Pith. sign in

REVIEW 2 major objections 4 minor 24 references

Grounding Multimodal Large Language Models with Quantitative Skin Attributes: A Retrieval Study

T0 review · 2 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Fine-tuning an MLLM to predict quantitative skin attributes grounds its embeddings in those attributes, and the resulting image-text embeddings retrieve images that match both appearance and the specified attribute.

desk verdict Honest, useful study: LoRA-tuning Qwen2-VL on 16 SLICE-3D attributes yields held-out R2 0.71–0.96, and the retrieval demo is a genuine extension of composed retrieval, but the retrieval evaluation uses the training set as the gallery, which makes the case for embedding-space grounding weaker than claimed. read the letter →

arxiv 2508.20188 v1 pith:SBWGJNCP submitted 2025-08-27 cs.CV cs.LG

Max Torop , Masih Eskandar , Nicholas Kurtansky , Jinyang Liu , Jochen Weber , Octavia Camps , Veronica Rotemberg , Jennifer Dy
show 1 more author
Kivanc Kose
This is my paper · ORCID
classification cs.CVcs.LG
keywords multimodallargelanguagemodelsquantitativeskinattributescontent-basedimageretrievalattributegroundingLoRAfine-tuninginterpretabilitySLICE-3Dnumericalreasoning
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

The paper claims that fine-tuning a multimodal large language model (MLLM) to predict numerical skin attributes—such as lesion area, border irregularity, and color contrast—can ground the model's embedding space in those attributes. It demonstrates this through a retrieval case study on a dataset of about 400,000 skin lesion images: after LoRA fine-tuning Qwen 2 VL on image-question-answer tuples, the embedding of the final token conditioned on an attribute question retrieves images that match the query both visually and in that attribute. Test-set R² values for predicting the 16 attributes range from 0.71 to 0.96. If the grounding holds, it offers a path toward more interpretable, interactive dermatology AI that can justify its decisions in terms of measurable lesion properties.

What carries the argument

The load-bearing object is the attribute-conditioned embedding h(im,a)(I) = f_{-1}([g(I), ψ(Q(a))]), the final decoder token's feature when the input is the image followed by the attribute question. Because the final token attends to every earlier token, this single vector is meant to fuse the whole image with the meaning of the requested attribute. LoRA fine-tuning on numeric answer tokens is what forces that fused representation to carry attribute information. The image-only baseline h(im)(I) averages all image-token outputs and captures general visual similarity, while a hierarchical search first narrows with h(im) and then re-ranks with h(im,a), trading compute for per-attribute precisio

What would settle it

Repeat the retrieval experiment with the image database restricted to patient-disjoint held-out images that were never used in LoRA fine-tuning. If the attribute-conditioned embeddings no longer beat the untuned Qwen and medical image-text baselines on attribute match, the reported grounding is an artifact of training-set familiarity rather than a general property of the embedding space.

Watch

Extended reading notes

Core claim

The paper's central claim is that an MLLM's embedding space can be grounded in quantitative skin attributes by fine-tuning the model to predict those attributes from images. Using Qwen 2 VL 7B, the authors apply LoRA to jointly fine-tune the vision encoder and decoder on 1.4 million (image, attribute question, numeric answer) tuples sampled from the SLICE-3D dataset. They then use the decoder's last-layer final-token embedding, computed from the image plus an attribute-specific question, as an attribute-conditioned image embedding. In retrieval over the training set, this embedding returns top-5 images whose attribute values are closer to the query's than those returned by image-only embeddi

Load-bearing premise

The argument assumes retrieval over the training set is a fair test, even though the fine-tuned model was optimized on those exact images while the untuned baselines were not.

Editorial extensions

If this is right

  • Attribute-conditioned retrieval consistently beats image-only retrieval on the 16 tested attributes, so an MLLM can serve as an attribute-aware image search engine after question-answer fine-tuning.
  • The hierarchical variant, which searches 200 initial images before re-ranking, preserves retrieval quality while avoiding one database per attribute, making the approach practical at SLICE-3D scale.
  • Because the text prompt can hold several attribute questions at once, multi-attribute retrieval—finding images matching a query on area and border jaggedness simultaneously—is a direct extension of the same embedding recipe.
  • The reported test-set R² values between 0.71 and 0.96 across all 16 attributes support the claim that the model has learned the numerical concepts on held-out patients.

Reading between the lines

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

  • Because the retrieval database is the training set itself, part of the fine-tuned model's retrieval advantage could come from having seen those exact images during LoRA training; a held-out retrieval database would cleanly separate attribute grounding from training-set familiarity.
  • The paper uses a single fixed question per attribute, so retrieval quality may shift with question wording; the reported grounding is for these particular prompts rather than for attribute concepts in general.
  • High attribute-prediction accuracy does not by itself guarantee that the embedding space is ordered by attribute value; the retrieval percentiles are the direct geometric evidence, and a ranking or linear-probe test on the embeddings would make the geometric claim more explicit.
  • If the grounding transfers to other image domains with numeric ground-truth attributes, the same question-conditioned final-token recipe could make retrieval systems in radiology or pathology explainable by the quantitative features they match on.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper explores grounding multimodal large language models (MLLMs) in quantitative skin attributes. It fine-tunes Qwen2-VL 7B with LoRA on the SLICE-3D dataset to predict 16 numerical attributes (e.g., lesion area, border jaggedness) from images. The authors propose two embedding functions: an image-only embedding h(im) (average of image-token features) and an attribute-conditioned image-text embedding h(im,a) (final-token feature after appending an attribute-specific question). They evaluate attribute prediction on held-out patients (R2 from 0.71 to 0.96) and present a retrieval case study in which test-set queries retrieve images from the training set, comparing fine-tuned embeddings against vanilla Qwen and MONET baselines. The conclusion claims that the retrieval results provide evidence that MLLM embedding spaces can be grounded in quantitative attributes.

Significance. If the retrieval evidence were sound, the paper would make a useful contribution: it would show that an MLLM's embedding space can be steered by attribute questions, which is relevant for interpretable and interactive dermatology AI. Strengths include the use of a large real-world dataset (SLICE-3D), a held-out patient split for the R2 evaluation, a simple and reproducible fine-tuning setup, and comparison against a medical VLM. However, the central retrieval demonstration is weakened by a training-set gallery and the absence of significance testing. The paper is better viewed as a promising case study than as a fully established demonstration of attribute grounding in embedding space.

major comments (2)
  1. [§3.3, §4.1, §4.2 (Fig. 4)] The retrieval database is explicitly set to I' = Itr (§3.3), and Itr is exactly the set of images used for LoRA fine-tuning (§4.1). The fine-tuned model therefore had direct supervised access to every gallery image and its attribute values during training, whereas the vanilla Qwen and MONET baselines did not. The consistent ordering in Fig. 4 may reflect training-set familiarity rather than a generally attribute-grounded embedding geometry. This is load-bearing because the conclusion in §5 bases the grounding claim on the retrieval results. Please re-run the retrieval with a held-out gallery (e.g., query from a subset of Ite and retrieve from the remaining Ite, or split Itr into a fine-tuning set and a disjoint gallery), and/or add a baseline that is fine-tuned on Itr without attribute conditioning. If the advantage persists in that setting, the conclusion will be much better supported.
  2. [§4.2, Fig. 4] The narrative claims that the fine-tuned image-only embeddings are 'significantly more effective' than the baselines and that image-text embeddings are 'most effective', but Fig. 4 reports only boxplots of percentile ranks. No confidence intervals, standard errors, or paired significance tests are provided. Since all methods are evaluated on the same query images, paired bootstrap or Wilcoxon signed-rank tests are straightforward. Please add such statistics for at least the key comparisons (fine-tuned image-only vs. MONET, and image-text vs. image-only). Without them, the stated ordering is not quantitatively established.
minor comments (4)
  1. [§3.3, Eq. (3)] The notation h(im,a)(I) = f^{-1}(...) is incorrect as written: f is defined in §3.1 as a feature extractor (a sequence-to-sequence map), not an invertible function. The intended object appears to be the last row of f([g(I), ψ(Q(a))]), i.e., f_{N_I+|Q(a)|}(...). Please fix the definition, as the current form is not reproducible.
  2. [§4.2, Fig. 4] Please define 'Image Untuned' explicitly in the text or figure caption, and state whether it uses the same image-token averaging as h(im) on the vanilla Qwen model. The current label is ambiguous.
  3. [§4.2, 'Quantitative Prediction'] The R2 computation is described only as 'predict all 16 attributes and compare to ground truth.' For reproducibility, please specify the decoding strategy (e.g., greedy sampling), how the numeric answer is parsed from generated tokens, and whether R2 is computed per value or per attribute after any calibration. Reporting confidence intervals for the R2 values would also be useful.
  4. [§3.3, Hierarchical Retrieval] The choice of b = 200 is introduced without justification or sensitivity analysis. Since the first-stage image-only embedding h(im) is itself trained with attribute supervision, the hierarchical method is not attribute-neutral; the paper should acknowledge this when interpreting why 'the hierarchical modification does not lower quality.'

Circularity Check

1 steps flagged · score 4.0 of 10

Retrieval evidence is partly self-referential: the gallery is the training set and the attribute-conditioned embeddings are trained on the same labels, though held-out R2 provides independent support.

  1. fitted input called prediction [Section 3.3 (Quantitative Image Retrieval) and Section 4.2 (Quantitative Retrieval)]
    "In this work we set I′ = Itr. ... For each method we retrieve the top-5 images from Itr for each Iq ∈ Ite."

    The model is fine-tuned on Itr (Secs. 3.2, 4.1) to predict the same 16 attributes used to score retrieval, and the image-text embedding h(im,a) (Eq. 3) is the final token representation optimized for that prediction. Using Itr as the gallery means the retrieval ranks images whose embeddings were directly fit to the attribute labels. The improvement over untuned baselines therefore reflects the model's fitted knowledge of the gallery, not an independent test of embedding-space grounding. The held-out R2 on Ite (Fig. 2) is not circular, but the retrieval case study—the concrete embedding-space evidence—is partly self-referential.

full rationale

The paper's held-out attribute prediction (R2 0.71–0.96 on patient-disjoint Ite) is a legitimate, non-circular evaluation: test patients were excluded from fine-tuning, so the model must generalize to unseen images. The retrieval study, however, uses Itr as the image database for all methods, and the fine-tuned model was optimized on exactly those images to predict the same attribute values that define retrieval quality. The image-text embedding h(im,a) is the penultimate representation of the attribute-prediction head, so ranking gallery images by this embedding is closely tied to the training objective. This makes the retrieval comparison against untuned baselines partly a measure of training-set familiarity rather than of a generalizable attribute-grounded embedding space. There is no load-bearing self-citation: references [8] and [10] are by overlapping authors but are used as dataset/motivation sources, not as justification for the retrieval claim. No uniqueness theorem is invoked. Overall, the central claim retains independent support from the held-out R2, so the circularity is partial and confined to the retrieval demonstration.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The paper's central claim depends on the reliability of SLICE-3D attributes, the validity of the held-out split, and the assumption that decoder token embeddings carry the needed similarity information. The main free parameters are training and retrieval hyperparameters, none of which are rigorously ablated. No new physical entities are introduced.

free parameters (4)
  • LoRA rank = 8
    Hand-chosen rank for low-rank adaptation; affects model capacity and the degree of attribute grounding. No sensitivity analysis is reported.
  • Number of sampled attributes per image W = 5
    The paper samples W=5 attributes per training image to form training tuples. This choice affects training data size and may influence which attributes are best learned.
  • Hierarchical retrieval first-stage size b = 200
    The hierarchical retrieval first retrieves 200 images before attribute-specific search. This is a hand-picked tradeoff between speed and coverage; no sensitivity analysis is given.
  • Training hyperparameters = LR 1e-4, batch 16, 1 epoch, 4 GPUs
    Learning rate, batch size, and epoch count are chosen without ablations. They affect the resulting embedding quality and retrieval performance.
assumptions (5)
  • domain assumption SLICE-3D attribute values are reliable ground truth.
    Section 4.1 uses these attributes as supervised labels for training and as ground truth for R2 and retrieval evaluation.
  • domain assumption Patient-stratified splitting prevents identity leakage between training and test sets.
    Section 4.1 states the split is 'stratifying by patients to make sure there is no overlap', which is necessary for the test R2 values to be meaningful.
  • domain assumption R2 on held-out images is a valid measure of attribute grounding.
    Section 4.2 equates high prediction R2 with the model having 'effectively learned to predict the attributes from images'.
  • domain assumption Mean image-token embeddings and final question-token embeddings capture visual and attribute-specific similarity.
    Section 3.3 defines h(im) and h(im,a) using these decoder features and assumes they are suitable similarity substrates for retrieval.
  • ad hoc to paper Attribute question templates Q(a) are valid conditioning signals.
    The questions are designed by the authors and used to specialize embeddings; no ablation shows that the exact wording matters, but the method depends on it.
invented entities (1)
  • Attribute-conditioned embedding functions h(im) and h(im,a)
    purpose: They are embedding constructions used for retrieval; they are methodological constructs rather than physical entities.
    These are not newly postulated physical or conceptual entities, just feature extraction procedures defined in Section 3.3, so no independent falsifiable handle is expected.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Grounding Multimodal Large Language Models with Quantitative Skin Attributes: A Retrieval Study." pith.science (2026). https://pith.science/paper/SBWGJNCP

@misc{pith2026250820188,
  author       = {Pith},
  title        = {Pith review of: Grounding Multimodal Large Language Models with Quantitative Skin Attributes: A Retrieval Study},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SBWGJNCP}},
  note         = {Machine review of arXiv:2508.20188}
}
read the original abstract

Artificial Intelligence models have demonstrated significant success in diagnosing skin diseases, including cancer, showing the potential to assist clinicians in their analysis. However, the interpretability of model predictions must be significantly improved before they can be used in practice. To this end, we explore the combination of two promising approaches: Multimodal Large Language Models (MLLMs) and quantitative attribute usage. MLLMs offer a potential avenue for increased interpretability, providing reasoning for diagnosis in natural language through an interactive format. Separately, a number of quantitative attributes that are related to lesion appearance (e.g., lesion area) have recently been found predictive of malignancy with high accuracy. Predictions grounded as a function of such concepts have the potential for improved interpretability. We provide evidence that MLLM embedding spaces can be grounded in such attributes, through fine-tuning to predict their values from images. Concretely, we evaluate this grounding in the embedding space through an attribute-specific content-based image retrieval case study using the SLICE-3D dataset.

Figures

Figures reproduced from arXiv: 2508.20188 by the authors.

Figure 1
Figure 1. Exemplar illustration of the image-only and area-specific image-text embedding functions. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Test set attribute prediction [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Qualitative retrieval results. The left shows query images [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Retrieval results for each attribute. We visualize the distribution of the percentile of the [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 15 canonical work pages

  1. [1]

    Medimageinsight: An open-source embedding model for general domain medical imaging

    Noel CF Codella, Ying Jin, Shrey Jain, Yu Gu, Ho Hin Lee, Asma Ben Abacha, Alberto Santamaria-Pang, Will Guyman, Naiteek Sangani, Sheng Zhang, et al. Medimageinsight: An open-source embedding model for general domain medical imaging. arXiv preprint arXiv:2410.06542, 2024

    arXiv 2024

  2. [2]

    Validation of artificial intelligence prediction models for skin cancer diagnosis using dermoscopy images: the 2019 international skin imaging collaboration grand challenge

    Marc Combalia, Noel Codella, Veronica Rotemberg, Cristina Carrera, Stephen Dusza, David Gutman, Brian Helba, Harald Kittler, Nicholas R Kurtansky, Konstantinos Liopyris, et al. Validation of artificial intelligence prediction models for skin cancer diagnosis using dermoscopy images: the 2019 international skin imaging collaboration grand challenge. The La...

    work page 2019

  3. [3]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, G Heigold, S Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. In International Conference on Learning Representations, 2020

    2020

  4. [4]

    Analysis of trends in geographic distribution of us dermatology workforce density

    Alex M Glazer and Darrell S Rigel. Analysis of trends in geographic distribution of us dermatology workforce density. JAMA dermatology, 153(5):472–473, 2017

    work page 2017

  5. [5]

    LoRA: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. In International Conference on Learning Representations, 2022. URL https://openreview. net/forum?id=nZeVKeeFYf9

    2022

  6. [6]

    Transparent medical image ai via an image–text foundation model grounded in medical literature

    Chanwoo Kim, Soham U Gadgil, Alex J DeGrave, Jesutofunmi A Omiye, Zhuo Ran Cai, Roxana Daneshjou, and Su-In Lee. Transparent medical image ai via an image–text foundation model grounded in medical literature. Nature Medicine, 30(4):1154–1165, 2024

    work page 2024

  7. [7]

    Human-ai interaction in skin cancer diagnosis: a systematic review and meta-analysis

    Isabelle Krakowski, Jiyeong Kim, Zhuo Ran Cai, Roxana Daneshjou, Jan Lapins, Hanna Eriksson, Anastasia Lykou, and Eleni Linos. Human-ai interaction in skin cancer diagnosis: a systematic review and meta-analysis. NPJ Digital Medicine, 7(1):78, 2024

    work page 2024

  8. [8]

    The slice-3d dataset: 400,000 skin lesion image crops extracted from 3d tbp for skin cancer detection

    Nicholas R Kurtansky, Brian M D’Alessandro, Maura C Gillis, Brigid Betz-Stablein, Sara E Cerminara, Rafael Garcia, Marcela Alves Girundi, Elisabeth Victoria Goessinger, Philippe Gottfrois, Pascale Guitera, et al. The slice-3d dataset: 400,000 skin lesion image crops extracted from 3d tbp for skin cancer detection. Scientific Data, 11(1):884, 2024

    work page 2024

Show all 24 references
  1. [9]

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In International conference on machine learning, pages 19730–19742. PMLR, 2023

    2023

  2. [10]

    3d whole-body skin imaging for automated melanoma detection

    MA Marchetti, ZH Nazir, JK Nanda, SW Dusza, BM D’Alessandro, J DeFazio, AC Halpern, VM Rotemberg, and AA Marghoob. 3d whole-body skin imaging for automated melanoma detection. Journal of the European Academy of Dermatology and Venereology, 37(5):945–950, 2023

    2023

  3. [11]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, pa...

    2021

  4. [12]

    Steering llama 2 via contrastive activation addition

    Nina Rimsky, Nick Gabrieli, Julian Schulz, Meg Tong, Evan Hubinger, and Alexander Turner. Steering llama 2 via contrastive activation addition. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15504–15522, 2024

    2024

  5. [13]

    Grad-cam: Visual explanations from deep networks via gradient-based localization

    Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad-cam: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision, pages 618–626, 2017

    2017

  6. [14]

    Socioeconomic and geographic barriers to dermatology care in urban and rural us populations

    Toral Vaidya, Lindsey Zubritsky, Ali Alikhan, and Anne Housholder. Socioeconomic and geographic barriers to dermatology care in urban and rural us populations. Journal of the American Academy of Dermatology, 78(2):406–408, 2018

    2018

  7. [15]

    Composing text and image for image retrieval-an empirical odyssey

    Nam V o, Lu Jiang, Chen Sun, Kevin Murphy, Li-Jia Li, Li Fei-Fei, and James Hays. Composing text and image for image retrieval-an empirical odyssey. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6439–6448, 2019

    2019

  8. [16]

    Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution

    Peng Wang, Shuai Bai, Sinan Tan, Shijie Wang, Zhihao Fan, Jinze Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, et al. Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution. arXiv preprint arXiv:2409.12191, 2024

    2024 arXiv

  9. [17]

    Visnumbench: Evaluating number sense of multimodal large language models

    Tengjin Weng, Jingyi Wang, Wenhao Jiang, and Zhong Ming. Visnumbench: Evaluating number sense of multimodal large language models. arXiv preprint arXiv:2503.14939, 2025

    2025 arXiv

  10. [18]

    Derm1m: A million-scale vision-language dataset aligned with clinical ontology knowledge for dermatology

    Siyuan Yan, Ming Hu, Yiwen Jiang, Xieji Li, Hao Fei, Philipp Tschandl, Harald Kittler, and Zongyuan Ge. Derm1m: A million-scale vision-language dataset aligned with clinical ontology knowledge for dermatology. arXiv preprint arXiv:2503.14911, 2025

    2025 arXiv

  11. [19]

    A multimodal vision foundation model for clinical dermatology

    Siyuan Yan, Zhen Yu, Clare Primiero, Cristina Vico-Alonso, Zhonghua Wang, Litao Yang, Philipp Tschandl, Ming Hu, Lie Ju, Gin Tan, et al. A multimodal vision foundation model for clinical dermatology. Nature Medicine, pages 1–12, 2025

    2025

  12. [20]

    Qwen2 technical report

    An Yang, Baosong Yang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Zhou, Chengpeng Li, Chengyuan Li, Dayiheng Liu, Fei Huang, et al. Qwen2 technical report. arXiv preprint arXiv:2407.10671, 2024

    2024 arXiv

  13. [21]

    Visrag: Vision-based retrieval-augmented generation on multi-modality documents

    Shi Yu, Chaoyue Tang, Bokai Xu, Junbo Cui, Junhao Ran, Yukun Yan, Zhenghao Liu, Shuo Wang, Xu Han, Zhiyuan Liu, and Maosong Sun. Visrag: Vision-based retrieval-augmented generation on multi-modality documents. In The Thirteenth International Conference on Learning Representati...

    2025

  14. [22]

    Mm-skin: Enhancing dermatology vision-language model with an image-text dataset derived from textbooks

    Wenqi Zeng, Yuqi Sun, Chenxi Ma, Weimin Tan, and Bo Yan. Mm-skin: Enhancing dermatology vision-language model with an image-text dataset derived from textbooks. arXiv preprint arXiv:2505.06152, 2025

    2025 arXiv

  15. [23]

    Revisiting the trustworthiness of saliency methods in radiology ai

    Jiajin Zhang, Hanqing Chao, Giridhar Dasegowda, Ge Wang, Mannudeep K Kalra, and Pingkun Yan. Revisiting the trustworthiness of saliency methods in radiology ai. Radiology: Artificial Intelligence, 6(1):e220221, 2023

    2023

  16. [24]

    Pre-trained multimodal large language model enhances dermatological diagnosis using skingpt-4

    Juexiao Zhou, Xiaonan He, Liyuan Sun, Jiannan Xu, Xiuying Chen, Yuetan Chu, Longxi Zhou, Xingyu Liao, Bin Zhang, Shawn Afvari, et al. Pre-trained multimodal large language model enhances dermatological diagnosis using skingpt-4. Nature Communications, 15(1):5649, 2024. 9

    2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.