Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

UniSearch: Rethinking Search System with a Unified Generative Architecture

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A single generative model, trained end-to-end, can replace the multi-stage search cascade and beat it in production.

desk verdict A credible industrial-scale generative search deployment with a real +3.31% TPC gain, but weak statistical reporting and unmeasured semantic-ID churn under live-stream dynamics keep it from being fully convincing. read the letter →

arxiv 2509.06887 v2 pith:L5NMLE7R submitted 2025-09-08 cs.IR

Jiahui Chen , Xiaoze Jiang , Zhibo Wang , Quanzhi Zhu , Junyao Zhao , Feng Hu , Kang Pan , Ao Xie
show 13 more authors
Maohua Pei Zhiheng Qin Hongjing Zhang Zhixin Zhai Xiaobo Guo Runbin Zhou Kefeng Wang Mingyang Geng Cheng Chen Jingshan Lv Yupeng Huang Xiao Liang Han Li
This is my paper · ORCID
classification cs.IR
keywords GenerativeSearchInformationRetrievalEnd-to-EndArchitectureSemanticIdentifiersVectorQuantizationPreferenceOptimizationIndustrialDeploymentMulti-stageCascade
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

UniSearch sets out to show that industrial search does not need the conventional three-stage pipeline of recall, pre-ranking, and ranking. It replaces that pipeline with one generative model that reads a text query plus user context and directly emits short semantic identifiers for relevant videos, while a jointly trained video encoder builds and discretizes those identifiers. The paper argues that training both halves in one loop rather than tokenizing items first and training a generator second removes objective mismatch, and that a post-training stage using real user feedback aligns the outputs with clicks and watch time. If its results hold, unified generative search is production-viable at industrial scale, with a reported 3.31% total play-count gain in live search and larger relative gains for long-tail queries and new users.

What carries the argument

The load-bearing object is the semantic identifier (SID): a short sequence of codebook tokens that stands for a video. It is produced by a VQ-VAE video encoder and predicted one token at a time by the search generator, so it is simultaneously the index entry that makes a video retrievable and the output unit the model learns to emit. The paper's key mechanism is training the codebook and the generator jointly so that encoding and generation cannot drift apart; residual contrastive learning keeps successive SID tokens complementary rather than collapsed onto one path, and the prefix tree over valid SIDs converts open-ended decoding into constrained, valid-path decoding.

What would settle it

Measure retrieval recall for videos created 0, 1, 6, and 24 hours before the query, comparing the prefix-tree-constrained generator against a cascade that indexes items immediately; if freshly created searchable videos are often absent from the valid identifier tree or receive unstable identifiers, the unified generative design would miss them and the reported gains would not survive fast item churn.

Watch

Extended reading notes

Core claim

The paper's central claim is that the two jobs of generative search, turning videos into discrete semantic identifiers and turning queries into those identifiers, should be optimized as one objective instead of two. The video encoder maps multi-modal video signals into residual latent embeddings and quantizes them with a vector-quantization codebook into a short semantic-ID sequence; the generator autoregressively predicts that sequence from the query. Residual contrastive learning with a coarse-to-fine negative-sampling schedule aligns query and video embeddings, and reject-sampling reweights the generator's next-token loss toward high-quality items. A post-training stage, Search Preference Optimization, scores beam-searched candidates with a reward model plus real user interactions and applies a reward-advantage-weighted update while staying close to the pretrained policy. At inference a prefix tree over valid semantic-ID paths constrains decoding, and the authors report that the deployed system beats strong cascaded and two-stage generative baselines on offline metrics and in online tests.

Load-bearing premise

The approach assumes that the automatically learned short identifiers for videos remain stable, valid, and complete as the video corpus changes in real time; if identifiers go stale or newly created videos have no valid identifier before the prefix tree refreshes, those videos cannot be generated as search results.

Editorial extensions

If this is right

  • Replacing recall, pre-ranking, and ranking with one model removes the need to maintain and tune several stage-specific models and removes objective mismatches between them.
  • Unified training appears to scale: 6-, 12-, and 24-layer versions each beat same-scale two-stage generative baselines on recall and ranking accuracy (MRR).
  • Online Search Preference Optimization turns the deployed search system into a feedback loop that learns directly from clicks, watch time, and system-predicted quality signals.
  • The architecture transfers from a half-million-candidate live-stream search to a roughly billion-candidate short-video search, indicating the design can be scaled by enlarging the codebook and model rather than adding pipeline stages.
  • The reported user-side effects, higher total play counts, earlier and more satisfied interaction, and reduced query reformulation, suggest the unified model changes what users experience in search, not just internal system cost.

Reading between the lines

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

  • If the same long-tail advantage holds, the design should be most valuable for sparse, semantically rich query traffic, where a cascade's head-item bias is worst; the paper only splits long-tail and head queries into two buckets.
  • The paper does not measure identifier freshness; tracking retrieval recall of items created minutes before the query would quantify how much dynamic-corpus churn costs the architecture.
  • The acknowledged point-wise beam-search limitation suggests listwise generation is the natural next test: listwise sampling could raise diversity without sacrificing ranking accuracy, but that is an unverified direction.
  • The same joint tokenization-generation loop should transfer to other text-to-item searches, such as products, places, or documents, wherever a semantic item identifier can be learned; the paper only demonstrates video items.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper presents UniSearch, a unified generative search architecture for industrial search at Kuaishou. It replaces the traditional multi-stage cascaded pipeline with a single Search Generator that autoregressively outputs semantic identifiers (SIDs) of relevant items, together with a Video Encoder that produces the SIDs via VQ-VAE. The two components are trained jointly through a unified pre-training objective combining residual contrastive learning, codebook learning, and next-token prediction, followed by an online post-training step, Search Preference Optimization (SPO), which aligns the generator with rewards from the production ranker and real user interactions. Offline experiments on live-streaming search data show improvements in Recall@300 and MRR over several baselines, and online A/B tests in live and short-video search report gains in total play counts, click-through rate, and other business metrics, including +3.31% total play count in live search.

Significance. The paper addresses a timely and important problem: whether generative models can replace the entire cascaded retrieval-ranking pipeline in a large-scale industrial search engine. The proposed unified training framework, which jointly optimizes tokenization and generation, is a plausible way to avoid the objective mismatch in prior generative recommenders. The scale of the deployment and the reported online gains are substantial. If the results hold, the paper would be a strong demonstration that generative retrieval is production-viable at industrial scale. However, the central claim is only as strong as the reliability of the evaluation; the paper currently lacks statistical rigor, and several aspects of the experimental design create circularity between the reward signal and the offline test set. The paper does not provide code or data for reproducibility, which is common for industry papers, but the absence of error bars and significance tests is a genuine limitation.

major comments (4)
  1. [Section 4.4, Table 3] The online A/B test is reported without any statistical significance testing, confidence intervals, or variance estimates. The description of the 7-day, 10%-traffic experiment does not state whether the control and treatment groups were balanced on key covariates, whether concurrent system changes were controlled, or how the daily metric distributions varied. For a claim as strong as the largest single-experiment improvement in recent years, the absence of uncertainty quantification makes it impossible to rule out that the reported gains (e.g., +3.31% TPC) reflect noise or confounding. The same concern applies to the offline results in Table 1 and Figures 3-4, which report point estimates without error bars or significance tests. Please report per-day metrics with standard errors and a significance test, and describe the guardrails used to ensure that no other changes affected the comparison.
  2. [Section 4.1 and Section 3.3, Eq. (6)] The RK test set is defined as the set of top videos composed by the production search system (Section 4.1), and the SPO reward uses the production fine-ranking module's predicted scores as R_system (Eq. 6). As a result, the reported MRR gains on the RK set partly measure the degree to which UniSearch reproduces the decisions of the very cascade it is designed to replace. This circularity is load-bearing because one of the paper's central claims is that UniSearch outperforms the cascaded pipeline. To address it, the authors should either (i) evaluate on an independent ground truth (e.g., human relevance judgments or a held-out set of explicit user feedback not derived from system rankings), or (ii) ablate the effect of removing R_system from the reward and report the RK metrics without the production ranker's labels.
  3. [Section 3.4 and Section 4.3] The paper acknowledges that live streams start and end and that changing content can alter a stream's valid semantic identifiers, necessitating continuous Trie updates, but it provides no empirical characterization of the resulting churn. There are no measurements of the fraction of fresh items that lack valid SIDs at query time, no age-stratified retrieval metrics, and no analysis of how stale semantic codes degrade Recall@300 and MRR. In a live-search product, the ability to serve newly created or recently modified streams is essential; a 7-day A/B test cannot reveal slow decay of the semantic ID space. Please add an analysis of Trie coverage over time and report offline metrics stratified by item age or by time since the last content update.
  4. [Section 4.4] The short-video deployment does not fully replace the cascaded architecture; the text states that UniSearch is "first introduced as a new resource in the ranking stage, while several recall sources are gradually deactivated." This is materially weaker than the abstract's claim that UniSearch "replaces the cascaded pipeline." The paper should either restrict the replacement claim to the live-search scenario (where full replacement appears to have been performed) or describe the intermediate stage as a partial integration and present evidence that the final configuration is a full replacement.
minor comments (6)
  1. [Section 3.1] The Video Encoder is described as a unidirectional Transformer, but Section 4.1 states it is instantiated with BERT, which is bidirectional. Please clarify whether a unidirectional variant of BERT is used.
  2. [Figures 1 and 5] The figures contain stray Chinese characters ("保存" and "保存1") that should be removed or translated.
  3. [Table 1] The header row is confusing because the first data row (FSQ) has the codebook and generation model columns blank; please format the table so that each baseline's components are clearly aligned.
  4. [Abstract and Section 5] The claim that the deployment "yields the largest single-experiment improvement in recent years of our product's history" is not accompanied by any supporting data or historical context; please provide evidence or temper the claim.
  5. [Section 4.1] The definition of MRR uses a single ground-truth result per sample (K_i), but queries often have multiple relevant videos; please clarify how multiple ground-truth positives are handled in the MRR computation.
  6. [References] The ACM reference format includes a placeholder "In the Preprint Version of UniSearch, Beijing, China.ACM, New York, NY, USA" that should be replaced with the actual venue.

Circularity Check

1 steps flagged · score 5.0 of 10

SPO trains against the production fine-ranking module while RK is defined as that module's top outputs, making the headline offline RK gains partly self-referential; CK and online A/B remain independent.

  1. self definitional [Section 3.3 Eq. (6) and Section 4.1 Evaluation Metrics]
    "we adopt the existing fine-ranking module of the production search system as the Reward System... These scores constitute part of the reward signal... R=γ1 R_system+γ2 R_interaction... The RK test set includes the top videos composed to users by the search system, used to evaluate the alignment between generated videos and system preferences."

    The SPO objective in Eq. (6) explicitly maximizes a reward whose R_system component is the production fine-ranking module's score. The RK test set is then defined as the top videos produced by that same system. Thus the RK MRR numbers reported for the 'w/ Online SPO' rows measure how well the generator reproduces the exact ranker that supplied its training reward; improving RK MRR is a direct consequence of optimizing R_system, not an independent measure of retrieval quality. The paper labels RK as 'alignment with system preferences,' so the metric is internally consistent, but using it in the main results as evidence of superiority over the cascade is partially circular. CK and the online A/B test provide less self-referential evidence.

full rationale

Aside from the SPO/RK coupling, the paper's derivation is largely self-contained: the unified pre-training objective (Eq. 5) jointly optimizes contrastive alignment, codebook discretization, and next-token generation, and its offline gains on CK and online A/B metrics (TPC +3.31%, CTR, CQR, etc.) are measured against user behavior and the production baseline, not against the model's own training signal. References to prior work, including OneSug and the authors' own query auto-completion paper, are positioning citations and not load-bearing for the architecture's validity. The unmeasured semantic-ID churn under live-stream dynamics is a correctness risk but not a circularity. Overall, the central claim retains independent support, but the headline RK evaluation after SPO is partly self-referential, so the paper is not fully clean.

Assumptions & free parameters 6 free parameters · 4 assumptions · 1 invented entities

The central empirical claims rest mostly on the authors' own training and evaluation choices rather than external benchmarks. The most consequential hand-chosen values are codebook size and depth, loss and reward weights, KL coefficient, and beam size, none with sensitivity analysis. The evaluation assumes that system-generated labels and clicks are valid proxies for relevance, and that the production reward model transfers into user value. No code, data, or external replication is provided.

free parameters (6)
  • Contrastive temperature tau = not reported
    Appears in Eq. (2) controlling the sharpness of the InfoNCE-style contrastive loss; no value or sensitivity analysis is given.
  • Loss weights lambda_1, lambda_2, lambda_3 = not reported
    Balance contrastive, codebook, and generation losses in Eq. (5); values are never stated and no ablation is shown.
  • Reward weights gamma_1, gamma_2 = not reported
    Balance system-estimated and user-interaction rewards in Eq. (6); values are not reported.
  • KL regularization coefficient beta = not reported
    Controls deviation from the reference policy in SPO, Eq. (8); value is not stated.
  • Codebook depth and size = k=3, size 512 (live search) and 8192 (short-video search)
    Chosen by ablation in Figure 3; directly determines the capacity and coverage of the semantic ID space.
  • Beam size = 256
    Inference beam size, with no sensitivity analysis shown; larger beams trade latency against recall.
assumptions (4)
  • domain assumption VQ-VAE discrete semantic IDs preserve enough query-relevant information for generative retrieval
    The whole framework assumes a 3-level codebook can represent items well enough for recall and ranking; path validity is measured (99.8% with Trie), but semantic fidelity is not.
  • domain assumption The production fine-ranking reward model is a valid proxy for user preference
    SPO trains against R_system plus user interactions (Eq. 6); if the reward model is biased, preference alignment inherits that bias.
  • domain assumption Recall@300 and MRR on RK and CK test sets reflect real search quality
    RK labels come from the production system's own top videos and CK labels from clicks; both are proxies, not ground-truth relevance.
  • domain assumption Trie-constrained beam search has negligible negative effect on generation diversity
    The paper reports path validity improvements from the Trie but does not measure how often valid but undesirable paths are suppressed, or how diversity changes beyond one case study.
invented entities (1)
  • Semantic identifier (SID) sequences produced by the VQ-VAE codebook
    purpose: Discrete tokens that bridge queries and video items inside the unified generative model.
    SIDs are internal to UniSearch; no external falsifiable handle is provided beyond the system's own path-validity and retrieval metrics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UniSearch: Rethinking Search System with a Unified Generative Architecture." pith.science (2026). https://pith.science/paper/L5NMLE7R

@misc{pith2026250906887,
  author       = {Pith},
  title        = {Pith review of: UniSearch: Rethinking Search System with a Unified Generative Architecture},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L5NMLE7R}},
  note         = {Machine review of arXiv:2509.06887}
}
read the original abstract

Modern search systems play a crucial role in facilitating information acquisition. Traditional search engines typically rely on a cascaded architecture, where results are retrieved through recall, pre-ranking, and ranking stages. The complexity of designing and maintaining multiple modules makes it difficult to achieve holistic performance gains. Recent advances in generative recommendation have motivated the exploration of unified generative search as an alternative. However, existing approaches are not genuinely end-to-end: they typically train an item encoder to tokenize candidates first and then optimize a generator separately, leading to objective inconsistency and limited generalization. To address these limitations, we propose UniSearch, a unified generative search framework for Kuaishou Search. UniSearch replaces the cascaded pipeline with an end-to-end architecture that integrates a Search Generator and a Video Encoder. The Generator produces semantic identifiers of relevant items given a user query, while the Video Encoder learns latent item embeddings and provides their tokenized representations. A unified training framework jointly optimizes both components, enabling mutual enhancement and improving representation quality and generation accuracy. Furthermore, we introduce Search Preference Optimization (SPO), which leverages a reward model and real user feedback to better align generation with user preferences. Extensive experiments on industrial-scale datasets, together with online A/B testing in both short-video and live search scenarios, demonstrate the strong effectiveness and deployment potential of UniSearch. Notably, its deployment in live search yields the largest single-experiment improvement in recent years of our product's history, highlighting its practical value for real-world applications.

Figures

Figures reproduced from arXiv: 2509.06887 by the authors.

Figure 1
Figure 1. A schematic illustration of existing search architec [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The overall architecture of the UniSearch, where “Seq.” denotes “Sequence”, “Fea.” stands for “Feature”, “Learned Tok.” [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. The results of incorporating the Trie on RK and CK [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: The analysis of the improvements in Total Play [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Search results comparison between the Multi-stage [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Token-Native Storage: Read and Write in your Agent's Language

    cs.DB 2026-08 unverdicted novelty 5.0 of 10

    Storing text as BPE token IDs instead of UTF-8 bytes gives 1.6–3.4x compression and lets models read/write token IDs directly, avoiding repeated tokenization.

Reference graph

Works this paper leans on

44 extracted references · 21 canonical work pages · cited by 1 Pith paper

  1. [1]

    Yimeng Bai, Shunyu Zhang, Yang Zhang, Hu Liu, Wentian Bao, Enyun Yu, Fuli Feng, and Wenwu Ou. 2025. Unconstrained Monotonic Calibration of Predictions in Deep Ranking Systems. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval

    work page 2025

  2. [2]

    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)

    arXiv 2025

  3. [3]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 4171–4186

    2019

  4. [4]

    Luciano Floridi and Massimo Chiriatti. 2020. GPT-3: Its Nature, Scope, Limits, and Consequences.Minds and Machines30 (2020), 681–694

    work page 2020

  5. [5]

    Xian Guo, Ben Chen, Siyuan Wang, Ying Yang, Chenyi Lei, Yuqing Ding, and Han Li. 2025. OneSug: The Unified End-to-End Generative Framework for E-commerce Query Suggestion.arXiv preprint arXiv:2506.06913(2025)

    arXiv 2025

  6. [6]

    Ruidong Han, Bin Yin, Shangyu Chen, He Jiang, Fei Jiang, Xiang Li, Chi Ma, Mingcong Huang, Xiaoguang Li, Chunzhen Jing, Yueming Han, Menglei Zhou, Lei Yu, Chuan Liu, and Wei Lin. 2025. MTGR: Industrial-Scale Generative Rec- ommendation Framework in Meituan.arXiv preprint arXiv:2505.18654(2025)

    arXiv 2025

  7. [7]

    Yanhua Huang, Yuqi Chen, Xiong Cao, Rui Yang, Mingliang Qi, Yinghao Zhu, Qingchang Han, Yaowei Liu, Zhaoyu Liu, Xuefeng Yao, Yuting Jia, Leilei Ma, Yinqi Zhang, Taoyu Zhu, Liujie Zhang, Lei Chen, Weihang Chen, Min Zhu, Ruiwen Xu, and Lei Zhang. 2025. Towards Large-scale Generative Ranking.arXiv preprint arXiv:2505.04180(2025)

    arXiv 2025

  8. [8]

    Sujay Khandagale, Bhawna Juneja, Prabhat Agarwal, Aditya Subramanian, Jae- won Yang, and Yuting Wang. 2025. InteractRank: Personalized Web-Scale Search Pre-Ranking with Cross Interaction Features. InCompanion Proceedings of the ACM on Web Conference 2025. 287–295

    work page 2025

Show all 44 references
  1. [9]

    Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension. InProceedings of the 58th...

    2020

  2. [10]

    Mingming Li, Huimu Wang, Zuxu Chen, Guangtao Nie, Yiming Qiu, Guoyu Tang, Lin Liu, and Jingwei Zhuo. 2024. Generative Retrieval with Preference Optimization for E-commerce Search.arXiv preprint arXiv:2407.19829(2024)

    2024 arXiv

  3. [11]

    Xiaoxi Li, Jiajie Jin, Yujia Zhou, Yuyao Zhang, Peitian Zhang, Yutao Zhu, and Zhicheng Dou. 2025. From Matching to Generation: A Survey on Generative Information Retrieval.ACM Transactions on Information Systems(2025), 1–62

    2025

  4. [12]

    Xiaoxi Li, Yujia Zhou, and Zhicheng Dou. 2024. UniGen: A Unified Generative Framework for Retrieval and Question Answering with Large Language Models. InProceedings of the AAAI Conference on Artificial Intelligence. 8688–8696

    2024

  5. [13]

    Juexin Lin, Sachin Yadav, Feng Liu, Nicholas Rossi, Praveen R Suram, Satya Chem- bolu, Prijith Chandran, Hrushikesh Mohapatra, Tony Lee, Alessandro Magnani, et al. 2024. Enhancing Relevance of Embedding-based Retrieval at Walmart. In Proceedings of the 33rd ACM International C...

    2024

  6. [14]

    Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Cheng- gang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, et al. 2024. Deepseek-V3 Technical Report.arXiv preprint arXiv:2412.19437(2024)

    2024 arXiv

  7. [15]

    Yuxuan Liu, Tianchi Yang, Zihan Zhang, Minghui Song, Haizhen Huang, Weiwei Deng, Feng Sun, and Qi Zhang. 2024. ASI++: Towards Distributionally Balanced End-to-End Generative Retrieval.arXiv preprint arXiv:2405.14280(2024)

    2024 arXiv

  8. [16]

    Fabian Mentzer, David Minnen, Eirikur Agustsson, and Michael Tschannen. 2024. Finite Scalar Quantization: VQ-VAE Made Simple.The Twelfth International Conference on Learning Representations

    2024

  9. [17]

    Agnès Mustar, Sylvain Lamprier, and Benjamin Piwowarski. 2020. Using BERT and BART for Query Suggestion. InJoint Conference of the Information Retrieval Communities in Europe, Vol. 2621

    2020

  10. [18]

    OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, et al

  11. [19]

    Ming Pang, Chunyuan Yuan, Xiaoyu He, Zheng Fang, Donghao Xie, Fanyi Qu, Xue Jiang, Changping Peng, Zhangang Lin, Ching Law, and Jingping Shao. 2025. Generative Retrieval and Alignment Model: A New Paradigm for E-commerce Retrieval.arXiv preprint arXiv:2504.01403(2025)

    2025 arXiv

  12. [20]

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

  13. [21]

    Yubao Tang, Ruqing Zhang, Jiafeng Guo, Maarten de Rijke, Wei Chen, and Xueqi Cheng. 2024. Listwise Generative Retrieval Models via a Sequential Learning Process. (2024), 31 pages

    2024

  14. [22]

    Yi Tay, Vinh Tran, Mostafa Dehghani, Jianmo Ni, Dara Bahri, Harsh Mehta, Zhen Qin, Kai Hui, Zhe Zhao, Jai Gupta, et al. 2022. Transformer Memory as a Differentiable Search Index.Advances in Neural Information Processing Systems 35 (2022), 21831–21843

    2022

  15. [23]

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. LLaMA: Open and Efficient Foundation Language Models. arXiv preprint arXiv:2302.13971(2023)

    2023 arXiv

  16. [24]

    Aaron Van Den Oord, Oriol Vinyals, et al. 2017. Neural Discrete Representation Learning.Advances in Neural Information Processing Systems30 (2017)

    2017

  17. [25]

    Dongsheng Wang, Yuxi Huang, Shen Gao, Yifan Wang, Chengrui Huang, and Shuo Shang. 2025. Generative Next POI Recommendation with Semantic ID. arXiv preprint arXiv:2506.01375(2025)

    2025 arXiv

  18. [26]

    Zhibo Wang, Xiaoze Jiang, Zhiheng Qin, Enyun Yu, and Han Li. 2025. Personal- ized Query Auto-Completion for Long and Short-Term Interests with Adaptive Detoxification Generation. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD ’2...

    2025

  19. [27]

    Zhipeng Wei, Kuo Cai, Junda She, Jie Chen, Minghao Chen, Yang Zeng, Qiang Luo, Wengcong Zeng, Ruiming Tang, Kun Gai, and Guorui Zhou. 2025. OneLoc: Geo-Aware Generative Recommender Systems for Local Life Service.arXiv preprint arXiv:2508.14646(2025)

    2025 arXiv

  20. [28]

    Haoyi Xiong, Jiang Bian, Yuchen Li, Xuhong Li, Mengnan Du, Shuaiqiang Wang, Dawei Yin, and Sumi Helal. 2024. When Search Engine Services meet Large Lan- guage Models: Visions and Challenges.IEEE Transactions on Services Computing (2024)

    2024

  21. [29]

    Enqiang Xu, Yiming Qiu, Junyang Bai, Ping Zhang, Dadong Miao, Songlin Wang, Guoyu Tang, Lin Liu, and MingMing Li. 2024. Optimizing E-commerce Search: Toward a Generalizable and Rank-Consistent Pre-Ranking Model. InProceedings of the 47th International ACM SIGIR Conference on R...

    2024

  22. [30]

    An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al. 2025. Qwen3 Technical Report.arXiv preprint arXiv:2505.09388(2025)

    2025 arXiv

  23. [31]

    Tianchi Yang, Minghui Song, Zihan Zhang, Haizhen Huang, Weiwei Deng, Feng Sun, and Qi Zhang. 2023. Auto Search Indexer for End-to-End Document Retrieval. InFindings of the Association for Computational Linguistics: EMNLP 2023. 6955– 6970

    2023

  24. [32]

    Yuhao Yang, Zhi Ji, Zhaopeng Li, Yi Li, Zhonglin Mo, Yue Ding, Kai Chen, Zijian Zhang, Jie Li, Shuanglong Li, and Lin Liu. 2025. Sparse Meets Dense: Unified Generative Recommendations with Cascaded Sparse-Dense Representations. arXiv preprint arXiv:2503.02453(2025)

    2025 arXiv

  25. [33]

    Cheng Ye. 2024. Exploring a learning-to-rank approach to enhance the Retrieval Augmented Generation (RAG)-based electronic medical records search engines. Informatics and Health1, 2 (2024), 93–99

    2024

  26. [34]

    Di Yin, Jiwei Tan, Zhe Zhang, Hongbo Deng, Shujian Huang, and Jiajun Chen

  27. [35]

    Jiaqi Zhai, Lucy Liao, Xing Liu, Yueming Wang, Rui Li, Xuan Cao, Leon Gao, Zhao- jie Gong, Fangda Gu, Michael He, Yinghai Lu, and Yu Shi. 2024. Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations.arXiv preprint arXiv:2402.1...

    2024 arXiv

  28. [36]

    Shunyu Zhang, Yaobo Liang, Ming Gong, Daxin Jiang, and Nan Duan. 2022. Multi- View Document Representation Learning for Open-Domain Dense Retrieval. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 5990–6000

    2022

  29. [37]

    Bowen Zheng, Yupeng Hou, Hongyu Lu, Yu Chen, Wayne Xin Zhao, Ming Chen, and Ji-Rong Wen. 2024. Adapting Large Language Models by Integrating Collaborative Semantics for Recommendation. In2024 IEEE 40th International Conference on Data Engineering (ICDE). IEEE, 1435–1448

    2024

  30. [38]

    Kai Zheng, Haijun Zhao, Rui Huang, Beichuan Zhang, Na Mou, Yanan Niu, Yang Song, Hongning Wang, and Kun Gai. 2024. Full Stage Learning to Rank: A Unified Framework for Multi-Stage Systems. InProceedings of the ACM Web Conference

    2024

  31. [39]

    Zuowu Zheng, Ze Wang, Fan Yang, Jiangke Fan, Teng Zhang, and Xingxing Wang. 2025. EGA: A Unified End-to-End Generative Framework for Industrial Advertising Systems.arXiv preprint arXiv:2505.17549(2025)

    2025 arXiv

  32. [40]

    Guorui Zhou, Jiaxin Deng, Jinghao Zhang, Kuo Cai, Lejian Ren, Qiang Luo, Qian- qian Wang, Qigen Hu, Rui Huang, Shiyao Wang, et al. 2025. OneRec Technical Report.arXiv preprint arXiv:2506.13695(2025)

    2025

  33. [41]

    Yongxin Zhu, Bocheng Li, Yifei Xin, Zhihua Xia, and Linli Xu. 2024. Addressing Representation Collapse in Vector Quantized Models with One Linear Layer. arXiv preprint arXiv:2411.02038(2024)

    2024

  34. [2020]

    InProceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining

    Learning to Generate Personalized Query Auto-Completions via a Multi- View Multi-Task Attentive Approach. InProceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2998–3007

  35. [2023]

    Recommender Systems with Generative Retrieval.Advances in Neural Information Processing Systems36 (2023), 10299–10315

    2023

  36. [2024]

    GPT-4 Technical Report.arXiv preprint arXiv:2303.08774(2024)

    2024 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.