arxiv: 2605.11707 · v1 · submitted 2026-05-12 · 💻 cs.IR

Recognition: 2 theorem links

· Lean Theorem

Quality-Aware Collaborative Multi-Positive Contrastive Learning for Sequential Recommendation

Wei Wang

Authors on Pith no claims yet

Pith reviewed 2026-05-13 05:27 UTC · model grok-4.3

classification 💻 cs.IR

keywords sequential recommendationcontrastive learningcollaborative augmentationquality-aware weightingdata augmentationmulti-positiveuser sequence modeling

0 comments

The pith

Learnable collaborative augmentations with quality weighting improve contrastive learning for sequential recommendation

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

The paper aims to show that heuristic augmentations in contrastive learning for sequential recommendation often remove key items or break transition patterns, leading to semantic drift and false positives. It proposes generating two augmented views through a learnable module that draws on same-target sequences and similar sequences to increase diversity while holding intent steady. A quality-aware component then scores each view by how confident the augmentation process was and uses those scores to give stronger weight to reliable views during training. If this works, models can extract better positive signals from user histories without as much manual intervention in data changes. Readers would care because sequential recommendation drives many online systems, and stronger contrastive signals could raise prediction accuracy on next actions.

Core claim

We introduce Quality-aware Collaborative Multi-Positive Contrastive Learning. A learnable collaborative sequence augmentation module generates two augmented views under two complementary collaborative contexts, one based on same-target sequences and the other on similar sequences, thereby enhancing view diversity while preserving intent consistency. A quality-aware mechanism, tightly integrated into the model representations, estimates each view's quality from the confidence of its augmentation operations and assigns adaptive weights to ensure that high-confidence views contribute more supervision while low-confidence ones contribute less. Extensive experiments on three real-world datasets (

What carries the argument

Learnable collaborative sequence augmentation module that draws views from same-target and similar sequences, paired with a quality estimation mechanism that derives adaptive weights from augmentation confidence

If this is right

Views gain diversity from dual collaborative contexts while retaining semantic consistency with the original sequence intent
High-confidence views exert stronger influence on the contrastive loss, lowering the effect of low-quality or drifted views
Explicit modeling of quality differences across views reduces the false-positive problem in multi-positive contrastive setups
The full model outperforms prior CL-based sequential recommendation methods across the tested real-world datasets

Where Pith is reading between the lines

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

The dual-context augmentation idea could be adapted to session-based or graph-based recommendation settings where sequence patterns are also central
Extending the quality estimation to incorporate direct user feedback signals might further refine which views receive higher weight
The approach may reduce the need for dataset-specific manual tuning of augmentation strategies in production recommendation pipelines

Load-bearing premise

The confidence scores from the augmentation operations accurately reflect the true semantic usefulness of the generated views and do not introduce new biases into the learning process

What would settle it

If the performance gains disappear when the quality-aware weighting is removed or replaced by uniform weights on the same three real-world datasets, while keeping the collaborative augmentation module intact

Figures

Figures reproduced from arXiv: 2605.11707 by Wei Wang.

**Figure 2.** Figure 2: Overview of QCMP-CL. Stage I pre-trains the co-augmentation module via self-supervised reconstruction from [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Figure 3: Statistical Analysis on Beauty, Yelp and Sports [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

read the original abstract

The effectiveness of contrastive learning in sequential recommendation hinges on the construction of contrastive views, which ideally should be both semantically consistent and diverse. However, most existing CL-based methods rely on heuristic augmentations that are prone to removing crucial items or disrupting transition patterns, leading to semantic drift. While a few studies have explored learnable augmentations to improve view quality, they often suffer from limited diversity and still necessitate heuristic aids. Furthermore, the quality differences across views are rarely modeled explicitly and adaptively, aggravating the false-positive issue. To address these issues, we propose Quality-aware Collaborative Multi-Positive Contrastive Learning for sequential recommendation. First, we introduce a learnable collaborative sequence augmentation module that generates two augmented views under two complementary collaborative contexts, one based on same-target sequences and the other on similar sequences, thereby enhancing view diversity while preserving intent consistency.Second, we design a quality-aware mechanism, tightly integrated into the model representations, which estimates each view' s quality from the confidence of its augmentation operations and assigns adaptive weights to ensure that high-confidence views contribute more supervision while low-confidence ones contribute less.Extensive experiments on three real-world datasets demonstrate that QCMP-CL outperforms state-of-the-art CL-based sequential recommendation baselines.

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.

Desk Editor's Note private letter to a colleague

The paper pairs collaborative contexts for multi-positive views with a confidence-based quality weight, which is a reasonable step past pure heuristics, but the link between internal confidence and actual view usefulness stays unverified.

read the letter

The main new piece is the learnable augmentation that pulls two views from same-target sequences and similar sequences, then feeds an estimated quality score into the contrastive loss so higher-confidence positives get more weight. This directly targets the semantic drift and false-positive problems that come from dropping items or breaking transitions in standard augmentations. The design is straightforward and the motivation lines up with known weaknesses in CL-based sequential recommendation work. If the full experiments show the weighting actually improves consistency without adding bias, it could be a practical tweak for people building these models. The experiments are said to beat several baselines on three datasets, which is the expected bar in this area. The soft spot is the quality mechanism itself. It derives weights only from the model's own in the augmentation steps, with no separate check that high-confidence views really preserve user intent better than low-confidence ones. Without that correlation shown through ablations or qualitative checks, the adaptive weighting risks amplifying easy positives or creating new selection effects rather than fixing the original issue. The abstract also skips any mention of statistical significance or error bars, so the reported gains are difficult to assess for robustness. This is for researchers already working on contrastive methods in recommender systems who want concrete ideas for handling view quality. A reader in that niche would get value from the architecture even if the results need more backing. I would send it to peer review because the proposal is clear enough that referees can test whether the quality estimation holds up.

Referee Report

3 major / 1 minor

Summary. The paper proposes Quality-aware Collaborative Multi-Positive Contrastive Learning (QCMP-CL) for sequential recommendation. It introduces a learnable collaborative sequence augmentation module that generates two augmented views from complementary contexts (same-target sequences and similar sequences) to improve diversity while preserving intent. It also presents a quality-aware mechanism that estimates each view's quality from the model's confidence in its augmentation operations and applies adaptive weights in the contrastive loss. The central claim is that this framework outperforms state-of-the-art CL-based sequential recommendation baselines on three real-world datasets.

Significance. If the empirical claims hold after proper validation, the work could meaningfully advance contrastive learning for sequential recommendation by replacing heuristic augmentations with learnable collaborative ones and by explicitly modeling view quality to reduce false-positive positives. The integration of quality estimation directly into representation learning is a potentially useful direction for mitigating semantic drift.

major comments (3)

[Abstract] Abstract: The claim that 'QCMP-CL outperforms state-of-the-art CL-based sequential recommendation baselines' is presented without any quantitative results, ablation tables, statistical significance tests, or error bars. This absence makes it impossible to assess whether the quality-aware weighting actually improves performance or merely amplifies easy positives.
[Method] Method description (quality-aware mechanism): The quality score for each view is derived solely from the model's internal confidence in the same-target vs. similar-sequence augmentation operations. No independent check (e.g., correlation with transition-pattern preservation or human judgment of semantic consistency) is reported, leaving open the possibility that the adaptive weights introduce new selection bias rather than addressing semantic drift.
[Experiments] Experiments: No implementation details, hyper-parameter settings, or ablation studies isolating the contribution of the learnable augmentation module versus the quality-aware weighting are provided. Without these, the load-bearing claim that the proposed components jointly solve the false-positive problem cannot be evaluated.

minor comments (1)

[Abstract] The abstract refers to 'three real-world datasets' without naming them or describing their characteristics (e.g., sparsity, sequence length distribution), which would help readers assess the scope of the claimed improvements.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the thoughtful and constructive comments. We agree that the abstract can be strengthened with quantitative highlights, that the quality-aware mechanism would benefit from additional discussion of its design choices, and that experimental details should be more explicitly referenced. We address each major comment below and will incorporate the suggested changes in the revised manuscript.

read point-by-point responses

Referee: [Abstract] Abstract: The claim that 'QCMP-CL outperforms state-of-the-art CL-based sequential recommendation baselines' is presented without any quantitative results, ablation tables, statistical significance tests, or error bars. This absence makes it impossible to assess whether the quality-aware weighting actually improves performance or merely amplifies easy positives.

Authors: We acknowledge that the abstract would be more informative with concrete performance numbers. In the revision we will add specific relative improvements (e.g., average gains of X% on HR@10 and NDCG@10 across the three datasets) and explicitly state that all reported results include statistical significance testing and error bars from five random seeds. The full tables with these metrics, ablations, and significance tests already appear in Section 4; we will simply surface the key figures in the abstract itself. revision: yes
Referee: [Method] Method description (quality-aware mechanism): The quality score for each view is derived solely from the model's internal confidence in the same-target vs. similar-sequence augmentation operations. No independent check (e.g., correlation with transition-pattern preservation or human judgment of semantic consistency) is reported, leaving open the possibility that the adaptive weights introduce new selection bias rather than addressing semantic drift.

Authors: The quality score is deliberately computed from the model's own augmentation confidence so that weighting remains fully differentiable and end-to-end trainable. This design choice avoids the need for external labels while allowing the model to down-weight unreliable views during optimization. We agree that an explicit correlation study with human semantic judgments or transition preservation metrics would provide additional reassurance; however, such an analysis would require new annotation effort outside the current scope. In the revision we will add a dedicated paragraph in Section 3.3 discussing this design rationale, potential selection bias, and why internal confidence serves as a practical proxy, supported by the observed performance gains. revision: partial
Referee: [Experiments] Experiments: No implementation details, hyper-parameter settings, or ablation studies isolating the contribution of the learnable augmentation module versus the quality-aware weighting are provided. Without these, the load-bearing claim that the proposed components jointly solve the false-positive problem cannot be evaluated.

Authors: We apologize that the experimental protocol was not sufficiently prominent. Section 4.1 and Appendix A already list all hyper-parameters (learning rate, embedding size, temperature, augmentation probabilities, etc.) and the exact data splits. Section 4.3 contains ablation studies that remove the learnable collaborative augmentation and the quality-aware weighting in turn, showing their individual and joint contributions. In the revision we will (1) add a short table in the main text summarizing the key hyper-parameters, (2) expand the ablation subsection to include a direct comparison of the two modules' marginal gains, and (3) cross-reference these sections from the abstract and introduction. revision: yes

Circularity Check

0 steps flagged

New learnable augmentation and quality estimation are independent of prior fitted parameters

full rationale

The paper's core derivation introduces a learnable collaborative sequence augmentation module (generating views from same-target and similar-sequence contexts) and an integrated quality-aware weighting mechanism (estimating per-view quality directly from augmentation confidence scores). These are not defined in terms of previously fitted parameters from the same paper, nor do any equations reduce predictions to inputs by construction. No self-citation chains, uniqueness theorems, or ansatz smuggling appear as load-bearing elements in the abstract or described method. The central claim of outperformance rests on end-to-end experiments rather than tautological re-derivation, satisfying the default expectation of no significant circularity while warranting a low score for the unverified assumption that internal confidence correlates with semantic usefulness.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that collaborative contexts produce high-quality views and that confidence-based weighting improves supervision. The method introduces learnable parameters for augmentation and quality estimation that are fitted during training.

free parameters (2)

learnable augmentation parameters
Parameters in the collaborative sequence augmentation module are optimized during model training to generate the two views.
quality estimation parameters
Parameters that map augmentation operations to confidence scores and adaptive weights are learned as part of the model.

axioms (1)

domain assumption Views generated from same-target and similar sequences preserve user intent while adding diversity
Invoked in the description of the augmentation module as the basis for semantic consistency.

pith-pipeline@v0.9.0 · 5504 in / 1312 out tokens · 104919 ms · 2026-05-13T05:27:38.485540+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

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

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

quality-aware mechanism... estimates each view's quality from the confidence of its augmentation operations
IndisputableMonolith/Foundation/ArithmeticFromLogic.lean embed_injective unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Jaccard (S_u, S_v) similarity for collaborative sampling

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.

Reference graph

Works this paper leans on

43 extracted references · 43 canonical work pages

[1]

Shuqing Bian, Wayne Xin Zhao, Jinpeng Wang, and Ji-Rong Wen. 2022. A Relevant and Diverse Retrieval-enhanced Data Augmentation Framework for Sequential Recommendation. InProceedings of the 31st ACM International Confer- ence on Information & Knowledge Management(Atlanta, GA, USA)(CIKM ’22). 2923–2932

work page 2022
[2]

Shuqing Bian, Wayne Xin Zhao, Kun Zhou, Jing Cai, Yancheng He, Cunxiang Yin, and Ji-Rong Wen. 2021. Contrastive Curriculum Learning for Sequential User Behavior Modeling via Data Augmentation. InProceedings of the 30th ACM International Conference on Information & Knowledge Management(Virtual Event, Queensland, Australia)(CIKM ’21). 3737–3746

work page 2021
[3]

Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. Simclr: A simple framework for contrastive learning of visual representations. InInternational Conference on Learning Representations, Vol. 2. PMLR New York, NY, USA

work page 2020
[4]

Yongjun Chen, Zhiwei Liu, Jia Li, Julian McAuley, and Caiming Xiong. 2022. Intent Contrastive Learning for Sequential Recommendation. InProceedings of the ACM Web Conference 2022(Virtual Event, Lyon, France)(WWW ’22). 2172–2182

work page 2022
[5]

Ziqiang Cui, Haolun Wu, Bowei He, Ji Cheng, and Chen Ma. 2024. Context Matters: Enhancing Sequential Recommendation with Context-aware Diffusion- based Contrastive Learning. InProceedings of the 33rd ACM International Confer- ence on Information and Knowledge Management. 404–414

work page 2024
[6]

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. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval(Padua, Italy)(SIGIR ’25). 1466–1475

work page 2025
[7]

Hanwen Du, Huanhuan Yuan, Pengpeng Zhao, Fuzhen Zhuang, Guanfeng Liu, Lei Zhao, Yanchi Liu, and Victor S. Sheng. 2023. Ensemble Modeling with Contrastive Knowledge Distillation for Sequential Recommendation. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval(Taipei, Taiwan)(SIGIR ’23). 58–67

work page 2023
[8]

Ruiming Guo, Mouxing Yang, Yijie Lin, Xi Peng, and Peng Hu. 2024. Robust contrastive multi-view clustering against dual noisy correspondence(NIPS ’24). Article 3857, 21 pages

work page 2024
[9]

R Devon Hjelm, Alex Fedorov, Samuel Lavoie-Marchildon, Karan Grewal, Phil Bachman, Adam Trischler, and Yoshua Bengio. 2019. Learning deep representa- tions by mutual information estimation and maximization. InInternational Confer- ence on Learning Representations. https://openreview.net/forum?id=Bklr3j0cKX

work page 2019
[10]

Wang-Cheng Kang and Julian McAuley. 2018. Self-attentive sequential recom- mendation. In2018 IEEE international conference on data mining (ICDM). IEEE, 197–206

work page 2018
[11]

Kingma and Jimmy Ba

Diederik P. Kingma and Jimmy Ba. 2015. Adam: A Method for Stochastic Opti- mization. In3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, Yoshua Bengio and Yann LeCun (Eds.)

work page 2015
[12]

Jake Lever, Martin Krzywinski, and Naomi Altman. 2016. Points of significance: model selection and overfitting.Nature methods13, 9 (2016), 703–705

work page 2016
[13]

Fei Li, Qingyun Gao, Yizhou Dang, Enneng Yang, Guibing Guo, Jianzhe Zhao, and Xingwei Wang. 2025. Denoising Multi-Interest-Aware Logical Reasoning for Long-Sequence Recommendation. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval(Padua, Italy)(SIGIR ’25). 1487–1496

work page 2025
[14]

Shikun Li, Xiaobo Xia, Shiming Ge, and Tongliang Liu. 2022. Selective-Supervised Contrastive Learning with Noisy Labels. In2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 316–325

work page 2022
[15]

Yuxin Liao, Yonghui Yang, Min Hou, Le Wu, Hefei Xu, and Hao Liu. 2025. Mitigat- ing Distribution Shifts in Sequential Recommendation: An Invariance Perspective. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval(Padua, Italy)(SIGIR ’25). 1603–1613

work page 2025
[16]

Yujie Lin, Chenyang Wang, Zhumin Chen, Zhaochun Ren, Xin Xin, Qiang Yan, Maarten de Rijke, Xiuzhen Cheng, and Pengjie Ren. 2023. A Self-Correcting Sequential Recommender. InProceedings of the ACM Web Conference 2023(Austin, TX, USA)(WWW ’23). 1283–1293

work page 2023
[17]

Zhiwei Liu, Yongjun Chen, Jia Li, Philip S Yu, Julian McAuley, and Caiming Xiong. 2021. Contrastive self-supervised sequential recommendation with robust augmentation.arXiv preprint arXiv:2108.06479(2021)

work page arXiv 2021
[18]

Julian McAuley, Christopher Targett, Qinfeng Shi, and Anton van den Hengel

work page
[19]

InProceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval(Santiago, Chile)(SIGIR ’15)

Image-based recommendations on styles and substitutes. InProceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval(Santiago, Chile)(SIGIR ’15). 43–52

work page
[20]

Xiuyuan Qin, Huanhuan Yuan, Pengpeng Zhao, Junhua Fang, Fuzhen Zhuang, Guanfeng Liu, Yanchi Liu, and Victor Sheng. 2023. Meta-optimized Contrastive Learning for Sequential Recommendation. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval (Taipei, Taiwan)(SIGIR ’23). 89–98

work page 2023
[21]

Xiuyuan Qin, Huanhuan Yuan, Pengpeng Zhao, Guanfeng Liu, Fuzhen Zhuang, and Victor S. Sheng. 2024. Intent Contrastive Learning with Cross Subsequences for Sequential Recommendation. InProceedings of the 17th ACM International Con- ference on Web Search and Data Mining(Merida, Mexico)(WSDM ’24). 548–556

work page 2024
[22]

Ruihong Qiu, Zi Huang, Hongzhi Yin, and Zijian Wang. 2022. Contrastive Learn- ing for Representation Degeneration Problem in Sequential Recommendation. InProceedings of the Fifteenth ACM International Conference on Web Search and Data Mining(Virtual Event, AZ, USA)(WSDM ’22). 813–823

work page 2022
[23]

Pengjie Ren, Zhumin Chen, Jing Li, Zhaochun Ren, Jun Ma, and Maarten de Rijke. 2019. RepeatNet: A repeat aware neural recommendation machine for session-based recommendation. InProceedings of the Thirty-Third AAAI Confer- ence on Artificial Intelligence and Thirty-First Innovative Applications of Artificial Intelligence Conference and Ninth AAAI Symposium...

work page 2019
[24]

Zhaochun Ren, Na Huang, Yidan Wang, Pengjie Ren, Jun Ma, Jiahuan Lei, Xinlei Shi, Hengliang Luo, Joemon Jose, and Xin Xin. 2023. Contrastive State Augmen- tations for Reinforcement Learning-Based Recommender Systems. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval(Taipei, Taiwan)(SIGIR ’23)...

work page 2023
[25]

David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams. 1986. Learning representations by back-propagating errors.nature323, 6088 (1986), 533–536

work page 1986
[26]

Jie Shuai, Kun Zhang, Le Wu, Peijie Sun, Richang Hong, Meng Wang, and Yong Li. 2022. A Review-aware Graph Contrastive Learning Framework for Recom- mendation. InProceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval(Madrid, Spain)(SIGIR ’22). 1283–1293

work page 2022
[27]

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

work page
[28]

InProceedings of the 28th ACM International Conference on Information and Knowledge Management(Beijing, China)(CIKM ’19)

BERT4Rec: Sequential recommendation with bidirectional encoder rep- resentations from transformer. InProceedings of the 28th ACM International Conference on Information and Knowledge Management(Beijing, China)(CIKM ’19). 1441–1450

work page
[29]

Haoyun Wang, Yongquan Fan, Yajun Du, Xianyong Li, and Xiaomin Wang. 2025. Improving contrastive learning with explanation method for sequential recom- mendation.Expert Systems with Applications291 (2025), 128534

work page 2025
[30]

Lei Wang, Ee-Peng Lim, Zhiwei Liu, and Tianxiang Zhao. 2022. Explana- tion Guided Contrastive Learning for Sequential Recommendation(CIKM ’22). 2017–2027

work page 2022
[31]

Wei Wang, Yujie Lin, Pengjie Ren, Zhumin Chen, Tsunenori Mine, Jianli Zhao, Qiang Zhao, Moyan Zhang, Xianye Ben, and Yujun Li. 2025. Privacy-Preserving Sequential Recommendation with Collaborative Confusion.ACM Trans. Inf. Syst. 43, 2, Article 50 (Jan. 2025), 25 pages

work page 2025
[32]

Wei Wang, Yujie Lin, Moyan Zhang, Hongyu Lu, Jianli Zhao, Jie Sun, Xianye Ben, Pengjie Ren, and Yujun Li. 2025. Triplet Contrastive Learning with Learn- able Sequence Augmentation for Sequential Recommendation. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval(Padua, Italy)(SIGIR ’25). 1519–1529

work page 2025
[33]

Ziyang Wang, Huoyu Liu, Wei Wei, Yue Hu, Xian-Ling Mao, Shaojian He, Rui Fang, and Dangyang Chen. 2022. Multi-level Contrastive Learning Framework for Sequential Recommendation. InProceedings of the 31st ACM International Conference on Information & Knowledge Management(Atlanta, GA, USA)(CIKM ’22). 2098–2107

work page 2022
[34]

Zhikai Wang, Yanyan Shen, Zexi Zhang, Li He, Yichun Li, Hao Gu, and Yinghua Zhang. 2024. Relative Contrastive Learning for Sequential Recommendation with Similarity-based Positive Sample Selection. InProceedings of the 33rd ACM International Conference on Information and Knowledge Management(Boise, ID, USA)(CIKM ’24). 2493–2502

work page 2024
[35]

Xu Xie, Fei Sun, Zhaoyang Liu, Shiwen Wu, Jinyang Gao, Jiandong Zhang, Bolin Ding, and Bin Cui. 2022. Contrastive learning for sequential recommendation. In 2022 IEEE 38th international conference on data engineering (ICDE). IEEE, 1259– 1273

work page 2022
[36]

Zuxiang Xie and Junyi Li. 2024. Simple Debiased Contrastive Learning for Sequential Recommendation.Knowledge-Based Systems300 (2024), 112257

work page 2024
[37]

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(Washington DC, USA)(SIGIR ’24). 1589–1598

work page 2024
[38]

Mengduo Yang, Yi Yuan, Jie Zhou, Meng Xi, Xiaohua Pan, Ying Li, Yangyang Wu, Jinshan Zhang, and Jianwei Yin. 2024. Adaptive Fusion of Multi-View for Quality-Aware Collaborative Multi-Positive Contrastive Learning for Sequential Recommendation Conference’17, July 2017, Washington, DC, USA Graph Contrastive Recommendation. InProceedings of the 18th ACM Conf...

work page 2024
[39]

Yuhao Yang, Chao Huang, Lianghao Xia, Chunzhen Huang, Da Luo, and Kangyi Lin. 2023. Debiased Contrastive Learning for Sequential Recommendation. In Proceedings of the ACM Web Conference 2023(Austin, TX, USA)(WWW ’23). 1063–1073

work page 2023
[40]

Dan Zhang, Yangliao Geng, Wenwen Gong, Zhongang Qi, Zhiyu Chen, Xing Tang, Ying Shan, Yuxiao Dong, and Jie Tang. 2024. RecDCL: Dual Contrastive Learning for Recommendation. InProceedings of the ACM Web Conference 2024 (Singapore, Singapore)(WWW ’24). 3655–3666

work page 2024
[41]

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. In Proceedings of the ACM Web Conference 2024(Singapore, Singapore)(WWW ’24). 3854–3863

work page 2024
[42]

Yuanhang Zhou, Kun Zhou, Wayne Xin Zhao, Cheng Wang, Peng Jiang, and He Hu. 2022. C2-CRS: Coarse-to-Fine Contrastive Learning for Conversational Recommender System. InProceedings of the Fifteenth ACM International Con- ference on Web Search and Data Mining(Virtual Event, AZ, USA)(WSDM ’22). 1488–1496

work page 2022
[43]

Guanghui Zhu, Wang Lu, Chunfeng Yuan, and Yihua Huang. 2023. AdaMCL: Adaptive Fusion Multi-View Contrastive Learning for Collaborative Filtering. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval(Taipei, Taiwan)(SIGIR ’23). 1076–1085

work page 2023