pith. sign in

arxiv: 2606.00422 · v1 · pith:O3PCERUSnew · submitted 2026-05-29 · 💻 cs.IR · cs.LG

UniPinRec: Unifying Generative Retrieval and Ranking at Pinterest Scale

Hanyu Li , Yi-Ping Hsu , Aditya Mantha , Prabhat Agarwal , Laksh Bhasin , Jialu Wang , Hongtao Lin , Bella Huang
show 12 more authors
Yaxin Li Xinyi Li Chuxi Wang Kousik Rajesh Hooshmand Shokri Razaghi Shunyao Li Zongyue Qin Jaewon Yang James Li Dhruvil Deven Badani Jiajing Xu Charles Rosenberg
This is my paper

Pith reviewed 2026-06-28 20:28 UTC · model grok-4.3

classification 💻 cs.IR cs.LG
keywords recommendation systemsretrievalrankingtransformermodel unificationserving efficiencyuser behavior sequencesproduction deployment
0
0 comments X

The pith

A single transformer model with shared user-history encoding and task-specific heads can replace separate retrieval and ranking models in production.

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

The paper establishes that retrieval and ranking stages, which have long been trained and served separately despite using similar transformer encoders on user behavior data, can be merged into one input format, one model, and one training stage. A shared transformer produces candidate-independent representations from action sequences that then branch to an ANN dot-product head for retrieval and a cross-attention head for ranking. This unification rests on Masked Action Modeling to keep context lengths manageable, blended training examples that pair sequences with impression slates for both objectives, and cross-stage KV cache reuse to avoid recomputing user history. When deployed on Pinterest core surfaces the unified system records a 1% engagement increase together with an 11.1% drop in end-to-end latency and a 63.6% rise in queries per second. A reader would care because the approach removes duplicated parameters, compute, and serving stacks while preserving or improving accuracy inside an existing production infrastructure.

Core claim

UniPinRec shows that full-stack unification of retrieval and ranking is possible by encoding user action sequences once in a shared transformer, then routing the resulting representations through task-specific heads—one performing ANN dot-product retrieval and the other performing cross-attention ranking—while using Masked Action Modeling, blended training examples, and KV cache sharing to satisfy both objectives in a single training run and serving path, yielding measurable efficiency gains and a modest engagement lift when run at Pinterest scale.

What carries the argument

Shared transformer encoder producing candidate-independent representations that branch into task-specific heads for dot-product retrieval and cross-attention ranking, enabled by Masked Action Modeling.

If this is right

  • One model and one training stage replace two independent pipelines, removing duplicated parameters and compute.
  • Cross-stage KV cache sharing lowers total FLOPs relative to serving two separate models.
  • The system integrates into existing serving infrastructure without requiring new infrastructure.
  • Online A/B tests record roughly 1% higher engagement, 11.1% lower latency, and 63.6% higher QPS.

Where Pith is reading between the lines

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

  • The same blending and cache-sharing pattern could extend to additional recommendation stages such as re-ranking or exploration if their input representations overlap.
  • Blended training examples may allow new objectives like diversity or freshness to be added by simply changing the paired slates without retraining separate models.
  • If user-action sequences remain the dominant signal, the approach could reduce the total number of distinct models maintained across an entire recommendation stack.

Load-bearing premise

The shared transformer plus the three supporting techniques can meet both retrieval and ranking accuracy targets at production scale without any degradation that would erase the claimed efficiency savings.

What would settle it

A side-by-side offline evaluation showing whether the unified model’s retrieval recall or ranking NDCG falls below the levels achieved by the prior separate models on the same user-action data.

Figures

Figures reproduced from arXiv: 2606.00422 by Aditya Mantha, Bella Huang, Charles Rosenberg, Chuxi Wang, Dhruvil Deven Badani, Hanyu Li, Hongtao Lin, Hooshmand Shokri Razaghi, Jaewon Yang, James Li, Jiajing Xu, Jialu Wang, Kousik Rajesh, Laksh Bhasin, Prabhat Agarwal, Shunyao Li, Xinyi Li, Yaxin Li, Yi-Ping Hsu, Zongyue Qin.

Figure 1
Figure 1. Figure 1: Prior methods unify the architecture but leave input format, training, and serving fragmented. Our approach unifies [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: End-to-end flow of UniPinRec: a single shared backbone is trained jointly on retrieval and ranking objectives, then [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Co-serving topology for UniPinRec. Retrieval and ranking are independent Triton ensemble nodes; an ANN lookup [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Hit@3 vs. relative FLOPs for depth and sequence [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
read the original abstract

Modern recommendation systems predominantly train retrieval and ranking as separate models despite both increasingly relying on large transformers encoding the same user behavior data, duplicating parameters, compute, and serving cost. Prior work unifies the model architecture but not the full pipeline: input formats, training procedures, and serving stacks remain fragmented across stages. We present UniPinRec, which achieves full-stack unification of retrieval and ranking at Pinterest: one input format, one model, one training stage, deployed within existing serving infrastructure. A shared transformer encodes the user action sequence into candidate-independent representations that branch into retrieval (ANN dot-product) and ranking (cross-attention) via task-specific heads. Three ideas make this work: (1) Masked Action Modeling (MAM) eliminates interleaving, enabling weight sharing without doubling context length; (2) Blended training examples pair action sequences with feedview impression slates to satisfy both objectives jointly; (3) Cross-stage KV cache sharing reuses user-history computation from retrieval for ranking, reducing total FLOPs versus serving two independent models. Deployed in the Pinterest core surfaces, UniPinRec delivers approximately +1% online engagement lift while cutting end-to-end serving latency by 11.1% and lifting QPS by 63.6%. To our knowledge, this is the first full-stack unification of retrieval and ranking, covering inputs, model, training and serving, deployed in a production recommendation system.

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

1 major / 0 minor

Summary. The paper presents UniPinRec, a full-stack unification of retrieval and ranking for Pinterest-scale recommendation systems. It uses a single shared transformer that encodes user action sequences into candidate-independent representations, branching to retrieval via ANN dot-product and ranking via cross-attention with task-specific heads. Three enabling techniques are described: Masked Action Modeling (MAM) to avoid context doubling, blended training examples pairing action sequences with impression slates, and cross-stage KV cache sharing for serving efficiency. The system is deployed in core surfaces and reports approximately +1% online engagement lift, 11.1% end-to-end latency reduction, and 63.6% QPS increase.

Significance. If the reported production metrics are supported by controlled evidence, the work demonstrates that unifying input formats, model architecture, training, and serving infrastructure across retrieval and ranking stages can simultaneously improve engagement and reduce compute/serving costs in large-scale transformer-based recsys without apparent negative transfer.

major comments (1)
  1. Abstract: the central claims of +1% engagement lift, 11.1% latency reduction, and 63.6% QPS increase are presented as deployment outcomes, yet the provided manuscript supplies no experimental details, baselines, ablation studies, or error analysis to support them; this is load-bearing for assessing whether the unification actually delivers the stated gains without offsetting degradation.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the review and the opportunity to address the concern regarding experimental support for the reported production metrics. We respond to the major comment below.

read point-by-point responses

  1. Referee: [—] Abstract: the central claims of +1% engagement lift, 11.1% latency reduction, and 63.6% QPS increase are presented as deployment outcomes, yet the provided manuscript supplies no experimental details, baselines, ablation studies, or error analysis to support them; this is load-bearing for assessing whether the unification actually delivers the stated gains without offsetting degradation.

    Authors: The full manuscript contains a dedicated 'Online Experiments' section describing the A/B test setup (including cohort sampling, test duration, and statistical testing) used to measure the engagement lift, as well as production serving benchmarks for the latency and QPS improvements relative to the prior two-model baseline. We agree that the abstract would benefit from a brief pointer to this evaluation methodology. However, comprehensive ablation studies and error analyses are not feasible to publish at Pinterest scale without compromising production stability or revealing proprietary infrastructure details. We will revise the abstract to reference the experiments section and add a short summary paragraph on the evaluation protocol. revision: partial

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper describes a production-scale engineering system for unifying retrieval and ranking stages via a shared transformer, MAM, blended examples, and KV-cache reuse. All load-bearing claims are architectural descriptions or measured online A/B metrics (+1% engagement, latency/QPS improvements) rather than any derivation, fitted parameter, or prediction that reduces to its own inputs by construction. No equations, self-citations, or uniqueness theorems are invoked in a load-bearing way that would create circularity; the results are externally falsifiable deployment outcomes.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the approach appears to rest on standard transformer components and ANN indexing whose details are not supplied.

pith-pipeline@v0.9.1-grok · 5873 in / 1301 out tokens · 15538 ms · 2026-06-28T20:28:39.678230+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

39 extracted references · 25 canonical work pages · 3 internal anchors

  1. [1]

    Prabhat Agarwal, Anirudhan Badrinath, Laksh Bhasin, Jaewon Yang, Edoardo Botta, Jiajing Xu, and Charles Rosenberg. 2025. PinRec: Outcome-Conditioned, Multi-Token Generative Retrieval for Industry-Scale Recommendation Systems. arXiv preprint arXiv:2504.10507(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  2. [2]

    Prabhat Agarwal, Minhazul Islam SK, Nikil Pancha, Kurchi Subhra Hazra, Jiajing Xu, and Chuck Rosenberg. 2024. OmniSearchSage: Multi-Task Multi-Entity Embeddings for Pinterest Search. InCompanion Proceedings of the ACM Web Conference 2024 (WWW ’24). ACM, 121–130. doi:10.1145/3589335.3648309

    work page doi:10.1145/3589335.3648309 2024

  3. [3]

    Anirudhan Badrinath, Alex Yang, Kousik Rajesh, Prabhat Agarwal, Jaewon Yang, Haoyu Chen, Jiajing Xu, and Charles Rosenberg. 2025. OmniSage: Large Scale, Multi-Entity Heterogeneous Graph Representation Learning. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (Toronto ON, Canada)(KDD ’25). Association for Computin...

    work page doi:10.1145/3711896.3737253 2025

  4. [4]

    Josh Beal, Eric Kim, Jinfeng Rao, Rex Wu, Dmitry Kislyuk, and Charles Rosenberg

  5. [5]

    arXiv:2603.03544 [cs.CV] https://arxiv.org/abs/2603.03544

    PinCLIP: Large-scale Foundational Multimodal Representation at Pinterest. arXiv:2603.03544 [cs.CV] https://arxiv.org/abs/2603.03544

    work page arXiv

  6. [6]

    Yang Cao, Changhao Zhang, Xiaoshuang Chen, Kaiqiao Zhan, and Ben Wang

  7. [7]

    InProceedings of the ACM Web Conference 2025 (WWW)

    xMTF: A Formula-Free Model for Reinforcement-Learning-Based Multi- Task Fusion in Recommender Systems. InProceedings of the ACM Web Conference 2025 (WWW)

    2025

  8. [8]

    Jiahui Chen, Xiaoze Jiang, Zhibo Wang, Quanzhi Zhu, Junyao Zhao, Feng Hu, Kang Pan, Ao Xie, Maohua Pei, Zhiheng Qin, Hongjing Zhang, Zhixin Zhai, Xiaobo Guo, Runbin Zhou, Kefeng Wang, Mingyang Geng, Cheng Chen, Jingshan Lv, Yupeng Huang, Xiao Liang, and Han Li. 2025. UniSearch: Rethinking Search System with a Unified Generative Architecture.arXiv preprint...

    work page arXiv 2025

  9. [9]

    Xiangyi Chen, Kousik Rajesh, Matthew Lawhon, Zelun Wang, Hanyu Li, Haomiao Li, Saurabh Vishwas Joshi, Pong Eksombatchai, Jaewon Yang, Yi-Ping Hsu, Jiajing Xu, and Charles Rosenberg. 2025. PinFM: Foundation Model for User Activity Sequences at a Billion-scale Visual Discovery Platform. InProceedings of the 19th ACM Conference on Recommender Systems (RecSys...

    work page doi:10.1145/3705328 2025

  10. [10]

    Lee, Khush- hall Chandra Mahajan, Ning Jiang, Kai Ren, Jinhui Li, and Wen-Yun Yang

    Zhimin Chen, Chenyu Zhao, Ka Chun Mo, Yunjiang Jiang, Jane H. Lee, Khush- hall Chandra Mahajan, Ning Jiang, Kai Ren, Jinhui Li, and Wen-Yun Yang. 2026. Massive Memorization with Hundreds of Trillions of Parameters for Sequen- tial Transducer Generative Recommenders. InProceedings of the International Conference on Learning Representations (ICLR)

    2026

  11. [11]

    Sunhao Dai, Jiakai Tang, Jiahua Wu, Kun Wang, Yuxuan Zhu, Bingjun Chen, Bangyang Hong, Yu Zhao, Cong Fu, Kangle Wu, Yabo Ni, Anxiang Zeng, Wenjie Wang, Xu Chen, Jun Xu, and See-Kiong Ng. 2025. OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking System.arXiv preprint arXiv:2509.18091(2025)

    work page arXiv 2025

  12. [12]

    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

  13. [13]

    Yijie Ding, Zitian Guo, Jiacheng Li, Letian Peng, Shuai Shao, Wei Shao, Xiaoqiang Luo, Luke Simon, Jingbo Shang, Julian McAuley, and Yupeng Hou. 2026. How Well Does Generative Recommendation Generalize? arXiv:2603.19809 [cs.IR] https://arxiv.org/abs/2603.19809

    work page arXiv 2026

  14. [14]

    Gao, Chen Xue, Marc Versage, Xie Zhou, Zhongruo Wang, Chao Li, Yeon Seonwoo, Nan Chen, Zhen Ge, Gourab Kundu, Weiqi Zhang, Tian Wang, Qingjun Cui, and Trishul Chilimbi

    Vianne R. Gao, Chen Xue, Marc Versage, Xie Zhou, Zhongruo Wang, Chao Li, Yeon Seonwoo, Nan Chen, Zhen Ge, Gourab Kundu, Weiqi Zhang, Tian Wang, Qingjun Cui, and Trishul Chilimbi. 2025. SynerGen: Contextualized Genera- tive Recommender for Unified Search and Recommendation.arXiv preprint arXiv:2509.21777(2025)

    work page arXiv 2025

  15. [15]

    Ruidong Han, Bin Yin, Shangyu Chen, He Jiang, Fei Jiang, Xiang Li, Chi Ma, Mincong 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. InProceedings of the 34th ACM Interna- tional Conference on Information and Knowledge Management (CI...

    work page doi:10.1145/3746252.3761565 2025

  16. [16]

    Horace He et al. 2024. Flex Attention: A Programming Model for Generating Optimized Attention Kernels.arXiv preprint arXiv:2412.05496(2024). https: //arxiv.org/abs/2412.05496

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  17. [17]

    Lars Hertel, Neil Daftary, Fedor Borisyuk, Aman Gupta, and Rahul Mazumder

  18. [18]

    InCompanion Proceedings of the ACM Web Confer- ence 2025 (WWW)

    Efficient User History Modeling with Amortized Inference for Deep Learn- ing Recommendation Models. InCompanion Proceedings of the ACM Web Confer- ence 2025 (WWW)

    2025

  19. [19]

    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. CoRRabs/2505.04180 (2025). doi:10.48550/ARXIV.2505.04180 arXiv:2505.04180

    work page doi:10.48550/arxiv.2505.04180 2025

  20. [20]

    Jordan, and Ion Stoica

    Philipp Moritz, Robert Nishihara, Stephanie Wang, Alexey Tumanov, Richard Liaw, Eric Liang, Melih Elibol, Zongheng Yang, William Paul, Michael I. Jordan, and Ion Stoica. 2018. Ray: A Distributed Framework for Emerging AI Applications. In13th USENIX Symposium on Operating Systems Design and Implementation (OSDI 18). 561–577

    2018

  21. [21]

    David Pardoe, Neil Daftary, Miro Furtado, Aditya Aiyer, Yu Wang, Liuqing Li, Tao Song, Lars Hertel, Young Jin Yun, Senthil Radhakrishnan, Zhiwei Wang, Tommy Li, Khai Tran, Ananth Nagarajan, Ali Naqvi, Yue Zhang, Renpeng Fang, Avi Romascanu, Arjun Kulothungun, Deepak Kumar, Praneeth Boda, Fedor Borisyuk, and Ruoyan Wang. 2026. CADET: Context-Conditioned Ad...

    work page arXiv 2026

  22. [22]

    Zhengyang Su, Isay Katsman, Yueqi Wang, Ruining He, Lukasz Heldt, Raghu- nandan Keshavan, Shao-Chuan Wang, Xinyang Yi, Mingyan Gao, Onkar Dalal, Lichan Hong, Ed Chi, and Ningren Han. 2026. STATIC: Vectorizing the Trie: Effi- cient Constrained Decoding for LLM-based Generative Retrieval on Accelerators. arXiv preprint arXiv:2602.22647(2026)

    work page arXiv 2026

  23. [23]

    Dekai Sun, Yiming Liu, Jiafan Zhou, Xun Liu, Chenchen Yu, Yi Li, Jun Zhang, Huan Yu, and Jie Jiang. 2026. OneRanker: Unified Generation and Ranking with One Model in Industrial Advertising Recommendation.arXiv preprint arXiv:2603.02999(2026)

    work page arXiv 2026

  24. [24]

    Yijia Sun, Shanshan Huang, Zhiyuan Guan, Qiang Luo, Ruiming Tang, Kun Gai, and Guorui Zhou. 2026. GRank: Towards Target-Aware and Streamlined Industrial Retrieval with a Generate-Rank Framework. InProceedings of the ACM Web Conference 2026 (WWW)

    2026

  25. [25]

    Jiarui Wang, Huichao Chai, Yuanhang Zhang, Zongjin Zhou, Wei Guo, Xingkun Yang, Qiang Tang, Bo Pan, Jiawei Zhu, Ke Cheng, Yuting Yan, Shulan Wang, Yingjie Zhu, Zhengfan Yuan, Jiaqi Huang, Yuhan Zhang, Xiaosong Sun, Zhinan Zhang, Hong Zhu, Yongsheng Zhang, Tiantian Dong, Zhong Xiao, Deliang Liu, Chengzhou Lu, Yuan Sun, Zhiyuan Chen, Xinming Han, Zaizhu Liu...

    work page arXiv 2026

  26. [26]

    Xue Xia, Pong Eksombatchai, Nikil Pancha, Dhruvil Deven Badani, Po-Wei Wang, Neng Gu, Saurabh Vishwas Joshi, Nazanin Farahpour, Zhiyuan Zhang, and An- drew Zhai. 2023. TransAct: Transformer-based Realtime User Action Model for Recommendation at Pinterest. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2023, Lon...

    work page doi:10.1145/3580305.3599918 2023

  27. [27]

    Xue Xia, Saurabh Joshi, Kousik Rajesh, Kangnan Li, Yangyi Lu, Nikil Pancha, Dhruvil Badani, Jiajing Xu, and Pong Eksombatchai. 2025. TransAct V2: Lifelong User Action Sequence Modeling on Pinterest Recommendation. InProceedings of the 34th ACM International Conference on Information and Knowledge Management (Seoul, Republic of Korea)(CIKM ’25). Associatio...

    work page doi:10.1145/3746252.3761433 2025

  28. [28]

    Bencheng Yan, Shilei Liu, Zhiyuan Zeng, Zihao Wang, Yizhen Zhang, Yujin Yuan, Langming Liu, Jiaqi Liu, Di Wang, Wenbo Su, Pengjie Wang, Jian Xu, and Bo Zheng. 2026. Unlocking Scaling Law in Industrial Recommendation Systems with a Three-step Paradigm based Large User Model. InProceedings of the 19th ACM International Conference on Web Search and Data Mini...

    2026

  29. [29]

    Huimin Yan, Longfei Xu, Junjie Sun, Ni Ou, Wei Luo, Xing Tan, Ran Cheng, Kaikui Liu, and Xiangxiang Chu. 2025. IntSR: An Integrated Generative Framework for Search and Recommendation.arXiv preprint arXiv:2509.21179(2025)

    work page arXiv 2025

  30. [30]

    Nowak, Xiaoli Gao, and Hamid Eghbalzadeh

    Liu Yang, Fabian Paischer, Kaveh Hassani, Jiacheng Li, Shuai Shao, Zhang Gabriel Li, Yun He, Xue Feng, Nima Noorshams, Sem Park, Bo Long, Robert D. Nowak, Xiaoli Gao, and Hamid Eghbalzadeh. 2024. Unifying Generative and Dense Retrieval for Sequential Recommendation.arXiv preprint arXiv:2411.18814(2024)

    work page arXiv 2024

  31. [31]

    Xiao Yang, Peifeng Yin, Abe Engle, Jinfeng Zhuang, and Ling Leng. 2025. MTMD: A Multi-Task Multi-Domain Framework for Unified Ad Lightweight Ranking at Pinterest. InProceedings of the AdKDD Workshop at the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining

    2025

  32. [32]

    Yufei Ye, Wei Guo, Jin Yao Chin, Hao Wang, Hong Zhu, Xi Lin, Yuyang Ye, Yong Liu, Ruiming Tang, Defu Lian, and Enhong Chen. 2025. FuXi-𝛼: Scaling Recommendation Model with Feature Interaction Enhanced Transformer. In Proceedings of the ACM Web Conference 2025 (WWW)

    2025

  33. [33]

    Yufei Ye, Wei Guo, Hao Wang, Hong Zhu, Yuyang Ye, Yong Liu, Huifeng Guo, Ruiming Tang, Defu Lian, and Enhong Chen. 2025. FuXi- 𝛽: Towards a Light- weight and Fast Large-Scale Generative Recommendation Model.arXiv preprint arXiv:2508.10615(2025)

    work page arXiv 2025

  34. [34]

    Jun Yuan, Guohao Cai, and Zhenhua Dong. 2024. A Parameter Update Balanc- ing Algorithm for Multi-task Ranking Models in Recommendation Systems. In Proceedings of the 2024 IEEE International Conference on Data Mining (ICDM)

    2024

  35. [35]

    Jiaqi Zhai, Lucy Liao, Xing Liu, Yueming Wang, Rui Li, Xuan Cao, Leon Gao, Zhaojie Gong, Fangda Gu, Michael He, Yinghai Lu, and Yu Shi. 2024. Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Gener- ative Recommendations. InProceedings of the 41st International Conference on Machine Learning (ICML)

    2024

  36. [36]

    Jun Zhang, Yi Li, Yue Liu, Changping Wang, Yuan Wang, Yuling Xiong, Xun Liu, Haiyang Wu, Qian Li, Enming Zhang, Jiawei Sun, Xin Xu, Zishuai Zhang, Ruoran Liu, Suyuan Huang, Zhaoxin Zhang, Zhengkai Guo, Shuojin Yang, Meng-Hao Guo, Huan Yu, Jie Jiang, and Shi-Min Hu. 2025. GPR: Towards a Generative Pre-trained One-Model Paradigm for Large-Scale Advertising ...

    work page arXiv 2025

  37. [37]

    Luankang Zhang, Kenan Song, Yi Quan Lee, Wei Guo, Hao Wang, Yawen Li, Huifeng Guo, Yong Liu, Defu Lian, and Enhong Chen. 2025. Killing Two Birds with One Stone: Unifying Retrieval and Ranking with a Single Generative Rec- ommendation Model. InProceedings of the 48th International ACM SIGIR Confer- ence on Research and Development in Information Retrieval(...

    work page doi:10.1145/3726302.3730017 2025

  38. [38]

    Yukun Zhang, Si Dong, Xu Wang, Bo Chen, Qinglin Jia, Shengzhe Wang, Jinlong Jiao, Runhan Li, Jiaqing Liu, Chaoyi Ma, Ruiming Tang, Guorui Zhou, Han Li, and Kun Gai. 2026. SMES: Towards Scalable Multi-Task Recommendation via Expert Sparsity.arXiv preprint arXiv:2602.09386(2026)

    work page arXiv 2026

  39. [39]

    Yanyan Zou, Junbo Qi, Lunsong Huang, Yu Li, Kewei Xu, Jiabao Gao, Binglei Zhao, Xuanhua Yang, Sulong Xu, and Shengjie Li. 2026. GenRec: A Preference- Oriented Generative Framework for Large-Scale Recommendation. InProceedings of the 49th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR). 10

    2026