pith. sign in

arxiv: 2605.17357 · v1 · pith:5SKZPBJInew · submitted 2026-05-17 · 💻 cs.IR · cs.MM

Dual-Diffusional Generative Fashion Recommendation

Mingzhe Yu , Lei Wu , Qianru Sun , Yunshan Ma This is my paper

Pith reviewed 2026-05-19 23:09 UTC · model grok-4.3

classification 💻 cs.IR cs.MM
keywords fashion recommendationgenerative modelsdiffusion modelsmulti-modal learningpersonalized recommendationinterpretabilitytransformeroutfit recommendation
0
0 comments X

The pith

A dual-diffusion Transformer generates both fashion item images and textual descriptions for personalized recommendations.

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

The paper introduces DualFashion to overcome limitations in existing generative fashion recommenders that rely on implicit visual embeddings containing irrelevant information. These methods often fail to model user behavior adequately and lack interpretability by generating only images. DualFashion uses a dual-diffusion Transformer with image and text branches conditioned on structured attribute-level captions and visual outfit information from historical interactions. It generates both images and text for visual compatibility and semantic explanations, supported by a text-augmented fine-tuning strategy for diversity and efficiency. A sympathetic reader would care because this could lead to more accurate, understandable, and computationally efficient personalized fashion suggestions.

Core claim

DualFashion is a Dual-Diffusional Generative Fashion Recommendation Architecture that jointly models image and text modalities for personalized and explainable recommendation. It adopts a dual-diffusion Transformer with image and text branches, where structured attribute-level captions and visual outfit information are jointly used as conditioning signals to model user behavior. The architecture produces both fashion item images and textual descriptions, ensuring visual compatibility while providing explicit semantic interpretability, and uses a text-augmented fine-tuning strategy to enhance generation diversity and enable effective cross-modal knowledge transfer without heavy computational

What carries the argument

Dual-diffusion Transformer with image and text branches conditioned on structured attribute-level captions and visual outfit information from historical interactions.

Load-bearing premise

Conditioning the dual-diffusion Transformer on structured attribute-level captions and visual outfit information from historical interactions sufficiently removes preference-irrelevant information and accurately models user behavior.

What would settle it

Observing no significant improvement in recommendation accuracy or user preference alignment when comparing DualFashion outputs to baselines on the iFashion dataset would challenge the central claim.

Figures

Figures reproduced from arXiv: 2605.17357 by Lei Wu, Mingzhe Yu, Qianru Sun, Yunshan Ma.

Figure 1
Figure 1. Figure 1: Comparison between our dual-diffusional architec [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The multi-stage training of DualFashion consists of warm-up, matching-aware personalized multimodal training, and [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Ablation study about the alignment between fash [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Model-wise comparison of different models’ generative capabilities on the GOR task. Our DualFashion generates [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison on the PFITB task. Two generated im [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Experimental Evaluation. (a) Comparison of interpretability ability between our model architecture and post-hoc [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Time cost analysis of the baseline and our Dual [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
read the original abstract

Personalized generative recommender systems have emerged as a promising solution for fashion recommendation. However, existing methods primarily rely on implicit visual embeddings from historical interactions, which often contain preference-irrelevant information and result in insufficient user behavior modeling. Moreover, these models typically generate only item images, providing limited interpretability. To address these limitations, we propose DualFashion, a Dual-Diffusional Generative Fashion Recommendation Architecture that jointly models image and text modalities for personalized and explainable recommendation. DualFashion adopts a dual-diffusion Transformer with image and text branches, where structured attribute-level captions and visual outfit information are jointly used as conditioning signals to model user behavior. The proposed architecture produces both fashion item images and textual descriptions, ensuring visual compatibility while providing explicit semantic interpretability. Furthermore, we introduce a text-augmented fine-tuning strategy that enhances generation diversity and enables effective cross-modal knowledge transfer without incurring heavy computational costs. Extensive experiments on iFashion and Polyvore-U across Personalized Fill-in-the-Blank and Generative Outfit Recommendation tasks demonstrate that DualFashion achieves strong performance in behavior modeling, interpretability, and efficiency compared to state-of-the-art methods. Our code and model checkpoints are available at https://github.com/LinkMingzhe/DualFashion.

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 introduces DualFashion, a Dual-Diffusional Generative Fashion Recommendation Architecture using a dual-diffusion Transformer with image and text branches. Structured attribute-level captions and visual outfit information from historical interactions serve as conditioning signals to model user behavior and generate both item images and textual descriptions for personalized and explainable recommendations. A text-augmented fine-tuning strategy is proposed for diversity and cross-modal transfer. Experiments on iFashion and Polyvore-U for Personalized Fill-in-the-Blank and Generative Outfit Recommendation tasks show strong performance in behavior modeling, interpretability, and efficiency versus state-of-the-art methods.

Significance. If the results hold, DualFashion advances generative fashion recommenders by jointly handling modalities for better user modeling and explicit interpretability through text outputs. The open availability of code and checkpoints at the provided GitHub link is a positive aspect for the community. This could influence future work on diffusion models in recommendation systems.

major comments (2)
  1. The abstract asserts that joint conditioning on structured attribute-level captions and visual outfit information models user behavior more accurately by removing preference-irrelevant information from historical visual embeddings. However, the manuscript provides no explicit mechanism (e.g., attention masking, disentanglement loss) or diagnostic (e.g., attention maps, caption-quality ablation) to demonstrate suppression of irrelevant visual factors rather than mere co-presence; performance gains on the two tasks could therefore stem from the diffusion architecture or text-augmented fine-tuning instead.
  2. §5 (Experiments): the reported superiority on iFashion and Polyvore-U for both Personalized Fill-in-the-Blank and Generative Outfit Recommendation lacks reported statistical significance, run-to-run variance, or confirmation that baselines were re-implemented under identical hyper-parameter protocols, which is load-bearing for the central claim of improved behavior modeling.
minor comments (2)
  1. The notation and forward process for the dual-diffusion Transformer would benefit from an explicit equation or pseudocode block in the model section to improve clarity.
  2. Figure captions for qualitative generation examples should explicitly state the conditioning inputs used for each sample.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive summary and constructive major comments. We address each point below with the strongest honest defense and indicate planned revisions.

read point-by-point responses

  1. Referee: The abstract asserts that joint conditioning on structured attribute-level captions and visual outfit information models user behavior more accurately by removing preference-irrelevant information from historical visual embeddings. However, the manuscript provides no explicit mechanism (e.g., attention masking, disentanglement loss) or diagnostic (e.g., attention maps, caption-quality ablation) to demonstrate suppression of irrelevant visual factors rather than mere co-presence; performance gains on the two tasks could therefore stem from the diffusion architecture or text-augmented fine-tuning instead.

    Authors: We thank the referee for this precise observation. Section 3 details that the dual-diffusion Transformer processes the image branch under joint conditioning from both visual outfit embeddings and structured attribute captions produced by the text branch. The cross-attention layers between branches force the image denoising process to respect explicit semantic constraints (e.g., color, style, category), which inherently down-weights preference-irrelevant visual factors present in raw historical embeddings. This is not mere co-presence; the text branch supplies an independent supervisory signal that the image branch must satisfy at every diffusion step. While we did not add an explicit disentanglement loss or attention visualizations in the original submission, the architecture description and the text-augmented fine-tuning objective already encode this filtering effect. To make the claim fully explicit, we will add (i) qualitative attention maps between modalities and (ii) a caption-quality ablation in the revised manuscript. revision: partial

  2. Referee: §5 (Experiments): the reported superiority on iFashion and Polyvore-U for both Personalized Fill-in-the-Blank and Generative Outfit Recommendation lacks reported statistical significance, run-to-run variance, or confirmation that baselines were re-implemented under identical hyper-parameter protocols, which is load-bearing for the central claim of improved behavior modeling.

    Authors: We agree that statistical rigor strengthens the central claim. All reported numbers in §5 are averages over multiple random seeds; however, standard deviations and significance tests were omitted from the tables. We will revise the experimental section to report mean ± std over five independent runs and include paired t-test p-values against the strongest baseline for each metric. Regarding re-implementation, the baselines were executed from their official repositories (or re-coded from the original papers) using the exact dataset splits and task definitions provided in the respective works, with only minimal hyper-parameter adjustments needed for compatibility with our evaluation protocol. We will add an explicit paragraph in §5.1 documenting these protocols and the hyper-parameter tables used for each baseline. revision: yes

Circularity Check

0 steps flagged

No circularity: architecture and results are empirically grounded on external benchmarks

full rationale

The paper proposes DualFashion as a dual-diffusion Transformer conditioned on attribute-level captions and visual outfit data, then reports performance on iFashion and Polyvore-U for fill-in-the-blank and generative recommendation tasks. No equations, derivations, or fitted-parameter predictions appear in the abstract or described content. Claims rest on comparative experiments against prior methods rather than any self-referential definition, self-citation chain, or renaming of known results. The central modeling assumption is stated explicitly but is not shown to reduce to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides insufficient technical detail to enumerate specific free parameters, axioms, or invented entities; typical diffusion models involve many hyperparameters and the new architecture is presented as the main contribution.

pith-pipeline@v0.9.0 · 5749 in / 1021 out tokens · 37051 ms · 2026-05-19T23:09:21.779354+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.

Reference graph

Works this paper leans on

52 extracted references · 52 canonical work pages · 5 internal anchors

  1. [1]

    Andrew Brock, Jeff Donahue, and Karen Simonyan. 2019. Large Scale GAN Training for High Fidelity Natural Image Synthesis. InICLR. OpenReview.net

    work page 2019

  2. [2]

    Miaomiao Cai, Lei Chen, Yifan Wang, Haoyue Bai, Peijie Sun, Le Wu, Min Zhang, and Meng Wang. 2024. Popularity-Aware Alignment and Contrast for Mitigating Popularity Bias. InKDD. ACM, 187–198

    work page 2024

  3. [3]

    Miaomiao Cai, Zhijie Zhang, Junfeng Fang, Zhiyong Cheng, Xiang Wang, and Meng Wang. 2026. RMBRec: Robust Multi-Behavior Recommendation towards Target Behaviors. InWWW. ACM, 6731–6742

    work page 2026

  4. [4]

    Wen Chen, Pipei Huang, Jiaming Xu, Xin Guo, Cheng Guo, Fei Sun, Chao Li, Andreas Pfadler, Huan Zhao, and Binqiang Zhao. 2019. POG: Personalized Outfit Generation for Fashion Recommendation at Alibaba iFashion. InKDD. ACM, 2662–2670

    work page 2019

  5. [5]

    Hao Cheng, Shuo Wang, Wensheng Lu, Wei Zhang, Mingyang Zhou, Kezhong Lu, and Hao Liao. 2023. Explainable Recommendation with Personalized Re- view Retrieval and Aspect Learning. InACL (1). Association for Computational Linguistics, 51–64

    work page 2023

  6. [6]

    Yujuan Ding, Yunshan Ma, Wai Keung Wong, and Tat-Seng Chua. 2021. Lever- aging Two Types of Global Graph for Sequential Fashion Recommendation. In ICMR. ACM, 73–81

    work page 2021

  7. [7]

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, Dustin Podell, Tim Dockhorn, Zion English, and Robin Rombach. 2024. Scaling Rectified Flow Transformers for High-Resolution Image Synthesis. InICML. OpenReview.net

    work page 2024

  8. [8]

    Rinon Gal, Yuval Alaluf, Yuval Atzmon, Or Patashnik, Amit Haim Bermano, Gal Chechik, and Daniel Cohen-Or. 2023. An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion. InICLR. OpenReview.net

    work page 2023

  9. [9]

    Generative Adversarial Networks

    Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde- Farley, Sherjil Ozair, Aaron C. Courville, and Yoshua Bengio. 2014. Generative Adversarial Networks.CoRRabs/1406.2661 (2014)

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  10. [10]

    Xintong Han, Zuxuan Wu, Yu-Gang Jiang, and Larry S. Davis. 2017. Learning Fashion Compatibility with Bidirectional LSTMs. InACM Multimedia. ACM, 1078–1086

    work page 2017

  11. [11]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising Diffusion Probabilistic Models. InNeurIPS

    work page 2020

  12. [12]

    Jonathan Ho and Tim Salimans. 2022. Classifier-Free Diffusion Guidance.CoRR abs/2207.12598 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  13. [13]

    Yupeng Hou, An Zhang, Leheng Sheng, Zhengyi Yang, Xiang Wang, Tat-Seng Chua, and Julian J. McAuley. 2025. Generative Recommendation Models: Progress and Directions. InWWW (Companion Volume). ACM, 13–16

    work page 2025

  14. [14]

    Wang-Cheng Kang, Chen Fang, Zhaowen Wang, and Julian J. McAuley. 2017. Visually-Aware Fashion Recommendation and Design with Generative Image Models. InICDM. IEEE Computer Society, 207–216

    work page 2017

  15. [15]

    Tero Karras, Samuli Laine, and Timo Aila. 2019. A Style-Based Generator Ar- chitecture for Generative Adversarial Networks. InCVPR. Computer Vision Foundation / IEEE, 4401–4410

    work page 2019

  16. [16]

    Jeongho Kim, Gyojung Gu, Minho Park, Sunghyun Park, and Jaegul Choo. 2023. StableVITON: Learning Semantic Correspondence with Latent Diffusion Model for Virtual Try-On.CoRRabs/2312.01725 (2023)

    work page arXiv 2023

  17. [17]

    Zijie Li, Henry Li, Yichun Shi, Amir Barati Farimani, Yuval Kluger, Linjie Yang, and Peng Wang. 2025. Dual Diffusion for Unified Image Generation and Under- standing. InCVPR. Computer Vision Foundation / IEEE, 2779–2790

    work page 2025

  18. [18]

    Chang Liu, Yimeng Bai, Xiaoyan Zhao, Yang Zhang, Fuli Feng, and Wenge Rong

    work page

  19. [19]

    DiscRec: Disentangled Semantic-Collaborative Modeling for Generative Recommendation.CoRRabs/2506.15576 (2025)

    work page arXiv 2025

  20. [20]

    Xiaohao Liu, Zhulin Tao, Jiahong Shao, Lifang Yang, and Xianglin Huang. 2022. EliMRec: Eliminating Single-modal Bias in Multimedia Recommendation. InACM Multimedia. ACM, 687–695

    work page 2022

  21. [21]

    Zakirul Alam Bhuiyan

    Xiangyong Liu, Guojun Wang, and Md. Zakirul Alam Bhuiyan. 2022. Personalised context-aware re-ranking in recommender system.Connect. Sci.34, 1 (2022), 319–338

    work page 2022

  22. [22]

    Xiaohao Liu, Xiaobo Xia, See-Kiong Ng, and Tat-Seng Chua. 2025. Principled Multimodal Representation Learning.CoRRabs/2507.17343 (2025)

    work page arXiv 2025

  23. [23]

    Simian Luo, Yiqin Tan, Longbo Huang, Jian Li, and Hang Zhao. 2023. Latent Con- sistency Models: Synthesizing High-Resolution Images with Few-Step Inference. CoRRabs/2310.04378 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  24. [24]

    Yunshan Ma, Yingzhi He, Wenjun Zhong, Xiang Wang, Roger Zimmermann, and Tat-Seng Chua. 2024. CIRP: Cross-Item Relational Pre-training for Multimodal Product Bundling. InACM Multimedia. ACM, 9641–9649

    work page 2024

  25. [25]

    Yunshan Ma, Xiaohao Liu, Yinwei Wei, Zhulin Tao, Xiang Wang, and Tat-Seng Chua. 2024. Leveraging Multimodal Features and Item-level User Feedback for Bundle Construction. InWSDM. ACM, 510–519

    work page 2024

  26. [26]

    Mehdi Mirza and Simon Osindero. 2014. Conditional Generative Adversarial Nets.CoRRabs/1411.1784 (2014)

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  27. [27]

    Maryam Moosaei, Yusan Lin, Ablaikhan Akhazhanov, Huiyuan Chen, Fei Wang, and Hao Yang. 2022. OutfitGAN: Learning Compatible Items for Generative Fashion Outfits. InCVPR Workshops. IEEE, 2272–2276

    work page 2022

  28. [28]

    William Peebles and Saining Xie. 2023. Scalable Diffusion Models with Trans- formers. InICCV. IEEE, 4172–4182

    work page 2023

  29. [29]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning Transferable Visual Models From Natural Language Supervision. InICML (Proceedings of Machine Learning Research, Vol. 139). PMLR, 8748–8763

    work page 2021

  30. [30]

    Tran, Jonah Samost, Maciej Kula, Ed H

    Shashank Rajput, Nikhil Mehta, Anima Singh, Raghunandan Hulikal Keshavan, Trung Vu, Lukasz Heldt, Lichan Hong, Yi Tay, Vinh Q. Tran, Jonah Samost, Maciej Kula, Ed H. Chi, and Mahesh Sathiamoorthy. 2023. Recommender Systems with Generative Retrieval. InNeurIPS

    work page 2023

  31. [31]

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-Resolution Image Synthesis with Latent Diffusion Models. InCVPR. IEEE, 10674–10685

    work page 2022

  32. [32]

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. 2023. DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation. InCVPR. IEEE, 22500–22510

    work page 2023

  33. [33]

    Xiaoteng Shen, Rui Zhang, Xiaoyan Zhao, Jieming Zhu, and Xi Xiao. 2024. PMG : Personalized Multimodal Generation with Large Language Models. InWWW. ACM, 3833–3843

    work page 2024

  34. [34]

    Jiaming Song, Chenlin Meng, and Stefano Ermon. 2021. Denoising Diffusion Implicit Models. InICLR. OpenReview.net

    work page 2021

  35. [35]

    Kingma, Abhishek Kumar, Ste- fano Ermon, and Ben Poole

    Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Ste- fano Ermon, and Ben Poole. 2021. Score-Based Generative Modeling through Stochastic Differential Equations. InICLR. OpenReview.net

    work page 2021

  36. [36]

    Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, and Zbig- niew Wojna. 2016. Rethinking the Inception Architecture for Computer Vision. InCVPR. IEEE Computer Society, 2818–2826

    work page 2016

  37. [37]

    Gemini Team. 2025. Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities.CoRR abs/2507.06261 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  38. [38]

    Bohao Wang, Feng Liu, Jiawei Chen, Xingyu Lou, Changwang Zhang, Jun Wang, Yuegang Sun, Yan Feng, Chun Chen, and Can Wang. 2025. MSL: Not All Tokens Are What You Need for Tuning LLM as a Recommender. InSIGIR. ACM, 1912– 1922

    work page 2025

  39. [39]

    Bohao Wang, Feng Liu, Changwang Zhang, Jiawei Chen, Yudi Wu, Sheng Zhou, Xingyu Lou, Jun Wang, Yan Feng, Chun Chen, and Can Wang. 2026. LLM4DSR: Leveraging Large Language Model for Denoising Sequential Recommendation. ACM Trans. Inf. Syst.44, 1 (2026), 6:1–6:32

    work page 2026

  40. [40]

    Xiang Wang, Hongye Jin, An Zhang, Xiangnan He, Tong Xu, and Tat-Seng Chua

    work page

  41. [41]

    Disentangled Graph Collaborative Filtering. InSIGIR. ACM, 1001–1010

    work page

  42. [42]

    Yu Wang, Lei Sang, Yi Zhang, and Yiwen Zhang. 2025. Intent Representation Learning with Large Language Model for Recommendation. InSIGIR. ACM, 1870–1879

    work page 2025

  43. [43]

    Yu Wang, Yonghui Yang, Le Wu, Jiancan Wu, Hefei Xu, and Hui Lin. 2026. MLLMRec-R1: Incentivizing Reasoning Capability in Large Language Models for Multimodal Sequential Recommendation.CoRRabs/2603.06243 (2026)

    work page arXiv 2026

  44. [44]

    Yiyan Xu, Wenjie Wang, Fuli Feng, Yunshan Ma, Jizhi Zhang, and Xiangnan He

    work page

  45. [45]

    Diffusion Models for Generative Outfit Recommendation. InSIGIR. ACM, 1350–1359

    work page

  46. [46]

    Yiyan Xu, Wenjie Wang, Yang Zhang, Biao Tang, Peng Yan, Fuli Feng, and Xiangnan He. 2025. Personalized Image Generation with Large Multimodal Models. InWWW. ACM, 264–274

    work page 2025

  47. [47]

    Yiyan Xu, Jinghao Zhang, Alireza Salemi, Xinting Hu, Wenjie Wang, Fuli Feng, Hamed Zamani, Xiangnan He, and Tat-Seng Chua. 2025. Personalized Genera- tion In Large Model Era: A Survey. InACL (1). Association for Computational Linguistics, 24607–24649

    work page 2025

  48. [48]

    Kai Yang, Jian Tao, Jiafei Lyu, Chunjiang Ge, Jiaxin Chen, Qimai Li, Weihan Shen, Xiaolong Zhu, and Xiu Li. 2023. Using Human Feedback to Fine-tune Diffusion Models without Any Reward Model.CoRRabs/2311.13231 (2023)

    work page arXiv 2023

  49. [49]

    Mingzhe Yu, Yunshan Ma, Lei Wu, Kai Cheng, Xue Li, Lei Meng, and Tat-Seng Chua. 2024. Smart Fitting Room: A One-stop Framework for Matching-aware Virtual Try-On. InICMR. ACM, 184–192

    work page 2024

  50. [50]

    Mingzhe Yu, Yunshan Ma, Lei Wu, Changshuo Wang, Xue Li, and Lei Meng. 2025. FashionDPO: Fine-tune Fashion Outfit Generation Model using Direct Preference Optimization. InSIGIR. ACM, 212–222

    work page 2025

  51. [51]

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. 2023. Adding Conditional Control to Text-to-Image Diffusion Models. InICCV. IEEE, 3813–3824

    work page 2023

  52. [52]

    Lixi Zhu, Xiaowen Huang, and Jitao Sang. 2024. How Reliable is Your Simulator? Analysis on the Limitations of Current LLM-based User Simulators for Conver- sational Recommendation. InWWW (Companion Volume). ACM, 1726–1732

    work page 2024