pith. sign in

arxiv: 2509.13648 · v3 · pith:5A3POPVVnew · submitted 2025-09-17 · 💻 cs.LG · cs.IR

Sequential Data Augmentation for Generative Recommendation

Geon Lee , Bhuvesh Kumar , Clark Mingxuan Ju , Tong Zhao , Kijung Shin , Neil Shah , Liam Collins This is my paper

Pith reviewed 2026-05-21 22:39 UTC · model grok-4.3

classification 💻 cs.LG cs.IR
keywords data augmentationgenerative recommendationsequential datastochastic samplingtraining distributiongeneralizationbias controluser interaction sequences
0
0 comments X

The pith

GenPAS models data augmentation for generative recommendation as stochastic sampling over input-target pairs with three bias-controlled steps.

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

Generative recommendation models predict future user interactions from historical sequences, yet the construction of training examples from those histories strongly shapes model generalization. Experiments show that common augmentation choices produce large differences in performance by altering how well training data aligns with future targets and unseen inputs. GenPAS unifies these choices by representing augmentation as a stochastic process with three explicit steps—sequence sampling, target sampling, and input sampling—each equipped with bias controls. This structure lets designers adjust the resulting training distribution in a principled way rather than relying on ad-hoc rules. On both public benchmarks and industrial data, the approach delivers higher accuracy while using fewer examples and smaller models.

Core claim

The paper claims that data augmentation in generative recommendation can be formalized as a stochastic sampling process over input-target pairs governed by three bias-controlled operations—sequence sampling, target sampling, and input sampling—thereby recovering prior strategies as special cases and producing training distributions that support stronger alignment with future targets and better generalization to unseen inputs.

What carries the argument

GenPAS framework, which models augmentation as a stochastic sampling process over input-target pairs with three explicit bias-controlled steps.

If this is right

  • GenPAS produces higher accuracy than existing augmentation strategies on both benchmark and industrial datasets.
  • The same performance level is reached with smaller training sets, improving data efficiency.
  • Models trained under GenPAS require fewer parameters while maintaining accuracy, improving parameter efficiency.
  • Existing augmentation methods appear as special cases inside the three-step formulation.
  • Designers gain explicit levers to shape the training distribution rather than treating augmentation as a minor implementation detail.

Where Pith is reading between the lines

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

  • The same sampling view could be tested on other sequential prediction tasks such as next-item prediction in non-recommendation domains.
  • Explicit bias controls may offer a route to mitigate popularity bias or long-tail effects without changing the underlying model architecture.
  • The framework could be extended with learned or adaptive sampling probabilities that depend on the current model state during training.

Load-bearing premise

Modeling augmentation as a stochastic sampling process over input-target pairs with three explicit bias-controlled steps is sufficient to capture and control the key factors that determine generalization in generative recommendation models.

What would settle it

A new augmentation procedure that cannot be expressed inside the three-step sampling model yet still produces higher accuracy, data efficiency, and parameter efficiency than GenPAS on the same benchmark and industrial datasets would falsify the claim of sufficiency.

Figures

Figures reproduced from arXiv: 2509.13648 by Bhuvesh Kumar, Clark Mingxuan Ju, Geon Lee, Kijung Shin, Liam Collins, Neil Shah, Tong Zhao.

Figure 1
Figure 1. Figure 1: Different strategies yield distinct target distribu [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Different strategies produce distinct input-target [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The two parameters, 𝛼 and 𝛽, jointly shape the training distribution and have a substantial impact on model performance. Their impact patterns differ across datasets. (𝛼, 𝛽) = (0, ∞) [49], our results show performance varies with the target distribution, emphasizing the need for broader evaluation. Effectiveness of Components. We examine the impact of Gen￾PAS ’s key parameters [PITH_FULL_IMAGE:figures/ful… view at source ↗
Figure 4
Figure 4. Figure 4: GenPAS enhances the data efficiency. SASRec with GenPAS outperforms the full-data baseline without augmen￾tation, even when trained on 5, 10, 20% of the original data. on the test targets from a random subset of 200k training users (transductive setting) and a set of 200k users unseen during train￾ing (inductive setting). We randomly sample 2000 negative items for evaluation in both settings, and compare a… view at source ↗
Figure 8
Figure 8. Figure 8: SASRec trained on GenPAS-augmented data outper￾forms larger-parameter variants trained with other strate￾gies, demonstrating GenPAS’s parameter efficiency. Replace. Replaces the item at a randomly selected position 𝑘 ∗ ∈ {1, . . . , |𝑠 (𝑢) | − 1} with a randomly sampled item 𝑖 ∗ ∈ I: 𝑥˜ (𝑢) = h 𝑖 (𝑢) 1 , . . . ,𝑖 (𝑢) 𝑘 ∗−1 , 𝑖∗ , 𝑖 (𝑢) 𝑘 ∗ , . . . ,𝑖 (𝑢) |𝑠 (𝑢) |−1 i . Reorder. Shuffles a contiguous subseq… view at source ↗
Figure 7
Figure 7. Figure 7: The performance (NDCG@10) of SASRec tends to [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
read the original abstract

Generative recommendation plays a crucial role in personalized systems, predicting users' future interactions from their historical behavior sequences. A critical yet underexplored factor in training these models is data augmentation, the process of constructing training data from user interaction histories. By shaping the training distribution, data augmentation directly and often substantially affects model generalization and performance. Nevertheless, in much of the existing work, this process is simplified, applied inconsistently, or treated as a minor design choice, without a systematic and principled understanding of its effects. Motivated by our empirical finding that different augmentation strategies can yield large performance disparities, we conduct an in-depth analysis of how they reshape training distributions and influence alignment with future targets and generalization to unseen inputs. To systematize this design space, we propose GenPAS, a generalized and principled framework that models augmentation as a stochastic sampling process over input-target pairs with three bias-controlled steps: sequence sampling, target sampling, and input sampling. This formulation unifies widely used strategies as special cases and enables flexible control of the resulting training distribution. Our extensive experiments on benchmark and industrial datasets demonstrate that GenPAS yields superior accuracy, data efficiency, and parameter efficiency compared to existing strategies, providing practical guidance for principled training data construction in generative recommendation. Our code is available at https://github.com/snap-research/GenPAS.

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 / 2 minor

Summary. The manuscript proposes GenPAS, a generalized framework for data augmentation in generative recommendation. It models augmentation as a stochastic sampling process over input-target pairs consisting of three explicit bias-controlled steps (sequence sampling, target sampling, and input sampling). This formulation unifies existing strategies as special cases and is shown via experiments on benchmark and industrial datasets to deliver superior accuracy, data efficiency, and parameter efficiency.

Significance. If the empirical results hold under rigorous controls, the work offers a principled way to shape training distributions in generative recommendation, addressing an underexplored factor that substantially affects generalization. The open release of code supports reproducibility and practical adoption.

major comments (2)
  1. [§4.2] §4.2 (Experimental Setup): The description of baseline re-implementations and whether augmentation hyperparameters were tuned under identical computational budgets is insufficient to support the claim of fair superiority in accuracy and efficiency; without these details the reported gains cannot be confidently attributed to the GenPAS structure.
  2. [§5.3] §5.3 (Ablation Studies): No controlled ablation fixes one step (e.g., uniform target sampling without the proposed bias control) while keeping others fixed; this leaves open whether the three-step framework itself is load-bearing or whether results are driven primarily by tuned sampling probabilities.
minor comments (2)
  1. [§3] Notation for the bias parameters in the three sampling steps could be introduced earlier and used consistently to improve readability of the unification argument.
  2. [Figure 3] Figure 3 (training distribution visualizations) would benefit from explicit labels indicating the bias values used in each panel.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and insightful comments. We address each major point below, providing clarifications and committing to revisions that strengthen the experimental rigor and ablation analysis.

read point-by-point responses

  1. Referee: [§4.2] §4.2 (Experimental Setup): The description of baseline re-implementations and whether augmentation hyperparameters were tuned under identical computational budgets is insufficient to support the claim of fair superiority in accuracy and efficiency; without these details the reported gains cannot be confidently attributed to the GenPAS structure.

    Authors: We agree that the current description in §4.2 lacks sufficient detail on baseline re-implementations and hyperparameter tuning procedures. In the revised manuscript we will expand this section to explicitly document: (1) the precise augmentation strategies used for each baseline (including the specific sampling probabilities and bias controls applied), (2) the hyperparameter search spaces, and (3) confirmation that all methods—including baselines—were tuned under identical computational budgets via grid search over equivalent ranges of augmentation parameters, with final configurations selected by validation performance. These additions will allow readers to confidently attribute performance differences to the GenPAS framework. revision: yes

  2. Referee: [§5.3] §5.3 (Ablation Studies): No controlled ablation fixes one step (e.g., uniform target sampling without the proposed bias control) while keeping others fixed; this leaves open whether the three-step framework itself is load-bearing or whether results are driven primarily by tuned sampling probabilities.

    Authors: We thank the referee for this observation. To isolate the contribution of each step, we have conducted new controlled ablations in which sequence sampling and input sampling are held fixed at their GenPAS settings while varying only the target sampling step. In particular, we compare uniform target sampling (no bias control) against the proposed bias-controlled target sampling. The additional results show that bias-controlled target sampling yields consistent gains beyond those achievable by simply tuning sampling probabilities, confirming that the three-step structure is load-bearing. These new experiments and analysis will be added to the revised §5.3. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical framework with external validation

full rationale

The paper proposes GenPAS as a modeling framework for data augmentation in generative recommendation, representing it as a stochastic sampling process with three bias-controlled steps (sequence, target, and input sampling) that unifies prior strategies as special cases. Claims of superior accuracy, data efficiency, and parameter efficiency rest on experiments across benchmark and industrial datasets rather than any closed-form derivation. No equations or steps reduce by construction to fitted parameters, self-definitions, or load-bearing self-citations; the central contribution is an empirical systematization validated externally. The derivation chain is self-contained against independent benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on standard machine-learning assumptions about sequence data and generalization; no new physical entities or heavily fitted constants are introduced beyond typical hyperparameters.

axioms (1)
  • domain assumption User interaction histories can be treated as sequences from which input-target pairs can be sampled to form training examples.
    Invoked when defining the three-step sampling process in the abstract.

pith-pipeline@v0.9.0 · 5781 in / 1186 out tokens · 35468 ms · 2026-05-21T22:39:32.426973+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

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

  1. Pay Attention to Sequence Split: Uncovering the Impacts of Sub-Sequence Splitting on Sequential Recommendation Models

    cs.IR 2026-04 conditional novelty 6.0

    Sub-sequence splitting interferes with fair evaluation in sequential recommendation models and enhances performance only when paired with particular splitting, targeting, and loss function choices.

Reference graph

Works this paper leans on

57 extracted references · 57 canonical work pages · cited by 1 Pith paper · 4 internal anchors

  1. [1]

    Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. 2019. Optuna: A Next-Generation Hyperparameter Optimization Frame- work. InKDD

    work page 2019

  2. [2]

    Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenbin Ge, Yu Han, Fei Huang, et al. 2023. Qwen technical report.arXiv preprint arXiv:2309.16609(2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  3. [3]

    James Bergstra, Daniel Yamins, and David Cox. 2013. Making a science of model search: Hyperparameter optimization in hundreds of dimensions for vision architectures. InICML

    work page 2013

  4. [4]

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners.NeurIPS(2020)

    work page 2020

  5. [5]

    Jianxin Chang, Chen Gao, Yu Zheng, Yiqun Hui, Yanan Niu, Yang Song, Depeng Jin, and Yong Li. 2021. Sequential recommendation with graph neural networks. InSIGIR

    work page 2021

  6. [6]

    Runjin Chen, Mingxuan Ju, Ngoc Bui, Dimosthenis Antypas, Stanley Cai, Xi- aopeng Wu, Leonardo Neves, Zhangyang Wang, Neil Shah, and Tong Zhao

    work page

  7. [7]

    Enhancing item tokenization for generative recommendation through self-improvement.arXiv preprint arXiv:2412.17171(2024)

    work page arXiv 2024

  8. [8]

    Yongjun Chen, Zhiwei Liu, Jia Li, Julian McAuley, and Caiming Xiong. 2022. Intent contrastive learning for sequential recommendation. InWWW

    work page 2022

  9. [9]

    Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Yunxuan Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, et al

    work page

  10. [10]

    Scaling instruction-finetuned language models.JMLR25, 70 (2024), 1–53

    work page 2024

  11. [11]

    Yizhou Dang, Yuting Liu, Enneng Yang, Minhan Huang, Guibing Guo, Jianzhe Zhao, and Xingwei Wang. 2025. Data augmentation as free lunch: Exploring the test-time augmentation for sequential recommendation. InSIGIR

    work page 2025

  12. [12]

    Yizhou Dang, Enneng Yang, Yuting Liu, Guibing Guo, Linying Jiang, Jianzhe Zhao, and Xingwei Wang. 2024. Data Augmentation for Sequential Recommendation: A Survey.arXiv(2024)

    work page 2024

  13. [13]

    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

  14. [14]

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xi- aohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale.arXiv preprint arXiv:2010.11929(2020)

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  15. [15]

    F Maxwell Harper and Joseph A Konstan. 2015. The movielens datasets: History and context.Acm transactions on interactive intelligent systems (tiis)5, 4 (2015), 1–19

    work page 2015

  16. [16]

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick

    work page

  17. [17]

    Masked autoencoders are scalable vision learners. InCVPR

    work page

  18. [18]

    Ruining He and Julian McAuley. 2016. Fusing similarity models with markov chains for sparse sequential recommendation. InICDM

    work page 2016

  19. [19]

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

    work page

  20. [20]

    Session-based recommendations with recurrent neural networks.arXiv preprint arXiv:1511.06939(2015)

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  21. [21]

    Juyong Jiang, Peiyan Zhang, Yingtao Luo, Chaozhuo Li, Jae Boum Kim, Kai Zhang, Senzhang Wang, Xing Xie, and Sunghun Kim. 2023. AdaMCT: adaptive mixture of CNN-transformer for sequential recommendation. InCIKM. Geon Lee, Bhuvesh Kumar, Clark Mingxuan Ju, Tong Zhao, Kijung Shin, Neil Shah, and Liam Collins

    work page 2023

  22. [22]

    Bowen Jin, Hansi Zeng, Guoyin Wang, Xiusi Chen, Tianxin Wei, Ruirui Li, Zhengyang Wang, Zheng Li, Yang Li, Hanqing Lu, et al. 2023. Language models as semantic indexers.arXiv preprint arXiv:2310.07815(2023)

    work page arXiv 2023

  23. [23]

    Clark Mingxuan Ju, Liam Collins, Leonardo Neves, Bhuvesh Kumar, Louis Yufeng Wang, Tong Zhao, and Neil Shah. 2025. Generative Recommendation with Seman- tic IDs: A Practitioner’s Handbook. InProceedings of the 34th ACM International Conference on Information and Knowledge Management (CIKM)

    work page 2025

  24. [24]

    Clark Mingxuan Ju, Liam Collins, Leonardo Neves, Bhuvesh Kumar, Louis Yufeng Wang, Tong Zhao, and Neil Shah. 2025. Generative Recommendation with Semantic IDs: A Practitioner’s Handbook.arXiv preprint arXiv:2507.22224(2025)

    work page arXiv 2025

  25. [25]

    Wang-Cheng Kang and Julian McAuley. 2018. Self-attentive sequential recom- mendation. InICDM

    work page 2018

  26. [26]

    Sungyeon Kim, Xinliang Zhu, Xiaofan Lin, Muhammet Bastan, Doug Gray, and Suha Kwak. 2025. GENIUS: A generative framework for universal multimodal search. (2025). https://www.amazon.science/publications/genius-a-generative- framework-for-universal-multimodal-search

    work page 2025

  27. [27]

    VI Lcvenshtcin. 1966. Binary coors capable or ‘correcting deletions, insertions, and reversals. InSoviet physics-doklady, Vol. 10

    work page 1966

  28. [28]

    Sunkyung Lee, Minjin Choi, Eunseong Choi, Hye-young Kim, and Jongwuk Lee

    work page

  29. [29]

    GRAM: Generative Recommendation via Semantic-aware Multi-granular Late Fusion.arXiv preprint arXiv:2506.01673(2025)

    work page arXiv 2025

  30. [30]

    Jiacheng Li, Yujie Wang, and Julian McAuley. 2020. Time interval aware self- attention for sequential recommendation. InWSDM

    work page 2020

  31. [31]

    Piji Li, Zihao Wang, Zhaochun Ren, Lidong Bing, and Wai Lam. 2017. Neural rating regression with abstractive tips generation for recommendation. InSIGIR

    work page 2017

  32. [32]

    Jianghao Lin, Xinyi Dai, Yunjia Xi, Weiwen Liu, Bo Chen, Hao Zhang, Yong Liu, Chuhan Wu, Xiangyang Li, Chenxu Zhu, et al . 2025. How can recommender systems benefit from large language models: A survey.TOIS43, 2 (2025), 1–47

    work page 2025

  33. [33]

    Zhiwei Liu, Yongjun Chen, Jia Li, Philip S Yu, Julian McAuley, and Caiming Xiong. 2021. Contrastive self-supervised sequential recommendation with robust augmentation.arXiv(2021)

    work page 2021

  34. [34]

    Julian McAuley, Christopher Targett, Qinfeng Shi, and Anton Van Den Hengel

    work page

  35. [35]

    Image-based recommendations on styles and substitutes. InSIGIR

    work page

  36. [36]

    Aleksandr V Petrov and Craig Macdonald. 2023. Generative sequential recom- mendation with gptrec.arXiv preprint arXiv:2306.11114(2023)

    work page arXiv 2023

  37. [37]

    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

  38. [38]

    Recommender systems with generative retrieval.NeurIPS(2023)

    work page 2023

  39. [39]

    Steffen Rendle, Christoph Freudenthaler, and Lars Schmidt-Thieme. 2010. Factor- izing personalized markov chains for next-basket recommendation. InWWW

    work page 2010

  40. [40]

    Anima Singh, Trung Vu, Nikhil Mehta, Raghunandan Keshavan, Maheswaran Sathiamoorthy, Yilin Zheng, Lichan Hong, Lukasz Heldt, Li Wei, Devansh Tandon, et al. 2024. Better generalization with semantic ids: A case study in ranking for recommendations. InProceedings of the 18th ACM Conference on Recommender Systems. 1039–1044

    work page 2024

  41. [41]

    Joo-yeong Song and Bongwon Suh. 2022. Data augmentation strategies for improving sequential recommender systems.arXiv(2022)

    work page 2022

  42. [42]

    Fei Sun, Jun Liu, Jian Wu, Changhua Pei, Xiao Lin, Wenwu Ou, and Peng Jiang

    work page

  43. [43]

    InProceedings of the 28th ACM international conference on information and knowledge management

    BERT4Rec: Sequential recommendation with bidirectional encoder rep- resentations from transformer. InProceedings of the 28th ACM international conference on information and knowledge management. 1441–1450

    work page

  44. [44]

    Jiaxi Tang and Ke Wang. 2018. Personalized top-n sequential recommendation via convolutional sequence embedding. InWSDM

    work page 2018

  45. [45]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need.Advances in neural information processing systems30 (2017)

    work page 2017

  46. [46]

    2018.High-dimensional probability: An introduction with applications in data science

    Roman Vershynin. 2018.High-dimensional probability: An introduction with applications in data science. Vol. 47. Cambridge university press

    work page 2018

  47. [47]

    Nan Wang, Shoujin Wang, Yan Wang, Quan Z Sheng, and Mehmet Orgun. 2020. Modelling local and global dependencies for next-item recommendations. In WISE

    work page 2020

  48. [48]

    Shoujin Wang, Liang Hu, Yan Wang, Quan Z Sheng, Mehmet Orgun, and Long- bing Cao. 2020. Intention nets: psychology-inspired user choice behavior model- ing for next-basket prediction. InAAAI

    work page 2020

  49. [49]

    Zhenlei Wang, Jingsen Zhang, Hongteng Xu, Xu Chen, Yongfeng Zhang, Wayne Xin Zhao, and Ji-Rong Wen. 2021. Counterfactual data-augmented se- quential recommendation. InSIGIR

    work page 2021

  50. [50]

    Xu Xie, Fei Sun, Zhaoyang Liu, Shiwen Wu, Jinyang Gao, Jiandong Zhang, Bolin Ding, and Bin Cui. 2022. Contrastive learning for sequential recommendation. InICDE

    work page 2022

  51. [51]

    Liu Yang, Fabian Paischer, Kaveh Hassani, Jiacheng Li, Shuai Shao, Zhang Gabriel Li, Yun He, Xue Feng, Nima Noorshams, Sem Park, et al . 2024. Unifying gen- erative and dense retrieval for sequential recommendation.arXiv preprint arXiv:2411.18814(2024)

    work page arXiv 2024

  52. [52]

    Li Yujian and Liu Bo. 2007. A normalized Levenshtein distance metric.TPAMI 29, 6 (2007), 1091–1095

    work page 2007

  53. [53]

    Shengyu Zhang, Dong Yao, Zhou Zhao, Tat-Seng Chua, and Fei Wu. 2021. Causerec: Counterfactual user sequence synthesis for sequential recommen- dation. InSIGIR

    work page 2021

  54. [54]

    Wayne Xin Zhao, Yupeng Hou, Xingyu Pan, Chen Yang, Zeyu Zhang, Zihan Lin, Jingsen Zhang, Shuqing Bian, Jiakai Tang, Wenqi Sun, et al. 2022. Recbole 2.0: Towards a more up-to-date recommendation library. InCIKM

    work page 2022

  55. [55]

    Zihuai Zhao, Wenqi Fan, Jiatong Li, Yunqing Liu, Xiaowei Mei, Yiqi Wang, Zhen Wen, Fei Wang, Xiangyu Zhao, Jiliang Tang, et al. 2024. Recommender systems in the era of large language models (llms).TKDE36, 11 (2024), 6889–6907

    work page 2024

  56. [56]

    Kun Zhou, Hui Wang, Wayne Xin Zhao, Yutao Zhu, Sirui Wang, Fuzheng Zhang, Zhongyuan Wang, and Ji-Rong Wen. 2020. S3-rec: Self-supervised learning for sequential recommendation with mutual information maximization. InCIKM

    work page 2020

  57. [57]

    Peilin Zhou, You-Liang Huang, Yueqi Xie, Jingqi Gao, Shoujin Wang, Jae Boum Kim, and Sunghun Kim. 2024. Is contrastive learning necessary? a study of data augmentation vs contrastive learning in sequential recommendation. InWWW

    work page 2024