; without these the lifts cannot be independently assessed.
  2. [ItemLTV module] ItemLTV module (abstract description): counterfactual inference is used to isolate the long-term value increment from a single interaction, but no identification strategy (propensity-score weighting, difference-in-differences with item or user fixed effects, or controls for concurrent promotions/ranking changes) is stated; because MoPO directly optimizes against these estimates, any bias from time-varying confounders directly undermines the reported ecosystem gains.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'non-trivial 0.3% gain' would benefit from explicit comparison to typical variance in overall GMV to clarify its practical importance.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our work. We address each major comment below with clarifications from the manuscript and indicate where revisions will be made to improve transparency.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the reported production A/B test lifts (5.3% new-item GMV, 0.3% overall GMV) constitute the central empirical claim yet supply no information on baselines, statistical tests, data splits, or confounds; without these the lifts cannot be independently assessed.

    Authors: We agree that the abstract, due to length constraints, does not detail the A/B test protocol. The full manuscript describes the online experiments, with the baseline being the prior production retrieval system, a multi-week test period, and significance assessed via standard statistical methods on the GMV metrics. To address the concern directly, we will revise the abstract to briefly note the controlled A/B testing setup and overall statistical significance while preserving conciseness. revision: yes

  2. Referee: [ItemLTV module] ItemLTV module (abstract description): counterfactual inference is used to isolate the long-term value increment from a single interaction, but no identification strategy (propensity-score weighting, difference-in-differences with item or user fixed effects, or controls for concurrent promotions/ranking changes) is stated; because MoPO directly optimizes against these estimates, any bias from time-varying confounders directly undermines the reported ecosystem gains.

    Authors: The referee is correct that the abstract provides only a high-level description of the counterfactual approach in ItemLTV without specifying the identification strategy. The manuscript body outlines the use of counterfactual inference to estimate long-term value increments, but we acknowledge that explicit discussion of controls for time-varying confounders (such as promotions or ranking changes) and fixed effects would strengthen the presentation. We will expand the ItemLTV section in the revision to detail the identification assumptions and mitigation steps employed. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation chain remains independent of its outputs

full rationale

The paper presents ItemLTV as a separate counterfactual-inference module that quantifies long-term value increments from single interactions, then feeds those estimates into the MoPO objective inside MultiGR. No equations, definitions, or self-citations are shown that make the long-term value estimates depend on the MoPO loss or on the final retrieval ranking by construction. The reported 5.3% and 0.3% lifts are attributed to production A/B testing rather than to any internal re-derivation of the same quantities. Because the central claims rest on externally measured deployment outcomes and do not reduce to fitted parameters renamed as predictions or to self-referential definitions, the derivation is self-contained.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 2 invented entities

The framework rests on modeling assumptions for counterfactual value estimation and the effectiveness of multi-stage policy optimization; no explicit free parameters or new physical entities are detailed.

free parameters (1)
  • MoPO training hyperparameters
    Policy optimization parameters are tuned to align multi-stage values and are not derived from first principles.
axioms (1)
  • domain assumption Counterfactual inference can isolate the causal long-term value increment attributable to a single user interaction
    Invoked in the ItemLTV module description.
invented entities (2)
  • ItemLTV module no independent evidence
    purpose: Quantify long-term transaction value via counterfactual inference
    New component introduced to address growth potential measurement gap.
  • MultiGR module no independent evidence
    purpose: Perform multi-value-aware generative retrieval
    Core retrieval architecture tailored to cascade signals.

pith-pipeline@v0.9.0 · 5828 in / 1441 out tokens · 50154 ms · 2026-05-20T00:44:50.857803+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

43 extracted references · 43 canonical work pages · 3 internal anchors

  1. [1]

    Keqin Bao, Jizhi Zhang, Yang Zhang, Wenjie Wang, Fuli Feng, and Xiangnan He. 2023. Tallrec: An effective and efficient tuning framework to align large language model with recommendation. InProceedings of the 17th ACM conference on recommender systems. 1007–1014

    work page 2023

  2. [2]

    Hao Chen, Zefan Wang, Feiran Huang, Xiao Huang, Yue Xu, Yishi Lin, Peng He, and Zhoujun Li. 2022. Generative adversarial framework for cold-start item recommendation. InProceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. 2565–2571

    work page 2022

  3. [3]

    Zeyu Cui, Jianxin Ma, Chang Zhou, Jingren Zhou, and Hongxia Yang. 2022. M6-rec: Generative pretrained language models are open-ended recommender systems.arXiv preprint arXiv:2205.08084(2022)

    work page arXiv 2022

  4. [4]

    Jiaxin Deng, Shiyao Wang, Kuo Cai, Lejian Ren, Qigen Hu, Weifeng Ding, Qiang Luo, and Guorui Zhou. 2025. Onerec: Unifying retrieve and rank with generative recommender and iterative preference alignment.arXiv preprint arXiv:2502.18965 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    Ali Mamdouh Elkahky, Yang Song, and Xiaodong He. 2015. A multi-view deep learning approach for cross domain user modeling in recommendation systems. InProceedings of the 24th international conference on world wide web. 278–288

    work page 2015

  6. [6]

    Francesco Fabbri, Maria Luisa Croci, Francesco Bonchi, and Carlos Castillo. 2022. Exposure inequality in people recommender systems: The long-term effects. In Proceedings of the international AAAI conference on web and social media, Vol. 16. 194–204

    work page 2022

  7. [7]

    Shijie Geng, Shuchang Liu, Zuohui Fu, Yingqiang Ge, and Yongfeng Zhang. 2022. Recommendation as language processing (rlp): A unified pretrain, personalized prompt & predict paradigm (p5). InProceedings of the 16th ACM conference on recommender systems. 299–315

    work page 2022

  8. [8]

    Shijie Geng, Juntao Tan, Shuchang Liu, Zuohui Fu, and Yongfeng Zhang. 2023. Vip5: Towards multimodal foundation models for recommendation.arXiv preprint arXiv:2305.14302(2023)

    work page arXiv 2023

  9. [9]

    Ruidong Han, Bin Yin, Shangyu Chen, He Jiang, Fei Jiang, Xiang Li, Chi Ma, Mincong Huang, Xiaoguang Li, Chunzhen Jing, et al . 2025. Mtgr: Industrial- scale generative recommendation framework in meituan. InProceedings of the 34th ACM International Conference on Information and Knowledge Management. 5731–5738

    work page 2025

  10. [10]

    Zhankui He, Zhouhang Xie, Rahul Jha, Harald Steck, Dawen Liang, Yesu Feng, Bodhisattwa Prasad Majumder, Nathan Kallus, and Julian McAuley. 2023. Large language models as zero-shot conversational recommenders. InProceedings of the 32nd ACM international conference on information and knowledge management. 720–730

    work page 2023

  11. [11]

    Feiran Huang, Yuanchen Bei, Zhenghang Yang, Junyi Jiang, Hao Chen, Qijie Shen, Senzhang Wang, Fakhri Karray, and Philip S Yu. 2025. Large Language Model Simulator for Cold-Start Recommendation. InProceedings of the Eighteenth ACM International Conference on Web Search and Data Mining. 261–270

    work page 2025

  12. [12]

    Feiran Huang, Zefan Wang, Xiao Huang, Yufeng Qian, Zhetao Li, and Hao Chen

    work page

  13. [13]

    InProceedings of the 46th international ACM SIGIR conference on research and development in information retrieval

    Aligning distillation for cold-start item recommendation. InProceedings of the 46th international ACM SIGIR conference on research and development in information retrieval. 1147–1157

    work page

  14. [14]

    Kalervo Järvelin and Jaana Kekäläinen. 2002. Cumulated gain-based evaluation of IR techniques.ACM Transactions on Information Systems (TOIS)20, 4 (2002), 422–446

    work page 2002

  15. [15]

    Hao Jiang, Chuanzhen Li, Juanjuan Cai, Runyu Tian, and Jingling Wang. 2023. Self-supervised contrastive enhancement with symmetric few-shot learning towers for cold-start news recommendation. InProceedings of the 32nd ACM International Conference on Information and Knowledge Management. 945–954

    work page 2023

  16. [16]

    Jinri Kim, Eungi Kim, Kwangeun Yeo, Yujin Jeon, Chanwoo Kim, Sewon Lee, and Joonseok Lee. 2024. Content-based graph reconstruction for cold-start item recommendation. InProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval. 1263–1273

    work page 2024

  17. [17]

    Doyup Lee, Chiheon Kim, Saehoon Kim, Minsu Cho, and Wook-Shin Han. 2022. Autoregressive image generation using residual quantization. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 11523–11532

    work page 2022

  18. [18]

    Jiacheng Li, Ming Wang, Jin Li, Jinmiao Fu, Xin Shen, Jingbo Shang, and Julian McAuley. 2023. Text is all you need: Learning language representations for sequential recommendation. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 1258–1267

    work page 2023

  19. [19]

    Jinming Li, Wentao Zhang, Tian Wang, Guanglei Xiong, Alan Lu, and Gerard Medioni. 2023. GPT4Rec: A generative framework for personalized recommen- dation and user interests interpretation.arXiv preprint arXiv:2304.03879(2023)

    work page arXiv 2023

  20. [20]

    Zida Liang, Changfa Wu, Dunxian Huang, Weiqiang Sun, Ziyang Wang, Yuliang Yan, Jian Wu, Yuning Jiang, Bo Zheng, Ke Chen, et al. 2025. Tbgrecall: A generative retrieval model for e-commerce recommendation scenarios. InProceedings of the 34th ACM International Conference on Information and Knowledge Management. 5863–5870

    work page 2025

  21. [21]

    Chi Liu, Jiangxia Cao, Rui Huang, Kai Zheng, Qiang Luo, Kun Gai, and Guorui Zhou. 2024. KuaiFormer: Transformer-Based Retrieval at Kuaishou.arXiv preprint arXiv:2411.10057(2024)

    work page arXiv 2024

  22. [22]

    Taichi Liu, Chen Gao, Zhenyu Wang, Dong Li, Jianye Hao, Depeng Jin, and Yong Li. 2023. Uncertainty-aware Consistency Learning for Cold-Start Item Recommendation. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. 2466–2470

    work page 2023

  23. [23]

    Weiming Liu, Chaochao Chen, Xinting Liao, Mengling Hu, Jiajie Su, Yanchao Tan, and Fan Wang. 2024. User distribution mapping modelling with collaborative filtering for cross domain recommendation. InProceedings of the ACM Web Conference 2024. 334–343

    work page 2024

  24. [24]

    Yunze Luo, Yuezihan Jiang, Yinjie Jiang, Gaode Chen, Jingchi Wang, Kaigui Bian, Peiyi Li, and Qi Zhang. 2025. Online item cold-start recommendation with popularity-aware meta-learning. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 1. 927–937

    work page 2025

  25. [25]

    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. InThe 41st International ACM SIGIR Conference on Research & Development in Information Retrieval. 1137–1140

    work page 2018

  26. [26]

    Haoyu Pang, Fausto Giunchiglia, Ximing Li, Renchu Guan, and Xiaoyue Feng

    work page

  27. [27]

    InProceedings of the ACM Web Conference

    PNMTA: A pretrained network modulation and task adaptation approach for user cold-start recommendation. InProceedings of the ACM Web Conference

    work page

  28. [28]

    Shashank Rajput, Nikhil Mehta, Anima Singh, Raghunandan Hulikal Keshavan, Trung Vu, Lukasz Heldt, Lichan Hong, Yi Tay, Vinh Tran, Jonah Samost, et al

    work page

  29. [29]

    Recommender systems with generative retrieval.Advances in Neural Information Processing Systems36 (2023), 10299–10315

    work page 2023

  30. [30]

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Yang Wu, et al. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300(2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  31. [31]

    Qijie Shen, Yuanchen Bei, Zihong Huang, Jialin Zhu, Keqin Xu, Boya Du, Jiawei Tang, Yuning Jiang, Feiran Huang, Xiao Huang, et al. 2025. AliBoost: Ecological Boosting Framework in Alibaba Platform. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2. 4827–4838

    work page 2025

  32. [32]

    Xuehan Sun, Tianyao Shi, Xiaofeng Gao, Yanrong Kang, and Guihai Chen. 2021. FORM: follow the online regularized meta-leader for cold-start recommendation. InProceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval. 1177–1186

    work page 2021

  33. [33]

    Jizhe Wang, Pipei Huang, Huan Zhao, Zhibo Zhang, Binqiang Zhao, and Dik Lun Lee. 2018. Billion-scale commodity embedding for e-commerce recommendation in alibaba. InProceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining. 839–848

    work page 2018

  34. [34]

    Xuansheng Wu, Huachi Zhou, Yucheng Shi, Wenlin Yao, Xiao Huang, and Ning- hao Liu. 2024. Could small language models serve as recommenders? towards data-centric cold-start recommendation. InProceedings of the ACM Web Confer- ence 2024. 3566–3575

    work page 2024

  35. [35]

    Xiaolong Xu, Hongsheng Dong, Lianyong Qi, Xuyun Zhang, Haolong Xiang, Xiaoyu Xia, Yanwei Xu, and Wanchun Dou. 2024. Cmclrec: Cross-modal con- trastive learning for user cold-start sequential recommendation. InProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval. 1589–1598

    work page 2024

  36. [36]

    Guipeng Xv, Chen Lin, Wanxian Guan, Jinping Gou, Xubin Li, Hongbo Deng, Jian Xu, and Bo Zheng. 2023. E-commerce search via content collaborative graph neural network. InProceedings of the 29th ACM SIGKDD conference on knowledge discovery and data mining. 2885–2897

    work page 2023

  37. [37]

    Jiaqi Zhai, Lucy Liao, Xing Liu, Yueming Wang, Rui Li, Xuan Cao, Leon Gao, Zhao- jie Gong, Fangda Gu, Michael He, et al. 2024. Actions speak louder than words: Trillion-parameter sequential transducers for generative recommendations.arXiv preprint arXiv:2402.17152(2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  38. [38]

    Chao Zhang, Shiwei Wu, Haoxin Zhang, Tong Xu, Yan Gao, Yao Hu, and Enhong Chen. 2024. Notellm: A retrievable large language model for note recommenda- tion. InCompanion Proceedings of the ACM Web Conference 2024. 170–179

    work page 2024

  39. [39]

    Weizhi Zhang, Yuanchen Bei, Liangwei Yang, Henry Peng Zou, Peilin Zhou, Aiwei Liu, Yinghui Li, Hao Chen, Jianling Wang, Yu Wang, et al. 2025. Cold-start rec- ommendation towards the era of large language models (llms): A comprehensive survey and roadmap.arXiv preprint arXiv:2501.01945(2025)

    work page arXiv 2025

  40. [40]

    Wenlin Zhang, Chuhan Wu, Xiangyang Li, Yuhao Wang, Kuicai Dong, Yichao Wang, Xinyi Dai, Xiangyu Zhao, Huifeng Guo, and Ruiming Tang. 2025. Llmtreerec: Unleashing the power of large language models for cold-start recom- mendations. InProceedings of the 31st International Conference on Computational Linguistics. 886–896

    work page 2025

  41. [41]

    Zhaoqi Zhang, Haolei Pei, Jun Guo, Tianyu Wang, Yufei Feng, Hui Sun, Shaowei Liu, and Aixin Sun. 2025. OneTrans: Unified Feature Interaction and Sequence Modeling with One Transformer in Industrial Recommender.arXiv preprint arXiv:2510.26104(2025)

    work page arXiv 2025

  42. [42]

    Zhihui Zhou, Lilin Zhang, and Ning Yang. 2023. Contrastive collaborative filtering for cold-start item recommendation. InProceedings of the ACM web conference

    work page 2023

  43. [43]

    head-item

    928–937. Conference acronym ’XX, June 03–05, 2018, Woodstock, NY WANG et al. ... Search Logs Search Data Pre-processingModel Fine-Tuning Deployment & Serving Platforms for LLMs New Item PoolMatch & RankEngine Search Server Main RecommenderAds Recommender Redis Offline Training RecordDaily Update Request Real-time Data GrowthPrediction Final Item ListReque...

    work page arXiv 2018