Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Inter-interaction time gaps, tokenized and interleaved with semantic item codes, fix temporal blindness in generative recommenders and lift next-item accuracy.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-11 23:01 UTC pith:W45TWHTP

load-bearing objection Clean, controlled extension of ReSID that adds gap tokens and a light temporal aux loss; modest consistent gains, especially on long gaps, but missing a dummy/random-gap control for capacity confounds. the 3 major comments →

arxiv 2607.03918 v1 pith:W45TWHTP submitted 2026-07-04 cs.IR

Beyond Item Order: Temporal Gap Tokenization for Generative Recommendation with Semantic IDs

Chengkai Huang , Tianqi Gao , Hongtao Huang , Quan Z. Sheng , Lina Yao This is my paper
classification cs.IR
keywords Recommender SystemsGenerative RecommendationSemantic IDsTemporal ModelingSequential RecommendationGap TokensTime-Aware Representation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Semantic-ID generative recommenders turn next-item prediction into short code generation, but they usually feed the model only a static sequence of item codes and ignore how much time passed between clicks. The paper argues that those elapsed gaps carry real signal about whether a user is still in the same intent or has drifted. ChronoSID keeps the existing three-stage pipeline (representation, quantization, generation) and adds two light temporal pieces: an auxiliary time-gap prediction loss while learning item embeddings, and fixed log-scale gap tokens interleaved with the semantic IDs that the generator sees. On eight Amazon review domains this consistently beats the strongest same-pipeline baseline and other generative methods, with the largest lifts when the wait until the next interaction is long.

Core claim

Static semantic-ID sequences are temporally incomplete; injecting discretized historical inter-interaction gaps as encoder tokens, plus a light time-gap auxiliary loss on item representations, improves generative next-item recommendation while preserving compact SID generation, and the gains are clearest under long-gap conditions where preference drift is more likely.

What carries the argument

ChronoSID’s gap-token interleaving: each historical interaction is preceded by a discrete log-scale gap token (<1 h, 1 h–1 d, 1 d–1 w, 1 w–1 mo, ≥1 mo, plus a start token) so the T5 encoder input becomes gap-SID-SID-SID repeated, giving the generator explicit temporal context without changing the decoder target or the quantization stage.

Load-bearing premise

A single hand-chosen set of five fixed log-scale time bins works well enough across product domains that the measured gains can be credited to temporal awareness rather than to a lucky binning.

What would settle it

Re-run the same leave-one-out Amazon splits with the identical ReSID quantization and T5 backbone, but replace the fixed log bins by random or equal-frequency bins of the same cardinality; if ChronoSID’s lift over ReSID disappears or reverses, the central claim that the temporal signal itself is responsible fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • SID-based generative recommenders that ignore elapsed time systematically under-perform when users return after long absences.
  • Gap-token injection alone accounts for most of the lift; the representation-level time-gap loss is a smaller complementary regularizer.
  • Encoder length grows only from 3L to 4L tokens while the decoder target stays a three-code SID, so the accuracy gain comes at moderate extra cost.
  • The same interleaving pattern can be dropped onto other SID pipelines that keep frozen item codes and a sequence-to-sequence generator.

Where Pith is reading between the lines

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

  • If fixed bins already help, learned or adaptive temporal vocabularies could capture domain-specific rhythms (seasonality, shopping cycles) without enlarging the decoder.
  • The same gap tokens could condition not only next-item codes but also next-time prediction, turning the generator into a joint item-and-timing model.
  • Long-gap robustness may matter most in sparse catalogs or cold-start return visits, suggesting a natural stress test beyond Amazon leave-one-out.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes ChronoSID, a lightweight temporal augmentation of the standard three-stage semantic-ID (SID) generative recommendation pipeline. It keeps Globally Aligned Orthogonal Quantization (GAOQ) identical to ReSID and injects time in two places: (i) TA-FAMAE, which adds a log-gap regression auxiliary loss to field-aware masked auto-encoding of item features (§4.1, Eqs. 12–15), and (ii) fixed log-scale gap-token interleaving on the T5 encoder input, expanding each history from 3L to 4L tokens while leaving the decoder target as the next item’s three-level SID (§3 Eqs. 3–7, §4.3.2 Eq. 30). On eight Amazon-2023 leave-one-out splits the method reports consistent Recall@K / NDCG@K gains over ReSID and other SID generative baselines (Table 2), with ablations attributing most of the lift to gap tokens (Table 3), clearer relative gains on long target-side gaps (Figure 3), sensitivity on λ and G (Table 4), efficiency overhead (Table 5), popularity splits (Figure 4), and output-level SID accuracy with bootstrap CIs (Table 6).

Significance. Temporal intervals are a classical signal in sequential recommendation but remain under-used once both inputs and targets are discrete SID code sequences. ChronoSID’s controlled design—identical GAOQ, same T5 backbone and beam protocol as ReSID—makes the empirical comparison unusually clean for this literature. The experimental package is thorough (eight domains, component ablations, long-gap diagnostics, hyperparameter sweeps, popularity and output-level analyses, efficiency numbers). If the gains are genuinely driven by temporal semantics rather than encoder length or token-type diversity, the work supplies a simple, practical recipe that preserves the compact SID generation paradigm. That would be a useful, if incremental, contribution to generative recommendation systems.

major comments (3)
  1. [§4.3.2, Table 3, Figure 3] Central attribution claim (§1, §3 Eq. 6, §4.3.2 Eq. 30, abstract): gains over ReSID are ascribed to “temporal awareness” via Disc(Δt) gap tokens. Ablations (Table 3) and long-gap groups (Figure 3) support usefulness of the injected tokens, but the design also lengthens every encoder sequence from 3L to 4L and introduces a new discrete token type with its own embeddings. No control replaces real gap labels by (i) a constant dummy gap, (ii) randomly permuted gap labels, or (iii) a non-temporal filler that preserves length and vocabulary size. Without such a control, improvements—and the long-gap pattern—could arise from extra encoder capacity or a generic separator/recency-bucket signal rather than the semantic content of the chosen {1h,1d,1w,1mo} thresholds. This is load-bearing for the paper’s main scientific claim and should be added (at least on MI/VG).
  2. [§3, §4.3.2, Table 4] Discretization universality (§3, §4.3.2, Table 4): the method freezes G=5 log-scale bins with hand-chosen thresholds for all eight domains. Sensitivity only varies bin count G inside the same log-scale family; it does not test alternative partitions (linear, quantile, domain-adaptive, or learned). Given that absolute gains vary across domains (Table 2) and that the weakest modeling assumption is precisely this fixed binning, a short domain-adaptive or quantile-bin comparison (or an explicit limitation statement with a negative result) is needed before the gains can be confidently attributed to a general temporal mechanism rather than a binning that fits Amazon-2023 leave-one-out splits.
  3. [Table 2, §5.1] Statistical reporting of main results (Table 2, §5.1): results are “averaged over five runs,” yet no standard deviations, paired significance tests, or confidence intervals appear in the main table (bootstrap CIs are given only for the MI output-level analysis in Table 6). For a multi-domain claim of “consistent” improvement over the direct baseline ReSID, at least pairwise significance (or mean±std) on the primary metrics should be reported so readers can judge whether small absolute lifts (e.g., BPC, TG) are reliable.
minor comments (6)
  1. [Figure 1] Figure 1 caption and panel (b) use β / temporal decay coefficient without stating the exact functional form fitted (e.g., same-category rate ~ exp(−β·gap) or similar). A one-line equation would make the diagnostic reproducible.
  2. [§3 Eq. (4), §4.3.2 Eq. (28)] Notation: Disc(·) is introduced in Eq. (4) and reused as Disc(t_l − t_{l−1}) in Eq. (28); the latter has an extra closing parenthesis. Also g^{(Disc(...))} vs g_l notation is slightly inconsistent between §3 and §4.3.2.
  3. [Table 2 vs Table 3] Table 3 ChronoSID numbers on MI/VG are slightly higher than the corresponding ChronoSID entries in Table 2 (e.g., MI R@5 0.0419 vs 0.0417). Clarify whether these are different seeds, different checkpoints, or a reporting inconsistency.
  4. [§2.1–2.2] Related work (§2.1) cites time-interval-aware self-attention (Li et al., 2020) and other temporal sequential models; a short explicit contrast—why gap tokens on SID sequences differ from continuous time embeddings in item-ID Transformers—would sharpen the positioning.
  5. [Table 5, §5.5] Efficiency Table 5: training-time ratio on VG is 1.29× while MI is 1.07×; a brief note on sequence-length distribution or batching would help readers interpret the overhead.
  6. [Abstract, Figure 1, §4] Typos / polish: “Time-A ware” / “Field-A ware” spacing artifacts in the abstract and §4; “sequence-to sequence” missing hyphen in the abstract; “T emporal” in Figure 1(b) axis label.

Circularity Check

0 steps flagged

No circular derivation: ChronoSID is an empirical temporal augmentation of SID generative rec; metrics are independent leave-one-out rankings, not quantities forced by definition or fit.

full rationale

The paper’s load-bearing chain is design-plus-evaluation, not a first-principles derivation that collapses into its inputs. Inter-interaction gaps Δt_l are computed from timestamps (Eq. 3), discretized into fixed log-scale bins (Eq. 4, §4.3.2), and interleaved as encoder-side tokens (Eq. 6, 30); the decoder still predicts the next item’s semantic ID under teacher-forced CE (Eq. 32), and evaluation is standard leave-one-out Recall/NDCG against held-out next items. Neither the reported ranking metrics nor the long-gap diagnostic groups are defined from the gap-token construction or from the TA-FAMAE auxiliary regression (Eq. 13–15). Hyperparameters λ and G are selected on validation (Table 4); that is ordinary model selection, not a fitted free parameter renamed as a prediction of a closely related target. ReSID is kept as a controlled backbone with unchanged GAOQ so that gains isolate Stages 1 and 3; ReSID is external work, not a self-citation uniqueness theorem that forces ChronoSID’s result. Author self-citations in related work (prior temporal sequential rec) are background and not load-bearing for the central claim. Figure 1’s fitted decay coefficients β are motivational only and do not enter the recommender or the reported metrics. Concerns about missing dummy/permuted-gap controls (capacity vs temporal semantics) are identification/correctness issues, not circularity. No equation reduces R@K/N@K to Disc(Δt) or λ by construction. Score 0 is therefore the honest finding.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 2 invented entities

The central claim is empirical and rests on standard sequential-recommendation modeling choices plus a small set of hand-chosen temporal hyper-parameters. No new physical or mathematical entities are postulated; the free parameters are ordinary ML knobs whose values are selected by validation/sensitivity rather than derived.

free parameters (4)
  • temporal auxiliary weight λ = 0.1
    Controls strength of time-gap regression in TA-FAMAE; default 0.1 chosen by sensitivity (Table 4); claim depends on a positive but not too large λ.
  • number of real gap bins G and log thresholds = G=5; thresholds {1h,1d,1w,1mo}
    G=5 with fixed cut-points {1 h, 1 d, 1 w, 1 mo} defines the discrete temporal vocabulary; selected by sensitivity and used for all main results.
  • GAOQ codebook sizes (b1=32, b2, b3=n_max) = b1=32 (others data-dependent)
    Inherited from ReSID and kept frozen; still free design choices that shape the SID space the generator predicts.
  • per-level beam sizes [50,50,50] = [50,50,50]
    Inference hyper-parameter that affects which SID candidates are scored; fixed for all generative comparisons.
axioms (4)
  • domain assumption Leave-one-out chronological split on Amazon-2023 subsets with users/items filtered to ≥5 interactions is a valid proxy for sequential recommendation quality.
    Standard in the cited generative-rec literature (§5); load-bearing for all reported metrics.
  • domain assumption Inter-interaction time gaps carry behavioral signal about interest continuity and drift that is not already fully captured by item-order SIDs.
    Motivated by Figure 1 same-category decay analysis; if false, gap tokens would be pure noise.
  • ad hoc to paper Keeping Stage-2 GAOQ identical to ReSID isolates the contribution of the two temporal components.
    Explicit design choice (§4, §5) that makes the ReSID comparison controlled but assumes quantization quality is independent of temporal regularization.
  • domain assumption Teacher-forced cross-entropy on three-level SID tokens plus beam-search lookup is a sufficient training/inference protocol for ranking evaluation.
    Inherited from TIGER/ReSID-style generative recommendation (§4.3.3).
invented entities (2)
  • ChronoSID framework (TA-FAMAE + gap-token interleaving) no independent evidence
    purpose: Package the two temporal augmentations as a drop-in extension of the three-stage SID pipeline.
    Method-level construct; no independent physical existence claimed beyond the empirical gains on the reported benchmarks.
  • Fixed log-scale gap tokens as encoder-side symbols no independent evidence
    purpose: Inject elapsed time into the generative input without enlarging the decoder target vocabulary.
    Discretization scheme invented for this paper; falsifiable only via the same ranking experiments, not via an external measurement.

pith-pipeline@v1.1.0-grok45 · 25369 in / 3390 out tokens · 32529 ms · 2026-07-11T23:01:20.275173+00:00 · methodology

0 comments
read the original abstract

Semantic-ID-based generative recommendation has recently emerged as a scalable paradigm for sequential recommendation, where each item is represented by a compact sequence of discrete codes and next-item prediction is formulated as code generation. Existing methods, however, typically construct user histories as sequences of static item identifiers, leaving the elapsed time between consecutive interactions outside the generative input. This temporal blindness is problematic because inter-interaction gaps provide useful cues about interest continuity and preference drift. In this paper, we propose ChronoSID, a lightweight temporal augmentation framework for semantic-ID-based generative recommendation. ChronoSID injects temporal signals into the standard three-stage semantic-ID pipeline from two complementary perspectives. First, we introduce Time-Aware Field-Aware Masked Auto-Encoding (TA-FAMAE), which regularizes item representation learning with an auxiliary time-gap prediction objective. Second, we discretize historical interaction intervals into fixed log-scale gap tokens and interleave them with semantic ID tuples as the encoder input of the sequence-to sequence generator. This design preserves the compact SID generation paradigm while enabling the model to capture time-aware transition patterns. Experiments on Amazon review benchmarks show that ChronoSID consistently improves over ReSID and other competitive generative recommendation baselines. Ablation studies further verify the contribution of both temporal components, and diagnostic analyses show clearer gains under long-gap scenarios where user interests are more likely to drift.

Figures

Figures reproduced from arXiv: 2607.03918 by Chengkai Huang, Hongtao Huang, Lina Yao, Quan Z. Sheng, Tianqi Gao.

Figure 1
Figure 1. Figure 1: Inter-interaction time gaps provide useful cues [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of ChronoSID. Stage 1 learns item representations with Time-Aware Field-Aware Masked Auto-Encoding [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Temporal robustness analysis on MI and VG. Test cases are grouped by the elapsed time between the last historical [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Relative improvement of ChronoSID over ReSID [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. SPARC: Sequence-aware Progressive Attribute Routing and Compression Framework for Generative Recommendation

    cs.IR 2026-07 conditional novelty 6.0

    A sequence-aware compression framework lets generative recommender models retain context-dependent side information about each past interaction without increasing input length.

Reference graph

Works this paper leans on

47 extracted references · 8 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Tianqi Gao, Chengkai Huang, Zihan Wang, Cao Liu, Ke Zeng, and Lina Yao. 2026. Factorized Latent Reasoning for LLM-based Recommendation.arXiv preprint arXiv:2604.26760(2026)

    Pith/arXiv arXiv 2026

  2. [2]

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

  3. [3]

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

    Pith/arXiv arXiv 2015

  4. [4]

    Yupeng Hou, Jiacheng Li, Zhankui He, An Yan, Xiusi Chen, and Julian McAuley

  5. [5]

    Bridging language and items for retrieval and recommendation.arXiv preprint arXiv:2403.03952(2024)

    Pith/arXiv arXiv 2024

  6. [6]

    Yupeng Hou, Jiacheng Li, Ashley Shin, Jinsung Jeon, Abhishek Santhanam, Wei Shao, Kaveh Hassani, Ning Yao, and Julian McAuley. 2025. Generating long semantic ids in parallel for recommendation. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2. 956–966

    2025

  7. [7]

    Chengkai Huang, Xiaodi Chen, Hongtao Huang, Quan Z Sheng, and Lina Yao

  8. [8]

    Generative Chain of Behavior for User Trajectory Prediction.arXiv preprint arXiv:2601.18213(2026)

    arXiv 2026

  9. [9]

    Chengkai Huang, Shoujin Wang, Xianzhi Wang, and Lina Yao. 2023. Dual con- trastive transformer for hierarchical preference modeling in sequential recom- mendation. InProceedings of the 46th international acm sigir conference on research and development in information retrieval. 99–109

    2023

  10. [10]

    Chengkai Huang, Shoujin Wang, Xianzhi Wang, and Lina Yao. 2023. Modeling temporal positive and negative excitation for sequential recommendation. In Proceedings of the ACM Web Conference 2023. 1252–1263

    2023

  11. [11]

    Chengkai Huang, Junda Wu, Yu Xia, Zixu Yu, Ruhan Wang, Tong Yu, Ruiyi Zhang, Ryan A Rossi, Branislav Kveton, Dongruo Zhou, et al . 2025. Towards agentic recommender systems in the era of multimodal large language models.arXiv preprint arXiv:2503.16734(2025)

    Pith/arXiv arXiv 2025

  12. [12]

    Chengkai Huang, Tong Yu, Kaige Xie, Shuai Zhang, Lina Yao, and Julian McAuley

  13. [13]

    Foundation models for recommender systems: A survey and new perspec- tives.arXiv preprint arXiv:2402.11143(2024)

    Pith/arXiv arXiv 2024

  14. [14]

    Hongtao Huang, Chengkai Huang, Junda Wu, Tong Yu, Julian McAuley, and Lina Yao. 2026. Listwise preference diffusion optimization for user behavior trajectories prediction.Advances in Neural Information Processing Systems38 (2026), 159383–159408

    2026

  15. [15]

    Hongtao Huang, Chengkai Huang, Tong Yu, Xiaojun Chang, Wen Hu, Julian McAuley, and Lina Yao. 2026. Dual Conditional Diffusion for Sequential Recom- mendation. InProceedings of the Nineteenth ACM International Conference on Web Search and Data Mining. 206–216

    2026

  16. [16]

    Jeff Johnson, Matthijs Douze, and Hervé Jégou. 2019. Billion-scale similarity search with GPUs.IEEE Transactions on Big Data7, 3 (2019), 535–547

    2019

  17. [17]

    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. 6420–6425

    2025

  18. [18]

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

    2018

  19. [19]

    Jing Li, Pengjie Ren, Zhumin Chen, Zhaochun Ren, Tao Lian, and Jun Ma. 2017. Neural Attentive Session-based Recommendation. InProceedings of the 2017 ACM on Conference on Information and Knowledge Management (CIKM ’17). Association for Computing Machinery, New York, NY, USA, 1419–1428. doi:10.1145/3132847. 3132926

    doi:10.1145/3132847 2017

  20. [20]

    Jiacheng Li, Yujie Wang, and Julian McAuley. 2020. Time interval aware self- attention for sequential recommendation. InProceedings of the 13th international conference on web search and data mining. 322–330

    2020

  21. [21]

    Yu Liang, Zhongjin Zhang, Yuxuan Zhu, Kerui Zhang, Zhiluohan Guo, Wenhang Zhou, Zonqi Yang, Kangle Wu, Yabo Ni, Anxiang Zeng, et al. 2026. Rethinking Generative Recommender Tokenizer: Recsys-Native Encoding and Semantic Quantization Beyond LLMs.arXiv preprint arXiv:2602.02338(2026)

    arXiv 2026

  22. [22]

    Enze Liu, Bowen Zheng, Cheng Ling, Lantao Hu, Han Li, and Wayne Xin Zhao

  23. [23]

    InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval

    Generative recommender with end-to-end learnable item tokenization. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval. 729–739

  24. [24]

    Chen Ma, Peng Kang, and Xue Liu. 2019. Hierarchical gating networks for sequential recommendation. InProceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining. 825–833

    2019

  25. [25]

    Mikko I Malinen and Pasi Fränti. 2014. Balanced k-means for clustering. InJoint IAPR international workshops on statistical techniques in pattern recognition (SPR) and structural and syntactic pattern recognition (SSPR). Springer, 32–41

    2014

  26. [26]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer.Journal of machine learning research21, 140 (2020), 1–67

    2020

  27. [27]

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

  28. [28]

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

    2023

  29. [29]

    Zihua Si, Zhongxiang Sun, Jiale Chen, Guozhang Chen, Xiaoxue Zang, Kai Zheng, Yang Song, Xiao Zhang, Jun Xu, and Kun Gai. 2024. Generative retrieval with semantic tree-structured identifiers and contrastive learning. InProceedings of the 2024 Annual International ACM SIGIR Conference on Research and Development in Information Retrieval in the Asia Pacific...

    2024

  30. [30]

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

  31. [31]

    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

  32. [32]

    Yong Kiam Tan, Xinxing Xu, and Yong Liu. 2016. Improved recurrent neural networks for session-based recommendations. InProceedings of the 1st workshop on deep learning for recommender systems. 17–22

    2016

  33. [33]

    Jiaxi Tang and Ke Wang. 2018. Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding. InProceedings of the Eleventh ACM International Conference on Web Search and Data Mining(Marina Del Rey, CA, USA)(WSDM ’18). Association for Computing Machinery, New York, NY, USA, 565–573. doi:10.1145/3159652.3159656

    doi:10.1145/3159652.3159656 2018

  34. [34]

    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)

    2017

  35. [35]

    Wenjie Wang, Honghui Bao, Xinyu Lin, Jizhi Zhang, Yongqi Li, Fuli Feng, See- Kiong Ng, and Tat-Seng Chua. 2024. Learnable item tokenization for generative recommendation. InProceedings of the 33rd ACM International Conference on Information and Knowledge Management. 2400–2409

    2024

  36. [36]

    Ye Wang, Jiahao Xun, Minjie Hong, Jieming Zhu, Tao Jin, Wang Lin, Haoyuan Li, Linjun Li, Yan Xia, Zhou Zhao, et al . 2024. Eager: Two-stream generative recommender with behavior-semantic collaboration. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 3245–3254

    2024

  37. [37]

    Shu Wu, Yuyuan Tang, Yanqiao Zhu, Liang Wang, Xing Xie, and Tieniu Tan

  38. [38]

    Session-based recommendation with graph neural networks. InProceedings of the Thirty-Third AAAI Conference on Artificial Intelligence and Thirty-First Innovative Applications of Artificial Intelligence Conference and Ninth AAAI Sym- posium on Educational Advances in Artificial Intelligence(Honolulu, Hawaii, USA) (AAAI’19/IAAI’19/EAAI’19). AAAI Press, Arti...

    doi:10.1609/aaai 2017

  39. [39]

    Longtao Xiao, Haozhao Wang, Cheng Wang, Linfei Ji, Yifan Wang, Jieming Zhu, Zhenhua Dong, Rui Zhang, and Ruixuan Li. 2025. Unger: Generative recommendation with a unified code via semantic and collaborative integration. ACM Transactions on Information Systems44, 2 (2025), 1–31

    2025

  40. [40]

    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:2411.18814 [cs.IR] https://arxiv. org/abs/2411.18814

    Pith/arXiv arXiv 2024

  41. [41]

    Yuhao Yang, Zhi Ji, Zhaopeng Li, Yi Li, Zhonglin Mo, Yue Ding, Kai Chen, Zijian Zhang, Jie Li, Shuanglong Li, et al. 2025. Sparse meets dense: Unified generative recommendations with cascaded sparse-dense representations.arXiv preprint arXiv:2503.02453(2025)

    Pith/arXiv arXiv 2025

  42. [42]

    Xiaoxin Ye, Chengkai Huang, Hongtao Huang, and Lina Yao. 2025. Beyond negative transfer: Disentangled preference-guided diffusion for cross-domain sequential recommendation.arXiv preprint arXiv:2509.00389(2025)

    Pith/arXiv arXiv 2025

  43. [43]

    Xiaoxin Ye, Chengkai Huang, Hongtao Huang, and Lina Yao. 2026. Gaussian Mixture Flow Matching with Domain Alignment for Multi-Domain Sequential Recommendation. InProceedings of the ACM Web Conference 2026. 6159–6170

    2026

  44. [44]

    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

  45. [45]

    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 se- quential recommendation with mutual information maximization. InProceedings of the 29th ACM international conference on information & knowledge management. 1893–1902

    2020

  46. [46]

    Jieming Zhu, Mengqun Jin, Qijiong Liu, Zexuan Qiu, Zhenhua Dong, and Xiu Li

  47. [47]

    InProceedings of the 18th ACM Conference on Recommender Systems

    Cost: Contrastive quantization based semantic tokenization for generative recommendation. InProceedings of the 18th ACM Conference on Recommender Systems. 969–974