pith. sign in

arxiv: 2605.21987 · v1 · pith:ZRKUXADJnew · submitted 2026-05-21 · 💻 cs.IR

Generative Conversational Recommender System

Sixiao Zhang , Mingrui Liu , Cheng Long This is my paper

Pith reviewed 2026-05-22 04:31 UTC · model grok-4.3

classification 💻 cs.IR
keywords conversational recommender systemsgenerative modelsautoregressive generationsemantic IDsstructured generationdialog generationpersonalized recommendations
0
0 comments X

The pith

A single autoregressive model unifies recommendation and dialog generation by using semantic item IDs and a structured intent-target-response process.

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

The paper establishes that conversational recommenders can be built as a fully generative system inside one language model rather than combining separate recommendation and dialog components. Items are turned into discrete semantic IDs that the model predicts as part of its token sequence. The generation is factored into first deciding the response intent, then the specific item to recommend, and finally producing the full natural language response conditioned on those choices. This setup allows joint training and uses constrained decoding to keep recommendations accurate. A sympathetic reader would care because current systems often suffer from poor integration between what the user wants and what gets recommended, resulting in less natural conversations.

Core claim

The central claim is that representing items as discrete semantic IDs and integrating them directly into an autoregressive generation process, combined with a structured generation paradigm that first predicts the response intent and the recommendation target before generating the response, unifies recommendation and dialog generation within a single framework, enabling end-to-end optimization and faithful item generation via constrained decoding.

What carries the argument

The structured generation paradigm that factorizes conversational recommendation into predicting the response intent, then the recommendation target, and finally generating the response conditioned on them, with items represented as discrete semantic IDs.

If this is right

  • Joint prediction of items and responses becomes possible through next-token modeling in one model.
  • End-to-end optimization improves recommendation performance without separate retrieval pipelines.
  • Constrained decoding ensures faithful item recommendations while supporting coherent response generation.
  • Recommendation metrics such as Recall@1 see gains of up to 29% compared to strong baselines.
  • Dialog quality remains competitive with existing approaches.

Where Pith is reading between the lines

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

  • Such a unified model could handle multi-turn conversations more fluidly by maintaining a single dependency structure across turns.
  • Applying similar factorization might improve other generative systems that mix selection and language output, such as in task completion agents.
  • The approach opens a path to parameter-efficient fine-tuning for domain-specific recommenders without retraining separate modules.

Load-bearing premise

That encoding items as discrete semantic IDs and breaking generation into sequential intent, target, and response predictions will produce accurate recommendations through constrained decoding without degrading the naturalness or coherence of the dialog responses.

What would settle it

A drop in dialog coherence or naturalness scores, or cases where the final generated response recommends an item different from the predicted target despite constrained decoding, would show the factorization harms response quality.

Figures

Figures reproduced from arXiv: 2605.21987 by Cheng Long, Mingrui Liu, Sixiao Zhang.

Figure 1
Figure 1. Figure 1: Overview of the GCRS framework. During semantic ID construction, item metadata is [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
read the original abstract

Conversational recommender systems aim to provide personalized recommendations via natural language interactions. However, existing approaches either decouple recommendation from dialog generation or rely on retrieval-based pipelines, limiting the integration between recommendation and response generation and leading to suboptimal modeling of user intent. In this paper, we propose a fully generative conversational recommender system that unifies recommendation and dialog generation within a single autoregressive framework. Our approach represents items as discrete semantic IDs and integrates them directly into the generation process, enabling joint prediction of items and responses via next-token modeling. We further introduce a structured generation paradigm that factorizes conversational recommendation into a sequence of interdependent decisions, where the model first predicts the response intent and the recommendation target, and then generates the response conditioned on them. This design enables end-to-end optimization, enforces a more coherent dependency structure, and supports faithful item generation via constrained decoding. Extensive experiments demonstrate that our method consistently improves recommendation performance, achieving gains of up to 29% on Recall@1 over strong baselines, while maintaining competitive dialog quality.

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 paper proposes a fully generative conversational recommender system that unifies recommendation and dialog generation within a single autoregressive framework. Items are represented as discrete semantic IDs integrated directly into the generation process. The model factorizes conversational recommendation into predicting response intent, then the recommendation target, followed by generating the response conditioned on these predictions, with constrained decoding to enforce valid item generation. Extensive experiments are reported to show consistent improvements in recommendation performance, including gains of up to 29% on Recall@1 over strong baselines, while maintaining competitive dialog quality.

Significance. If the empirical gains hold under rigorous controls, the work would advance conversational recommender systems by demonstrating that a single autoregressive model with semantic item IDs and structured factorization can jointly optimize recommendation and response generation more effectively than decoupled or retrieval-based approaches. The constrained decoding mechanism and end-to-end training paradigm represent a promising direction for reducing error propagation between intent modeling and item selection.

major comments (2)
  1. Abstract: The reported 29% Recall@1 improvement is presented as evidence for the superiority of the intent-then-target-then-response factorization plus constrained decoding, yet no ablation results or error analysis on target-prediction accuracy versus final recommendation metrics are described. This leaves open whether the gains survive when the intermediate target step errs, which is load-bearing for the claim that the structured paradigm produces faithful recommendations without post-hoc fixes.
  2. Abstract: The central claim that the approach 'enforces a more coherent dependency structure' and supports 'faithful item generation' rests on the assumption that early intent and target predictions are sufficiently reliable; without quantitative results on target-prediction precision or its correlation with final Recall@1, the benefit of the three-step factorization over direct generation remains unverified.
minor comments (2)
  1. Abstract: The description of baselines and datasets is absent, making it difficult to assess the strength of the 29% Recall@1 claim relative to prior work.
  2. Abstract: Statistical significance, variance across runs, or confidence intervals for the reported gains are not mentioned, which is standard for experimental claims in this area.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our work. We address each major comment below and outline the revisions we will make to strengthen the evidence for our structured generation approach.

read point-by-point responses

  1. Referee: Abstract: The reported 29% Recall@1 improvement is presented as evidence for the superiority of the intent-then-target-then-response factorization plus constrained decoding, yet no ablation results or error analysis on target-prediction accuracy versus final recommendation metrics are described. This leaves open whether the gains survive when the intermediate target step errs, which is load-bearing for the claim that the structured paradigm produces faithful recommendations without post-hoc fixes.

    Authors: We agree that the abstract and current manuscript do not provide a dedicated ablation or error analysis isolating the target-prediction step and its effect on final Recall@1. While the overall gains and constrained decoding are reported, a direct examination of cases where target prediction errs would better substantiate the claim. In the revised version we will add an ablation study reporting target-prediction precision together with an error analysis showing how often and in what way downstream recommendation metrics degrade when the intermediate prediction is incorrect. revision: yes

  2. Referee: Abstract: The central claim that the approach 'enforces a more coherent dependency structure' and supports 'faithful item generation' rests on the assumption that early intent and target predictions are sufficiently reliable; without quantitative results on target-prediction precision or its correlation with final Recall@1, the benefit of the three-step factorization over direct generation remains unverified.

    Authors: We acknowledge that quantitative evidence linking target-prediction precision to final Recall@1 is not currently presented, leaving the incremental benefit of the three-step factorization less directly verified. The manuscript emphasizes end-to-end results and the design of constrained decoding, but does not include the requested correlation analysis. We will incorporate these metrics and the corresponding correlation analysis in the revision to allow readers to assess the reliability of the intermediate predictions and the added value of the factorization. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on experimental outcomes

full rationale

The paper presents a generative conversational recommender that unifies recommendation and dialog via semantic IDs, intent-target-response factorization, and constrained decoding. All performance claims (e.g., up to 29% Recall@1 gains) are framed as results of end-to-end training and empirical comparison against baselines rather than any closed-form derivation or prediction that reduces to fitted inputs by construction. No equations, uniqueness theorems, or self-citation chains appear in the abstract or described structure that would make the central modeling choices tautological. The approach is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The approach rests on standard autoregressive language modeling assumptions and the existence of pre-trained models capable of constrained decoding; no explicit free parameters, axioms, or invented entities are introduced in the abstract.

pith-pipeline@v0.9.0 · 5700 in / 1026 out tokens · 38775 ms · 2026-05-22T04:31:36.809032+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

43 extracted references · 43 canonical work pages · 5 internal anchors

  1. [1]

    Towards knowledge-based recommender dialog system.arXiv preprint arXiv:1908.05391, 2019

    Qibin Chen, Junyang Lin, Yichang Zhang, Ming Ding, Yukuo Cen, Hongxia Yang, and Jie Tang. Towards knowledge-based recommender dialog system.arXiv preprint arXiv:1908.05391, 2019

    work page arXiv 1908

  2. [2]

    Broadening the view: Demonstration- augmented prompt learning for conversational recommendation

    Huy Dao, Yang Deng, Dung D Le, and Lizi Liao. Broadening the view: Demonstration- augmented prompt learning for conversational recommendation. InProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 785–795, 2024

    work page 2024

  3. [3]

    OneRec: Unifying Retrieve and Rank with Generative Recommender and Iterative Preference Alignment

    Jiaxin Deng, Shiyao Wang, Kuo Cai, Lejian Ren, Qigen Hu, Weifeng Ding, Qiang Luo, and Guorui Zhou. Onerec: Unifying retrieve and rank with generative recommender and iterative preference alignment, 2025. URLhttps://arxiv.org/abs/2502.18965

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  4. [4]

    Qlora: Efficient finetuning of quantized llms.Advances in neural information processing systems, 36:10088– 10115, 2023

    Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. Qlora: Efficient finetuning of quantized llms.Advances in neural information processing systems, 36:10088– 10115, 2023

    work page 2023

  5. [5]

    Leveraging large language models in conversational recommender systems.arXiv preprint arXiv:2305.07961, 2023

    Luke Friedman, Sameer Ahuja, David Allen, Zhenning Tan, Hakim Sidahmed, Changbo Long, Jun Xie, Gabriel Schubiner, Ajay Patel, Harsh Lara, et al. Leveraging large language models in conversational recommender systems.arXiv preprint arXiv:2305.07961, 2023

    work page arXiv 2023

  6. [6]

    The Llama 3 Herd of Models

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. The llama 3 herd of models.arXiv preprint arXiv:2407.21783, 2024. URL https://huggingface.co/ meta-llama/Llama-3.1-8B-Instruct

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  7. [7]

    Inspired: Toward sociable recommendation dialog systems.arXiv preprint arXiv:2009.14306, 2020

    Shirley Anugrah Hayati, Dongyeop Kang, Qingxiaoyang Zhu, Weiyan Shi, and Zhou Yu. Inspired: Toward sociable recommendation dialog systems.arXiv preprint arXiv:2009.14306, 2020

    work page arXiv 2009

  8. [8]

    Large language models as zero-shot conversational recommenders

    Zhankui He, Zhouhang Xie, Rahul Jha, Harald Steck, Dawen Liang, Yesu Feng, Bod- hisattwa Prasad Majumder, Nathan Kallus, and Julian McAuley. Large language models as zero-shot conversational recommenders. InProceedings of the 32nd ACM international conference on information and knowledge management, pages 720–730, 2023

    work page 2023

  9. [9]

    Reindex-then-adapt: Improving large language models for conversational recommendation

    Zhankui He, Zhouhang Xie, Harald Steck, Dawen Liang, Rahul Jha, Nathan Kallus, and Julian McAuley. Reindex-then-adapt: Improving large language models for conversational recommendation. InProceedings of the Eighteenth ACM International Conference on Web Search and Data Mining, pages 866–875, 2025. 10

    work page 2025

  10. [10]

    Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, and William El Sayed. Mistral 7b, 2023. URL https://hugg...

    work page 2023

  11. [11]

    Open source strikes bread - new fluffy embeddings model, 2024

    Sean Lee, Aamir Shakir, Darius Koenig, and Julius Lipp. Open source strikes bread - new fluffy embeddings model, 2024. URL https://huggingface.co/mixedbread-ai/ mxbai-embed-large-v1

    work page 2024

  12. [12]

    Incorporating external knowledge and goal guidance for llm-based conversational recommender systems.arXiv preprint arXiv:2405.01868, 2024

    Chuang Li, Yang Deng, Hengchang Hu, Min-Yen Kan, and Haizhou Li. Incorporating external knowledge and goal guidance for llm-based conversational recommender systems.arXiv preprint arXiv:2405.01868, 2024

    work page arXiv 2024

  13. [13]

    Care: Contextual adaptation of recommenders for llm-based conversational recommendation.arXiv preprint arXiv:2508.13889, 2025

    Chuang Li, Yang Deng, Hengchang Hu, See-Kiong Ng, Min-Yen Kan, and Haizhou Li. Care: Contextual adaptation of recommenders for llm-based conversational recommendation.arXiv preprint arXiv:2508.13889, 2025

    work page arXiv 2025

  14. [14]

    Improving conversational recommendation with contextual adaptation of external recommenders and llm-based reranking

    Chuang Li, Weida Liang, Hengchang Hu, See-Kiong Ng, Min-Yen Kan, Haizhou Li, and Yang Deng. Improving conversational recommendation with contextual adaptation of external recommenders and llm-based reranking. InEuropean Conference on Information Retrieval, pages 204–221. Springer, 2026

    work page 2026

  15. [15]

    Bbqrec: Behavior-bind quantization for multi-modal sequential recommendation,

    Kaiyuan Li, Rui Xiang, Yong Bai, Yongxiang Tang, Yanhua Cheng, Xialong Liu, Peng Jiang, and Kun Gai. Bbqrec: Behavior-bind quantization for multi-modal sequential recommendation,

    work page

  16. [16]

    URLhttps://arxiv.org/abs/2504.06636

    work page arXiv

  17. [17]

    Towards deep conversational recommendations.Advances in neural information processing systems, 31, 2018

    Raymond Li, Samira Ebrahimi Kahou, Hannes Schulz, Vincent Michalski, Laurent Charlin, and Chris Pal. Towards deep conversational recommendations.Advances in neural information processing systems, 31, 2018

    work page 2018

  18. [18]

    Cola: Improving conversational recommender systems by collaborative augmentation

    Dongding Lin, Jian Wang, and Wenjie Li. Cola: Improving conversational recommender systems by collaborative augmentation. InProceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 4462–4470, 2023

    work page 2023

  19. [19]

    Ministral 3

    Alexander H Liu, Kartik Khandelwal, Sandeep Subramanian, Victor Jouault, Abhinav Rastogi, Adrien Sadé, Alan Jeffares, Albert Jiang, Alexandre Cahill, Alexandre Gavaudan, et al. Ministral 3.arXiv preprint arXiv:2601.08584, 2026. URL https://huggingface.co/mistralai/ Ministral-3-8B-Instruct-2512-BF16

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  20. [20]

    Revcore: Review-augmented conversational recommendation.arXiv preprint arXiv:2106.00957, 2021

    Yu Lu, Junwei Bao, Yan Song, Zichen Ma, Shuguang Cui, Youzheng Wu, and Xiaodong He. Revcore: Review-augmented conversational recommendation.arXiv preprint arXiv:2106.00957, 2021

    work page arXiv 2021

  21. [21]

    Sentence-t5: Scalable sentence encoders from pre-trained text-to-text models

    Jianmo Ni, Gustavo Hernandez Abrego, Noah Constant, Ji Ma, Keith Hall, Daniel Cer, and Yinfei Yang. Sentence-t5: Scalable sentence encoders from pre-trained text-to-text models. In Findings of the association for computational linguistics: ACL 2022, pages 1864–1874, 2022. URLhttps://huggingface.co/sentence-transformers/sentence-t5-large

    work page 2022

  22. [22]

    A call for clarity in reporting bleu scores

    Matt Post. A call for clarity in reporting bleu scores. InProceedings of the third conference on machine translation: Research papers, pages 186–191, 2018

    work page 2018

  23. [23]

    Recommender systems with generative retrieval

    Shashank Rajput, Nikhil Mehta, Anima Singh, Raghunandan Hulikal Keshavan, Trung Vu, Lukasz Heldt, Lichan Hong, Yi Tay, Vinh Tran, Jonah Samost, Maciej Kula, Ed Chi, and Maheswaran Sathiamoorthy. Recommender systems with generative retrieval. In A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, editors,Advances in Neu- ral Information Pr...

    work page 2023

  24. [24]

    Parameter-efficient con- versational recommender system as a language processing task

    Mathieu Ravaut, Hao Zhang, Lu Xu, Aixin Sun, and Yong Liu. Parameter-efficient con- versational recommender system as a language processing task. InProceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 152–165, 2024. 11

    work page 2024

  25. [25]

    Qwen2.5: A party of foundation models, September 2024

    Qwen Team. Qwen2.5: A party of foundation models, September 2024. URL https:// huggingface.co/Qwen/Qwen2.5-7B-Instruct

    work page 2024

  26. [26]

    Qwen3 technical report, 2025

    Qwen Team. Qwen3 technical report, 2025. URL https://huggingface.co/Qwen/ Qwen3-8B

    work page 2025

  27. [27]

    LLaMA: Open and Efficient Foundation Language Models

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Tim- othée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models.arXiv preprint arXiv:2302.13971, 2023. URL https://huggingface.co/huggyllama/llama-7b

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  28. [28]

    Text Embeddings by Weakly-Supervised Contrastive Pre-training

    Liang Wang, Nan Yang, Xiaolong Huang, Binxing Jiao, Linjun Yang, Daxin Jiang, Rangan Majumder, and Furu Wei. Text embeddings by weakly-supervised contrastive pre-training. arXiv preprint arXiv:2212.03533, 2022. URL https://huggingface.co/intfloat/ e5-large-v2

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  29. [29]

    Recindial: A unified framework for conversational recommendation with pretrained language models

    Lingzhi Wang, Huang Hu, Lei Sha, Can Xu, Daxin Jiang, and Kam-Fai Wong. Recindial: A unified framework for conversational recommendation with pretrained language models. InProceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing...

    work page 2022

  30. [30]

    Towards unified conversational recommender systems via knowledge-enhanced prompt learning

    Xiaolei Wang, Kun Zhou, Ji-Rong Wen, and Wayne Xin Zhao. Towards unified conversational recommender systems via knowledge-enhanced prompt learning. InProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 1929–1937, 2022

    work page 1929

  31. [31]

    Content-based collaborative generation for recommender systems

    Yidan Wang, Zhaochun Ren, Weiwei Sun, Jiyuan Yang, Zhixiang Liang, Xin Chen, Ruobing Xie, Su Yan, Xu Zhang, Pengjie Ren, Zhumin Chen, and Xin Xin. Content-based collaborative generation for recommender systems. InProceedings of the 33rd ACM International Conference on Information and Knowledge Management, CIKM ’24, page 2420–2430, New York, NY , USA, 2024...

    work page doi:10.1145/3627673.3679692 2024

  32. [32]

    Mscrs: Multi- modal semantic graph prompt learning framework for conversational recommender systems

    Yibiao Wei, Jie Zou, Weikang Guo, Guoqing Wang, Xing Xu, and Yang Yang. Mscrs: Multi- modal semantic graph prompt learning framework for conversational recommender systems. In Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 42–52, 2025

    work page 2025

  33. [33]

    Memocrs: Memory-enhanced sequential conversational recommender systems with large language models

    Yunjia Xi, Weiwen Liu, Jianghao Lin, Bo Chen, Ruiming Tang, Weinan Zhang, and Yong Yu. Memocrs: Memory-enhanced sequential conversational recommender systems with large language models. InProceedings of the 33rd ACM International Conference on Information and Knowledge Management, pages 2585–2595, 2024

    work page 2024

  34. [34]

    Unger: Generative recommendation with a unified code via semantic and collaborative integration.ACM Trans

    Longtao Xiao, Haozhao Wang, Cheng Wang, Linfei Ji, Yifan Wang, Jieming Zhu, Zhenhua Dong, Rui Zhang, and Ruixuan Li. Unger: Generative recommendation with a unified code via semantic and collaborative integration.ACM Trans. Inf. Syst., October 2025. ISSN 1046-8188. doi: 10.1145/3773771. URLhttps://doi.org/10.1145/3773771. Just Accepted

    work page doi:10.1145/3773771 2025

  35. [35]

    C-pack: Packaged re- sources to advance general chinese embedding, 2023

    Shitao Xiao, Zheng Liu, Peitian Zhang, and Niklas Muennighoff. C-pack: Packaged re- sources to advance general chinese embedding, 2023. URL https://huggingface.co/ BAAI/bge-large-en-v1.5

    work page 2023

  36. [36]

    Improving conversational rec- ommendation systems’ quality with context-aware item meta information.arXiv preprint arXiv:2112.08140, 2021

    Bowen Yang, Cong Han, Yu Li, Lei Zuo, and Zhou Yu. Improving conversational rec- ommendation systems’ quality with context-aware item meta information.arXiv preprint arXiv:2112.08140, 2021

    work page arXiv 2021

  37. [37]

    Step: Stepwise curriculum learning for context-knowledge fusion in conversational recommendation

    Zhenye Yang, Jinpeng Chen, Huan Li, Xiongnan Jin, Xuanyang Li, Junwei Zhang, Hongbo Gao, Kaimin Wei, and Senzhang Wang. Step: Stepwise curriculum learning for context-knowledge fusion in conversational recommendation. InProceedings of the 34th ACM International Conference on Information and Knowledge Management, pages 3824–3833, 2025

    work page 2025

  38. [38]

    Multimodal quantitative language for generative recommendation, 2025

    Jianyang Zhai, Zi-Feng Mai, Chang-Dong Wang, Feidiao Yang, Xiawu Zheng, Hui Li, and Yonghong Tian. Multimodal quantitative language for generative recommendation, 2025. URL https://arxiv.org/abs/2504.05314. 12

    work page arXiv 2025

  39. [39]

    Multi-aspect cross-modal quantization for generative recommendation, 2025

    Fuwei Zhang, Xiaoyu Liu, Dongbo Xi, Jishen Yin, Huan Chen, Peng Yan, Fuzhen Zhuang, and Zhao Zhang. Multi-aspect cross-modal quantization for generative recommendation, 2025. URLhttps://arxiv.org/abs/2511.15122

    work page arXiv 2025

  40. [40]

    SiloFuse: Cross-silo Synthetic Data Generation with Latent Tabular Diffusion Models ,

    Bowen Zheng, Yupeng Hou, Hongyu Lu, Yu Chen, Wayne Xin Zhao, Ming Chen, and Ji-Rong Wen. Adapting large language models by integrating collaborative semantics for recommen- dation. In2024 IEEE 40th International Conference on Data Engineering (ICDE), pages 1435–1448, 2024. doi: 10.1109/ICDE60146.2024.00118

    work page doi:10.1109/icde60146.2024.00118 2024

  41. [41]

    Improving conversational recommender systems via knowledge graph based semantic fusion

    Kun Zhou, Wayne Xin Zhao, Shuqing Bian, Yuanhang Zhou, Ji-Rong Wen, and Jingsong Yu. Improving conversational recommender systems via knowledge graph based semantic fusion. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining, pages 1006–1014, 2020

    work page 2020

  42. [42]

    C2-crs: Coarse-to-fine contrastive learning for conversational recommender system

    Yuanhang Zhou, Kun Zhou, Wayne Xin Zhao, Cheng Wang, Peng Jiang, and He Hu. C2-crs: Coarse-to-fine contrastive learning for conversational recommender system. InProceedings of the Fifteenth ACM International Conference on Web Search and Data Mining, pages 1488–1496, 2022

    work page 2022

  43. [43]

    Collaborative retrieval for large language model-based conversational recommender systems

    Yaochen Zhu, Chao Wan, Harald Steck, Dawen Liang, Yesu Feng, Nathan Kallus, and Jundong Li. Collaborative retrieval for large language model-based conversational recommender systems. InProceedings of the ACM on Web Conference 2025, pages 3323–3334, 2025. A Collision resolution Suppose N items collide. For each colliding item, we compute the distances betw...

    work page 2025