Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

FashionM3: Multimodal, Multitask, and Multiround Fashion Assistant based on Unified Vision-Language Model

T0 review · 5 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A 2B fashion model beats GPT-4o at styling advice

desk verdict The FashionRec dataset is a real contribution, but the superiority claim over GPT-4o rests on a self-referential evaluation that doesn't yet support it. read the letter →

arxiv 2504.17826 v1 pith:SXCM73B7 submitted 2025-04-24 cs.CV cs.AI

Kaicheng Pang , Xingxing Zou , Waikeung Wong This is my paper · ORCID
classification cs.CVcs.AI
keywords FashionM3recommendationvision-languagemodelRecdatasetmultimodaldialoguepersonalizationvirtualtry-ongenerative
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to show that one compact vision-language model, fine-tuned on a newly assembled dataset of fashion recommendation dialogues, can act as a complete fashion assistant: it recommends items to complete an outfit, tailors suggestions to a user's past interactions, proposes alternatives, generates product images, and revises its advice over conversation rounds. The authors argue this generative approach sidesteps the combinatorial explosion that makes ranking-based outfit recommendation impractical, because the model produces contextually relevant suggestions directly instead of scoring huge candidate sets. The result would matter because online fashion retail has struggled to combine visual understanding, natural-language dialogue, and personalization in a single system, and a small model that does all three runs cheaply enough for a single consumer GPU.

What carries the argument

FashionVLM is the load-bearing component: a unified vision-language model based on the Show-O architecture, fine-tuned on FashionRec with two objectives—next-token prediction for recommendation text and discrete-diffusion masked-token prediction for image generation. FashionRec is the other essential mechanism: a pipeline that decomposes human-curated outfits into partial outfits and target items, filters each user's interaction history with CLIP similarity and interaction frequency, finds replacement items across overlapping outfits, and asks an LLM to write natural dialogues for each scenario. Around these, FashionM3 uses the Model Context Protocol (MCP), a tool-orchestration standard, to connect FashionVLM to a user database, a product database, virtual try-on, and online search, enabling multiround refinement.

What would settle it

Run the same test prompts against reference responses written by professional fashion stylists instead of the LLM, then measure S-BERT and CLIP agreement; if FashionVLM no longer beats GPT-4o, the reported superiority is a property of the synthetic benchmark. Alternatively, run a larger user study with logging of whether participants actually wear or buy the recommended items.

Watch

Extended reading notes

Core claim

The central claim is that fine-tuning a unified vision-language model on 331,124 synthetic multimodal dialogues—covering basic, personalized, and alternative recommendation—teaches it both fashion aesthetics and conversational recommendation behavior, so that a single 2B-parameter model can generate text recommendations and product images that score higher on semantic and visual similarity to held-out targets than off-the-shelf 11B and 200B vision-language models. The paper treats the generated dialogues as distilled fashion knowledge: human-curated outfits supply aesthetic coherence, user interaction histories supply preference, and an LLM converts these into natural question-answer pairs. FashionVLM is trained with a next-token prediction loss over recommendation responses and a masked-token diffusion loss over image tokens, letting one network handle both understanding and generation.

Load-bearing premise

The evaluation assumes that agreeing with the LLM-generated reference dialogues and images is the same as giving good fashion advice, and the same generation pipeline produced both the training data and the test targets.

Editorial extensions

If this is right

  • Fashion recommendation no longer needs to enumerate and score candidate outfits; a single generative model can answer open-ended styling queries directly, bypassing the combinatorial explosion that limits ranking-based systems.
  • One fine-tuned model covers basic, personalized, and alternative recommendation, product image generation, and virtual try-on, so users can stay in one conversation instead of switching between task-specific tools.
  • With multiround interaction, the assistant can revise its text suggestion and regenerate the product image as the user adds feedback, as the system walkthrough demonstrates.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the test targets and the training dialogues come from the same LLM prompting pipeline, an evaluation against human-written styling advice would be a harder test; the paper does not provide one.
  • The personalization metric reuses CLIP-based preference averaging similar to the filtering logic used to build training data, so a cleaner test would log real user choices or purchases rather than similarity scores.
  • The same data-construction recipe should transfer to other curated-product domains (home decor, gifts, travel kits), with the caveat that the resulting assistant inherits both the LLM's dialogue style and the curators' taste.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes FashionM3, a multimodal, multitask, and multiround fashion assistant built by fine-tuning the 2B-parameter Show-O vision-language model on a new synthetic dialogue dataset called FashionRec (331,124 samples covering basic, personalized, and alternative recommendation tasks). The system integrates FashionVLM with user/product databases, image generation, virtual try-on, and online search via the Model Context Protocol. The authors report quantitative comparisons against off-the-shelf GPT-4o, LLaMA-3.2-vision, and Show-O using S-BERT, CLIP text similarity, CLIP image score, and a personalization metric, claim that FashionVLM matches or exceeds these baselines despite its small size, and present qualitative examples and a 12-participant user study to support practical value.

Significance. If the claims are substantiated, the paper makes a useful contribution by demonstrating how a compact 2B VLM fine-tuned on a large synthetic fashion dialogue dataset can approach or exceed much larger general-purpose models on fashion-specific recommendation tasks. The FashionRec dataset, with 331k samples across three task types, is a valuable resource for the community and is publicly released. The modular MCP-based system architecture is also a practical strength. However, the quantitative evaluation is fundamentally self-referential: training and test targets are produced by the same LLM prompting pipeline, and the personalization metric mirrors the training-data filtering logic. The central claim of 'superior performance' therefore depends on whether the metrics measure genuine fashion advice quality rather than imitation of the LLM's dialogue style. The paper would be strengthened by an independent human evaluation with baseline comparisons or by evaluation on a benchmark with human-authored ground truth.

major comments (5)
  1. [Section III and V-B, Table III] The quantitative evaluation is self-referential. Both the training and test target responses are generated by the same LLM prompting pipeline described in Section III and Appendix A, while the baselines (GPT-4o, LLaMA-3.2-vision, Show-O) are zero-shot and never saw this generation style. Consequently, FashionVLM's higher S-BERT/CTS/CIS scores may reflect fine-tuning to imitate the LLM's dialogue structure, phrasing, and preference-injection style rather than superior fashion advice. The paper needs an independent evaluation, e.g., a blind expert/user comparison of recommendations from FashionVLM and the baselines, or a test set with human-authored ground truth, to support the claim of superior recommendation effectiveness.
  2. [Section V-A, 'Personalization (Per.)' and Algorithm 1] The personalization metric is circular with the training-data construction. The Per. metric computes the CLIP cosine similarity between the averaged embedding of the user's historical items and the generated image, which is essentially the same scoring logic used in Algorithm 1 (lines 25-29) to select the filtered user history U'c before generating training dialogues. Thus a model that matches the algorithm's CLIP-based preferences will score higher by construction, and the Per. results in Table III do not independently establish better preference matching. The paper should either use a different personalization measure (e.g., human judgments or a held-out preference prediction task not derived from the same CLIP filter) or explicitly justify why this metric is not self-referential.
  3. [Section V-B, Table III] There are numerical inconsistencies between Table III and the text in Section V-B. For the Personalized Recommendation task, the text states that FashionVLM's Per. score of 84.08 'closely trails GPT-4o's 85.51,' but Table III lists GPT-4o's Per. score as 83.09 (and FashionVLM's 84.08 actually exceeds it). For the Alternative Recommendation task, the text reports GPT-4o's S-BERT as 68.05 and CIS as 77.28, whereas Table III lists 67.06 and 75.38, respectively. These discrepancies undermine the reliability of the reported comparisons and must be corrected, with the text and table reconciled.
  4. [Section V-A and V-D] No error bars or significance tests are reported for any of the quantitative metrics, and the user study has only 12 participants with no control condition or comparison to alternative systems. The margins over GPT-4o are small (e.g., personalized S-BERT 78.54 vs 76.61; Per. 84.08 vs 83.09), so without repeated runs, confidence intervals, or statistical tests, these differences cannot be distinguished from noise. Similarly, the user study cannot support 'superior performance' claims because participants only interacted with FashionM3 and had no baseline for comparison. Please report variance across evaluation subsets or multiple inference runs and include a controlled or comparative user study.
  5. [Section V-A, Baselines] The comparison against GPT-4o and LLaMA-3.2-vision is skewed because these models are used off-the-shelf and are not fine-tuned or otherwise adapted to the FashionRec dialogue distribution. Only Show-O (the base model) is a direct fine-tuning baseline. The observed gains may simply reflect in-distribution fine-tuning, not a general advantage in fashion capability. A fair test of the claim that FashionVLM outperforms larger models should include baselines that are fine-tuned on the same training data (or otherwise given in-context examples of the expected style), or the paper should limit its claim to 'fine-tuning on FashionRec improves a 2B VLM relative to zero-shot general VLMs.'
minor comments (6)
  1. [Section IV-B and Fig. 3] The role of GPT-4o in the FashionM3 architecture is not fully clear. The text says 'The OpenAI GPT-4o model then processes these inputs to interpret the query and invoke appropriate tools,' which suggests GPT-4o is the orchestrator, while FashionVLM is the recommendation engine. Please clarify the division of labor between GPT-4o and FashionVLM and whether user queries are always routed through GPT-4o, as this affects the interpretation of the user study and the system's practical deployment.
  2. [Section III, Algorithm 1 and Appendix A] The 'valid flag' instruction in the personalized recommendation prompt (Appendix A) asks the LLM to judge whether historical items align with the target item and to set 'valid' to 0 if they do not. It is unclear whether samples labeled invalid are excluded from the training and test splits. If they are excluded, the test set may be easier than real user scenarios, and this should be stated explicitly.
  3. [Appendix A, Alternative Recommendation prompt] The alternative-recommendation system prompt contains a copied guideline from the basic-recommendation prompt: 'Ensure the user's questions collectively mention all desired categories for recommendations, covering every target item in the set by the end of the conversation.' Since the alternative task explicitly requires exactly one round and one replaceable item, this guideline is confusing and should be removed or adapted.
  4. [Section V-B, Table III] The caption of Table III says 'off-the-shelf VLM' but Show-O is the base model that FashionVLM is fine-tuned from; consider calling it 'baseline VLM' or clarifying that Show-O is used without fine-tuning.
  5. [Section VI] There is a typo: 'user's prefence' should be 'user's preference.' Also, the limitations section is brief and does not mention the self-referential evaluation issue, which is a significant limitation of the current study.
  6. [Section V-D, Fig. 6] The user study questionnaire results are presented as stacked bar counts without significance testing or a control condition. Please provide at least descriptive statistics (e.g., means and standard deviations per question) and clarify whether the positively/negatively framed questions were reverse-coded before aggregation.

Circularity Check

2 steps flagged · score 6.0 of 10

Benchmark is partly self-referential: test ground-truth dialogues come from the same LLM prompt pipeline used to generate training data, and the Personalization metric reuses the CLIP history-filtering logic from Algorithm 1.

  1. fitted input called prediction [Section III (FashionRec Dataset), Section V-A/V-B (Datasets and Metrics), Appendix A]
    "We asked LLM to utilize the partial outfit P, target item set T, and filtered user history U′c as input to generate user queries Xq and corresponding responses Xa. ... We utilize the test split of FashionRec datasets to compare recommendation performance with baselines."

    The headline comparison (Table III) is scored against test references produced by the same LLM prompting pipeline (Appendix A) that generated the training dialogues. FashionVLM is fine-tuned with a next-token objective on exactly these LLM-written responses (Eq. 2), so S-BERT/CTS/CIS largely measure imitation of the data generator's phrasing, preference-injection style, and item-description format, rather than better fashion advice. The zero-shot GPT-4o and LLaMA-3.2 baselines never saw this generation distribution, so the reported margins (e.g., +6.66 S-BERT on Basic, +6.94 on Alternative) are partly an in-distribution advantage. Because the test split is disjoint this is not full circularity, but the evaluation target is defined by the same generative process as the training target.

  2. self definitional [Section V-A (Evaluation Metrics, Personalization) vs Algorithm 1 / Eq. (1)]
    "Personalization (Per.) measures how well recommendations align with user preferences. It first calculates the preference feature by averaging the CLIP image features of a user's past interactions. Then, it computes the cosine similarity between this preference feature and embeddings of generated images. ... for each item j∈Uc, we calculate a similarity score using cosine similarity... The top-k items in Uc, ranked by score(j) in descending order, form the filtered set U′c."

    The Per. metric's 'preference feature' is the same CLIP-average-of-user-history construction that Algorithm 1 uses to compute sim(j) and to select the filtered history U′c that is fed to the LLM when writing training dialogues. A model trained on U′c-conditioned dialogues is therefore, by construction, trained to produce items with high cosine similarity to the CLIP average of the user's history—exactly what Per. then measures. The personalization score is partly an echo of the data-construction filter rather than an independent check of whether the recommendation matches genuine user preferences.

full rationale

The central empirical claim—FashionVLM (2B) surpassing GPT-4o (200B) on recommendation effectiveness—rests on Table III, which is scored against FashionRec's test split. Both the training and test ground-truth dialogues are generated by the same LLM prompting pipeline described in Section III and Appendix A, so the S-BERT/CTS/CIS scores largely reward reproducing the LLM's dialogue style and preference-injection format. The zero-shot baselines are not adapted to this distribution, making the comparison skewed. The personalization metric is more sharply circular: it averages CLIP features of the user's history and computes cosine similarity to generated images, while Algorithm 1 uses the same CLIP cosine-similarity logic to filter which historical items enter the training dialogues. Thus the metric's definition is built from the same operation that shapes the training data. The user study is human-based and offers some external evidence of practical value, but it has no baseline arm, so it cannot independently establish superiority over GPT-4o. No load-bearing self-citation chain or imported uniqueness theorem was found; the authors' self-citations are ordinary prior-art references. Overall, the central comparative claim is partially circular, warranting a score of 6 rather than 8 because the test split is held out and the user study provides some independent signal.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

No new physical or conceptual entities are introduced; the only new artifact is the FashionRec dataset, which is not an invented entity in the sense of this ledger.

free parameters (5)
  • mu = 10
    Minimum user history interactions threshold in Algorithm 1, set based on empirical experience; 43% data excluded with thresholds.
  • mi = 3
    Minimum item co-occurrence threshold in Algorithm 1, set based on empirical experience.
  • alpha = 3
    Weight for outfit coherence term in selecting optimal pair r* (Section III), chosen to emphasize outfit coherence over user history size.
  • beta = 2.0
    Amplification factor for frequently interacted items in score(j) (Eq. 1), set without justification.
  • top-k = unspecified
    Number of items selected from ranked U_c in Algorithm 1; value is not stated in the paper.
assumptions (4)
  • domain assumption Human-curated outfits are aesthetically coherent.
    The FashionRec construction assumes that outfits in iFashion, Polyvore-519, and Fashion32 are ground-truth style examples (Section III).
  • domain assumption LLM-generated dialogues are high-quality fashion advice.
    All training and test responses are generated by an LLM from prompts in Appendix A; no human validation of the dialogues is reported.
  • domain assumption CLIP embeddings capture fashion compatibility and preference similarity.
    Used both in dataset filtering (cosine similarity) and in evaluation metrics (CTS, CIS, Per.) (Sections III and V-A).
  • domain assumption Similarity to LLM-generated ground truth is a valid measure of recommendation quality.
    S-BERT, CTS, and CIS compare model output against the same LLM-generated targets used in training (Section V-A).

how reviews work

0 comments
Cite this review

Pith. "Pith review of FashionM3: Multimodal, Multitask, and Multiround Fashion Assistant based on Unified Vision-Language Model." pith.science (2026). https://pith.science/paper/SXCM73B7

@misc{pith2026250417826,
  author       = {Pith},
  title        = {Pith review of: FashionM3: Multimodal, Multitask, and Multiround Fashion Assistant based on Unified Vision-Language Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SXCM73B7}},
  note         = {Machine review of arXiv:2504.17826}
}
read the original abstract

Fashion styling and personalized recommendations are pivotal in modern retail, contributing substantial economic value in the fashion industry. With the advent of vision-language models (VLM), new opportunities have emerged to enhance retailing through natural language and visual interactions. This work proposes FashionM3, a multimodal, multitask, and multiround fashion assistant, built upon a VLM fine-tuned for fashion-specific tasks. It helps users discover satisfying outfits by offering multiple capabilities including personalized recommendation, alternative suggestion, product image generation, and virtual try-on simulation. Fine-tuned on the novel FashionRec dataset, comprising 331,124 multimodal dialogue samples across basic, personalized, and alternative recommendation tasks, FashionM3 delivers contextually personalized suggestions with iterative refinement through multiround interactions. Quantitative and qualitative evaluations, alongside user studies, demonstrate FashionM3's superior performance in recommendation effectiveness and practical value as a fashion assistant.

Figures

Figures reproduced from arXiv: 2504.17826 by the authors.

Figure 1
Figure 1. Three examples from the FashionRec dataset illustrate its three task types. Blue and green text highlight key elements of the human and assistant [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Training pipeline of FashionVLM, showcasing the multimodal [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Overview of FashionM3’s architecture, orchestrating the flow of [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Qualitative results of generated images for personalized recommen [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Qualitative results of generated images for alternative recommendation [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Statistical results of user experience survey (n=12). The survey comprises ten questions on a 5-point Likert scale, ranging from “Strongly disagree” [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Interactive workflow of FashionM3 assisting Sophie in finding an outfit for a casual day out. The process unfolds in five steps (a-e). (a) Sophie’s [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: One example of basic recommendation task. The partial outfit is shown without a border, while the target items are indicated within a dashed border. [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: An example of the Personalized Recommendation task. The partial outfit is shown without a border, the target items are indicated within a dashed [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: One example of the Alternative Recommendation task. The overlapped items are shown without a border, while the changeable items are indicated [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. VOGUE: A Multimodal Dataset for Conversational Recommendation in Fashion

    cs.IR 2025-10 conditional novelty 6.0 of 10

    A multimodal fashion recommendation dataset with 60 organic dialogues and full rating distributions shows current MLLMs underperform human recommenders in preference calibration and generalization.

Reference graph

Works this paper leans on

64 extracted references · 48 canonical work pages · cited by 1 Pith paper

  1. [1]

    A comprehensive review of circular economy research in the textile and clothing industry,

    K. Saha, P. K. Dey, and V . Kumar, “A comprehensive review of circular economy research in the textile and clothing industry,” Journal of Cleaner Production, p. 141252, 2024

    work page 2024

  2. [2]

    Hybrid recommender systems: Survey and experiments,

    R. Burke, “Hybrid recommender systems: Survey and experiments,” User modeling and user-adapted interaction, vol. 12, pp. 331–370, 2002

    work page 2002

  3. [3]

    The evolution and future of retailing and retailing education,

    D. Grewal, S. Motyka, and M. Levy, “The evolution and future of retailing and retailing education,” Journal of Marketing Education , vol. 40, no. 1, pp. 85–93, 2018

    work page 2018

  4. [4]

    Knowledge management and fashion retail performance: the moderating role of product complexity,

    M. R. I. Rana and S.-y. Youn, “Knowledge management and fashion retail performance: the moderating role of product complexity,” Inter- national Journal of Retail & Distribution Management , vol. 52, no. 5, pp. 532–548, 2024

    work page 2024

  5. [5]

    Assembled or unassembled? different types of outfit coordination presentations in online fashion retailing,

    Y . Qu and E. Baek, “Assembled or unassembled? different types of outfit coordination presentations in online fashion retailing,” Fashion and Textiles, vol. 11, no. 1, p. 8, 2024

    work page 2024

  6. [6]

    What is the future of fashion retailing with generative ai? understanding consumer response through twitter data,

    S. Kwon and N. L. Kim, “What is the future of fashion retailing with generative ai? understanding consumer response through twitter data,” in International Textile and Apparel Association Annual Conference Proceedings, vol. 80, no. 1. Iowa State University Digital Press, 2024

    work page 2024

  7. [7]

    Modeling fashion compat- ibility with explanation by using bidirectional lstm,

    P. Kaicheng, Z. Xingxing, and W. K. Wong, “Modeling fashion compat- ibility with explanation by using bidirectional lstm,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 3894–3898

    work page 2021

  8. [8]

    Fashion forward with ai creations using gan,

    R. Ranjan, N. K. Walia, U. Anand, S. Puri et al. , “Fashion forward with ai creations using gan,” in 2024 Sixth International Conference on Computational Intelligence and Communication Technologies (CCICT) . IEEE, 2024, pp. 395–402

    work page 2024

Show all 64 references
  1. [9]

    Per- sonalized clothing recommendation fusing the 4-season color system and users’ biological characteristics,

    X. Su, J. Duan, J. Ren, Y . Li, M. Danner, M. R ¨atsch, and J. Peng, “Per- sonalized clothing recommendation fusing the 4-season color system and users’ biological characteristics,” Multimedia tools and applications, vol. 83, no. 5, pp. 12 597–12 625, 2024

    2024

  2. [10]

    The impact of servitization on perceived quality, purchase intentions and recommendation intentions in the ready- to-wear sector,

    P. Tosun and G. Tosun, “The impact of servitization on perceived quality, purchase intentions and recommendation intentions in the ready- to-wear sector,” Journal of Fashion Marketing and Management: An International Journal, vol. 28, no. 3, pp. 460–479, 2024

    2024

  3. [11]

    An intelligent recommendation system in e- commerce using ensemble learning,

    A. Shankar, P. Perumal, M. Subramanian, N. Ramu, D. Natesan, V . R. Kulkarni, and T. Stephan, “An intelligent recommendation system in e- commerce using ensemble learning,”Multimedia Tools and Applications, vol. 83, no. 16, pp. 48 521–48 537, 2024

    2024

  4. [12]

    Learning binary code for personalized fashion recommendation,

    Z. Lu, Y . Hu, Y . Jiang, Y . Chen, and B. Zeng, “Learning binary code for personalized fashion recommendation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2019, pp. 10 562–10 570

    2019

  5. [13]

    Learning similarity conditions without explicit supervision,

    R. Tan, M. I. Vasileva, K. Saenko, and B. A. Plummer, “Learning similarity conditions without explicit supervision,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2019, pp. 10 373–10 382

    2019

  6. [14]

    Fashion outfit complementary item retrieval,

    Y .-L. Lin, S. Tran, and L. S. Davis, “Fashion outfit complementary item retrieval,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 3311–3319

    2020

  7. [15]

    Learning type-aware embeddings for fashion compatibil- ity,

    M. I. Vasileva, B. A. Plummer, K. Dusad, S. Rajpal, R. Kumar, and D. Forsyth, “Learning type-aware embeddings for fashion compatibil- ity,” in Proceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 390–405

    2018

  8. [16]

    Language models are few-shot learners,

    T. B. Brown, “Language models are few-shot learners,” arXiv preprint arXiv:2005.14165, 2020

    2005 arXiv

  9. [17]

    Diffusion models beat gans on image synthesis,

    P. Dhariwal and A. Nichol, “Diffusion models beat gans on image synthesis,” Advances in neural information processing systems , vol. 34, pp. 8780–8794, 2021

    2021

  10. [18]

    Outfitgan: Learning compatible items for generative fashion outfits,

    M. Moosaei, Y . Lin, A. Akhazhanov, H. Chen, F. Wang, and H. Yang, “Outfitgan: Learning compatible items for generative fashion outfits,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 2273–2277

    2022

  11. [19]

    Diffusion models for generative outfit recommendation,

    Y . Xu, W. Wang, F. Feng, Y . Ma, J. Zhang, and X. He, “Diffusion models for generative outfit recommendation,” in Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , 2024, pp. 1350–1359

    2024

  12. [20]

    Generative rec- ommendation: Towards next-generation recommender paradigm,

    W. Wang, X. Lin, F. Feng, X. He, and T.-S. Chua, “Generative rec- ommendation: Towards next-generation recommender paradigm,” arXiv preprint arXiv:2304.03516, 2023

    2023 arXiv

  13. [21]

    Integrating domain knowledge into large lan- guage models for enhanced fashion recommendations,

    Z. Shi and S. Yang, “Integrating domain knowledge into large lan- guage models for enhanced fashion recommendations,” arXiv preprint arXiv:2502.15696, 2025

    2025 arXiv

  14. [22]

    Fashion recommendation systems, models and methods: A review,

    S. Chakraborty, M. S. Hoque, N. Rahman Jeem, M. C. Biswas, D. Bard- han, and E. Lobaton, “Fashion recommendation systems, models and methods: A review,” in Informatics, vol. 8, no. 3. MDPI, 2021, p. 49

    2021

  15. [23]

    A review of modern fashion recommender systems,

    Y . Deldjoo, F. Nazary, A. Ramisa, J. McAuley, G. Pellegrini, A. Bellogin, and T. D. Noia, “A review of modern fashion recommender systems,” ACM Comput. Surv. , vol. 56, no. 4, oct 2023. [Online]. Available: https://doi.org/10.1145/3624733

    2023 doi

  16. [24]

    Generative ai-based style recommendation using fashion item detection and classification,

    A. Kalinin, A. A. Jafari, E. Avots, C. Ozcinar, and G. Anbarjafari, “Generative ai-based style recommendation using fashion item detection and classification,” Signal, Image and Video Processing, vol. 18, no. 12, pp. 9179–9189, 2024

    2024

  17. [25]

    Fashioning consumer choices: recommendation, motivation, and purchase intention toward instagram commerce. a mediation analysis,

    D. Herzallah, F. Li ´ebana-Cabanillas, and F. Mu ˜noz-Leiva, “Fashioning consumer choices: recommendation, motivation, and purchase intention toward instagram commerce. a mediation analysis,” International Jour- nal of Fashion Design, Technology and Education , pp. 1–13, 2024

    2024

  18. [26]

    Image-based recommendations on styles and substitutes,

    J. McAuley, C. Targett, Q. Shi, and A. Van Den Hengel, “Image-based recommendations on styles and substitutes,” in Proceedings of the 38th international ACM SIGIR conference on research and development in information retrieval, 2015, pp. 43–52

    2015

  19. [27]

    Category-aware mul- timodal attention network for fashion compatibility modeling,

    P. Jing, K. Cui, W. Guan, L. Nie, and Y . Su, “Category-aware mul- timodal attention network for fashion compatibility modeling,” IEEE Transactions on Multimedia , vol. 25, pp. 9120–9131, 2023

    2023

  20. [28]

    Toward explainable fashion recommenda- tion,

    P. Tangseng and T. Okatani, “Toward explainable fashion recommenda- tion,” in Proceedings of the IEEE/CVF winter conference on applications of computer vision , 2020, pp. 2153–2162

    2020

  21. [29]

    Collaborative fashion recommendation: A functional tensor factorization approach,

    Y . Hu, X. Yi, and L. S. Davis, “Collaborative fashion recommendation: A functional tensor factorization approach,” in Proceedings of the 23rd ACM international conference on Multimedia , 2015, pp. 129–138

    2015

  22. [30]

    Personalized outfit recom- mendation with learnable anchors,

    Z. Lu, Y . Hu, Y . Chen, and B. Zeng, “Personalized outfit recom- mendation with learnable anchors,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2021, pp. 12 722–12 731

    2021

  23. [31]

    Pog: personalized outfit generation for fashion recommendation at alibaba ifashion,

    W. Chen, P. Huang, J. Xu, X. Guo, C. Guo, F. Sun, C. Li, A. Pfadler, H. Zhao, and B. Zhao, “Pog: personalized outfit generation for fashion recommendation at alibaba ifashion,” in Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining,...

    2019

  24. [32]

    Hierarchical fashion graph network for personalized outfit recommendation,

    X. Li, X. Wang, X. He, L. Chen, J. Xiao, and T.-S. Chua, “Hierarchical fashion graph network for personalized outfit recommendation,” in Pro- ceedings of the 43rd international ACM SIGIR conference on research and development in information retrieval , 2020, pp. 159–168

    2020

  25. [33]

    Learning visual body-shape-aware embeddings for fashion compatibility,

    K. Pang, X. Zou, and W. Wong, “Learning visual body-shape-aware embeddings for fashion compatibility,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 8056– 8065

    2024

  26. [34]

    What dress fits me best? fashion recommendation on the clothing style for personal body shape,

    S. C. Hidayati, C.-C. Hsu, Y .-T. Chang, K.-L. Hua, J. Fu, and W.- H. Cheng, “What dress fits me best? fashion recommendation on the clothing style for personal body shape,” in Proceedings of the 26th ACM international conference on Multimedia , 2018, pp. 438–446

    2018

  27. [35]

    Hairstyle suggestion using statisti- cal learning,

    W. Yang, M. Toyoura, and X. Mao, “Hairstyle suggestion using statisti- cal learning,” in Advances in Multimedia Modeling: 18th International Conference, MMM 2012, Klagenfurt, Austria, January 4-6, 2012. Pro- ceedings 18. Springer, 2012, pp. 277–287

    2012

  28. [36]

    Wow! you are so beautiful today!

    L. Liu, J. Xing, S. Liu, H. Xu, X. Zhou, and S. Yan, “Wow! you are so beautiful today!” ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM) , vol. 11, no. 1s, pp. 1–22, 2014

    2014

  29. [37]

    Show me the best outfit for a certain scene: A scene-aware fashion recommender system,

    T. Ye, L. Hu, Q. Zhang, Z. Y . Lai, U. Naseem, and D. D. Liu, “Show me the best outfit for a certain scene: A scene-aware fashion recommender system,” in Proceedings of the ACM Web Conference 2023 , 2023, pp. 1172–1180

    2023

  30. [38]

    A unified framework for outfit design and advice,

    A. Adewumi, A. Taiwo, S. Misra, R. Maskeliunas, R. Damasevicius, R. Ahuja, and F. Ayeni, “A unified framework for outfit design and advice,” in Data Management, Analytics and Innovation: Proceedings of ICDMAI 2019, Volume 2 . Springer, 2020, pp. 31–41

    2019

  31. [39]

    Explainable outfit recommendation with joint outfit matching and comment genera- tion,

    Y . Lin, P. Ren, Z. Chen, Z. Ren, J. Ma, and M. De Rijke, “Explainable outfit recommendation with joint outfit matching and comment genera- tion,” IEEE Transactions on Knowledge and Data Engineering , vol. 32, no. 8, pp. 1502–1516, 2019

    2019

  32. [40]

    Fashion outfit generation for e-commerce,

    E. M. Bettaney, S. R. Hardwick, O. Zisimopoulos, and B. P. Chamber- lain, “Fashion outfit generation for e-commerce,” in Machine Learning IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 12 and Knowledge Discovery in Databases. Applied Data Science and Demo Track: Eur...

    2020

  33. [41]

    Fashion compatibility modeling through a multi-modal try-on-guided scheme,

    X. Dong, J. Wu, X. Song, H. Dai, and L. Nie, “Fashion compatibility modeling through a multi-modal try-on-guided scheme,” in Proceedings of the 43rd international ACM SIGIR conference on research and development in information retrieval , 2020, pp. 771–780

    2020

  34. [42]

    Knowledge-guided compatibility mod- eling,

    X. Song, L. Nie, and Y . Wang, “Knowledge-guided compatibility mod- eling,” in Compatibility Modeling: Data and Knowledge Applications for Clothing Matching . Springer, 2020, pp. 27–43

    2020

  35. [43]

    Outfittransformer: Outfit representations for fashion rec- ommendation,

    R. Sarkar, N. Bodla, M. Vasileva, Y .-L. Lin, A. Beniwal, A. Lu, and G. Medioni, “Outfittransformer: Outfit representations for fashion rec- ommendation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 2263–2267

    2022

  36. [44]

    Leveraging multimodal features and item-level user feedback for bundle construc- tion,

    Y . Ma, X. Liu, Y . Wei, Z. Tao, X. Wang, and T.-S. Chua, “Leveraging multimodal features and item-level user feedback for bundle construc- tion,” in Proceedings of the 17th ACM International Conference on Web Search and Data Mining , 2024, pp. 510–519

    2024

  37. [45]

    AI-Yo: Embedding Psychosocial Aspects In the Fashion Stylist Chatbot Design,

    Z. Ye, M. Guo, J. Han, and J. Ma, “AI-Yo: Embedding Psychosocial Aspects In the Fashion Stylist Chatbot Design,” in Creativity and Cognition. Chicago IL USA: ACM, Jun. 2024, pp. 520–526

    2024

  38. [46]

    Multimodal conversational fashion recommendation with positive and negative natural-language feedback,

    Y . Wu, C. Macdonald, and I. Ounis, “Multimodal conversational fashion recommendation with positive and negative natural-language feedback,” in Proceedings of the 4th Conference on Conversational User Interfaces, 2022, pp. 1–10

    2022

  39. [47]

    Multi-modal dialog state tracking for interactive fashion recom- mendation,

    ——, “Multi-modal dialog state tracking for interactive fashion recom- mendation,” in Proceedings of the 16th ACM Conference on Recom- mender Systems, 2022, pp. 124–133

    2022

  40. [48]

    Show-o: One single transformer to unify multimodal understanding and generation,

    J. Xie, W. Mao, Z. Bai, D. J. Zhang, W. Wang, K. Q. Lin, Y . Gu, Z. Chen, Z. Yang, and M. Z. Shou, “Show-o: One single transformer to unify multimodal understanding and generation,” arXiv preprint arXiv:2408.12528, 2024

    2024 arXiv

  41. [49]

    Janusflow: Harmonizing autoregression and rectified flow for unified multimodal understanding and generation,

    Y . Ma, X. Liu, X. Chen, W. Liu, C. Wu, Z. Wu, Z. Pan, Z. Xie, H. Zhang, L. Zhao et al., “Janusflow: Harmonizing autoregression and rectified flow for unified multimodal understanding and generation,” arXiv preprint arXiv:2411.07975 , 2024

    2024 arXiv

  42. [50]

    Vitron: A unified pixel-level vision llm for understanding, generating, segmenting, edit- ing,

    H. Fei, S. Wu, H. Zhang, T.-S. Chua, and S. Yan, “Vitron: A unified pixel-level vision llm for understanding, generating, segmenting, edit- ing,” arXiv preprint arXiv:2412.19806 , 2024

    2024 arXiv

  43. [51]

    Unifashion: A unified vision-language model for multimodal fashion retrieval and generation,

    X. Zhao, Y . Zhang, W. Zhang, and X.-M. Wu, “Unifashion: A unified vision-language model for multimodal fashion retrieval and generation,” arXiv preprint arXiv:2408.11305 , 2024

    2024 arXiv

  44. [52]

    Fashionai: A hierarchical dataset for fashion understanding,

    X. Zou, X. Kong, W. Wong, C. Wang, Y . Liu, and Y . Cao, “Fashionai: A hierarchical dataset for fashion understanding,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops, 2019, pp. 0–0

    2019

  45. [53]

    Theme-matters: fashion compatibility learning via theme attention,

    J.-H. Lai, B. Wu, X. Wang, D. Zeng, T. Mei, and J. Liu, “Theme-matters: fashion compatibility learning via theme attention,” arXiv preprint arXiv:1912.06227, 2019

    1912 arXiv

  46. [54]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in International conference on machine learning . PmLR, 2021, pp. 8748–8763

    2021

  47. [55]

    Language model beats diffusion–tokenizer is key to visual generation,

    L. Yu, J. Lezama, N. B. Gundavarapu, L. Versari, K. Sohn, D. Minnen, Y . Cheng, V . Birodkar, A. Gupta, X. Gu et al. , “Language model beats diffusion–tokenizer is key to visual generation,” arXiv preprint arXiv:2310.05737, 2023

    arXiv

  48. [56]

    Textbooks are all you need,

    S. Gunasekar, Y . Zhang, J. Aneja, C. C. T. Mendes, A. Del Giorno, S. Gopi, M. Javaheripi, P. Kauffmann, G. de Rosa, O. Saarikivi et al. , “Textbooks are all you need,” arXiv preprint arXiv:2306.11644 , 2023

    2023 arXiv

  49. [57]

    Chainlit,

    L. AI, “Chainlit,” 2024. [Online]. Available: https://github.com/Chainlit/ chainlit

    2024

  50. [58]

    Llama-3.2-11B-Vision,

    Meta AI, “Llama-3.2-11B-Vision,” https://huggingface.co/meta-llama/ Llama-3.2-11B-Vision, 2024

    2024

  51. [59]

    Gpt-4o system card,

    A. Hurst, A. Lerer, A. P. Goucher, A. Perelman, A. Ramesh, A. Clark, A. Ostrow, A. Welihinda, A. Hayes, A. Radford et al., “Gpt-4o system card,” arXiv preprint arXiv:2410.21276 , 2024

    2024 arXiv

  52. [60]

    Medec: A benchmark for medical error detection and correction in clinical notes,

    A. B. Abacha, W.-w. Yim, Y . Fu, Z. Sun, M. Yetisgen, F. Xia, and T. Lin, “Medec: A benchmark for medical error detection and correction in clinical notes,” arXiv preprint arXiv:2412.19260 , 2024

    2024 arXiv

  53. [61]

    Sentence-bert: Sentence embeddings using siamese bert-networks,

    N. Reimers and I. Gurevych, “Sentence-bert: Sentence embeddings using siamese bert-networks,” arXiv preprint arXiv:1908.10084 , 2019. APPENDIX A. System Prompts for Generating Dialogues

    1908 arXiv

  54. [62]

    messages=[{‘role’: ‘system’, ‘content’: ‘‘‘ As a fashion expert, generate a user−system conversation for training a fashion stylist model

    For Basic Recommendation. messages=[{‘role’: ‘system’, ‘content’: ‘‘‘ As a fashion expert, generate a user−system conversation for training a fashion stylist model. Your goal is to create natural, concise, and relevant dialogues based on the provided partial outfit and target ...

  55. [63]

    messages=[{‘role’: ‘system’, ‘content’: ‘‘‘ Create a user−system conversation for training a personalized fashion stylist model

    For Personalized Recommendation. messages=[{‘role’: ‘system’, ‘content’: ‘‘‘ Create a user−system conversation for training a personalized fashion stylist model. Focus on developing natural, concise, and relevant dialogues using the provided partial outfit, target items, and u...

  56. [64]

    UN GAYVOE MVEREI LELS

    For Alternative Recommendation. messages=[{‘role’: ‘system’, ‘content’: ‘‘‘ As a fashion expert, generate a user−system conversation for training a fashion stylist model. Your goal is to create a natural, concise, and relevant dialogue based on a given outfit and a specified c...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.