Pith. sign in

REVIEW 4 major objections 8 minor 41 references

Fair-MoE: Fairness-Oriented Mixture of Experts in Vision-Language Models

T0 review · 4 major / 8 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read A mixture-of-experts layer plus a variance-matching fairness loss improves both accuracy and fairness of medical vision-language models on a glaucoma benchmark.

desk verdict The central claim — improvements across all four attributes — is contradicted by the paper's own Table 1; the MoE-for-fairness idea is plausible but this version overclaims and has reproducibility gaps. read the letter →

arxiv 2502.06094 v1 pith:3CQDEB2B submitted 2025-02-10 cs.CV

Peiran Wang , Linjie Tong , Jiaxiang Liu , Zuozhu Liu This is my paper · ORCID
classification cs.CV
keywords fairnessvision-languagemodelsmixtureofexpertsmedicalimageanalysiscontrastivelearningSinkhorndistanceglaucomademographicparity
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 proposes Fair-MoE, a vision-language model for medical image diagnosis that aims to improve diagnostic accuracy and demographic fairness at the same time. Tested on a glaucoma image-and-text dataset, the model reports better area under the ROC curve and better fairness metrics—demographic parity difference, equal opportunity difference, and equity-scaled AUC—than a standard contrastive vision-language model and a prior fairness-oriented version of it, across all four protected attributes: race, gender, ethnicity, and language. The authors claim the gains come from two design choices: replacing the last-layer MLPs with a sparse mixture of experts that can filter out biased patch embeddings, and adding a fairness loss that matches both the mean and the variance of gating weights between protected groups. If the reported results hold, the approach would give medical VLMs a way to reduce demographic disparities without adding parameters or sacrificing accuracy.

What carries the argument

The machinery has two parts. The Fairness-Oriented Mixture of Experts (FO-MoE) replaces the MLP in the final attention block with an embedding-based MoE and places a second feature-based MoE after the encoder; each uses multiple MLP experts and a gating network, keeps only the top-k expert weights per input, and, in the embedding-based layer, caps how many patch embeddings each expert may process (a capacity filter) so that supposedly biased, task-irrelevant patch embeddings are dropped. The Fairness-Oriented Loss (FOL) adds to the standard Sinkhorn distance loss four terms that each compute, for one of the four MoE modules, the squared difference between the variance of expert gate weights over the whole dataset and the variance over a protected group, summed over groups and experts. FOL is thus a load-balancing loss that also acts as a fairness regularizer by making the dispersion of expert usage similar across demographic groups.

What would settle it

Log the per-group variance of gate weights alongside DPD and EOD throughout training of Fair-MoE; if the fairness metrics improve without the variance-difference term shrinking, or the variance differences shrink while DPD and EOD stay flat, the claim that gate-weight dispersion drives fairness is refuted, and the same logging can reveal whether the capacity filter's dropped patches correlate with protected attributes.

Watch

Extended reading notes

Core claim

The central claim is that a fairness-oriented mixture-of-experts layer plus a loss that penalizes differences in the dispersion of gating weights across protected groups simultaneously improves effectiveness and fairness in medical vision-language models. Concretely, the paper reports that Fair-MoE with a ViT-L/14 encoder reaches an equity-scaled AUC of 72.53 on the race attribute compared with 67.53 for the FairCLIP baseline, with demographic parity difference dropping from 16.01 to 2.63, while overall AUC rises from 71.57 to 73.93; similar patterns appear for gender, ethnicity, and language. While the baseline FairCLIP regularizes only the Sinkhorn distance between group distributions, Fair-MoE adds variance-of-gate-weight matching, and the ablations show that removing this variance term degrades both accuracy and fairness. The paper also shows that both the embedding-level and feature-level expert layers contribute to the reported trade-off.

Load-bearing premise

The load-bearing premise is that the variance of expert-gate weights across a protected group is a meaningful stand-in for the demographic disparities that the fairness metrics measure: if evening out gate-weight variance does not actually track or cause equal outcome rates, the proposed loss is optimizing an irrelevant quantity and the reported fairness improvements would not be attributable to the mechanism claimed.

Editorial extensions

If this is right

  • If the reported results are correct, fairness gains in medical VLMs do not require retraining from scratch or adding parameters: the paper states parameter counts stay around 200M for the smaller encoder and 500M for the larger one.
  • The ablation results imply that matching the variance of gate weights between groups is load-bearing; removing it measurably lowers both AUC and equity-scaled AUC.
  • Because both the image-branch and text-branch MoE ablations hurt performance when removed, the paper's account predicts that debiasing a multimodal medical VLM requires intervening in both modalities, not only on the visual side.
  • The capacity filter is presented as the mechanism that discards biased patch embeddings; if that is true, the same sparse-expert recipe could be transferred to other contrastive VLMs to remove spurious correlates before they enter the task representation.

Reading between the lines

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

  • The paper does not show that gate-weight variance tracks the demographic parity or equal-opportunity disparities it optimizes; a natural next step would be to measure that correlation during training and test whether a control loss of equal strength on randomly shuffled group labels reproduces the fairness gains.
  • A testable extension is to apply the same FO-MoE-plus-FOL design to other medical imaging tasks and check whether the fairness gains generalize beyond glaucoma or depend on the particular group sizes in the benchmark dataset.
  • Because the capacity filter is not analyzed in terms of which patches are dropped, the stated bias-filtering mechanism could be checked by comparing the distribution of dropped patch locations with protected-attribute content; if the dropped patches do not correlate with protected content, the effect may come from regularization rather than targeted filtering.
  • If variance matching is the operative ingredient, connecting FOL to a second-moment moment-matching or optimal-transport constraint could give a principled account of when the loss helps and when it might conflict with task accuracy.
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

4 major / 8 minor

Summary. The paper proposes Fair-MoE, a vision-language model for medical image analysis that replaces the final MLP layers in CLIP with a mixture-of-experts architecture (FO-MoE) and adds a fairness-oriented loss (FOL) that combines Sinkhorn distance with a penalty on the variance difference of gate weights between protected attribute groups. The authors claim that Fair-MoE improves both accuracy (AUC) and fairness (DPD, EOD, ES-AUC) over CLIP and FairCLIP across all four protected attributes (race, gender, ethnicity, language) on the Harvard-FairVLMed dataset. The paper includes a main results table, ablation studies for FO-MoE and FOL, and component-wise ablations for the four variance-based loss terms.

Significance. If the empirical claims held, Fair-MoE would offer a simple and general recipe for debiasing medical VLMs: route patches and features through a sparse mixture of experts and regularize the gate-weight variance across protected groups. The idea of using gate-weight dispersion as a fairness signal is interesting, and the ablation structure is comprehensive in intent. However, the headline claim is contradicted by the paper's own main results table, and the fairness metric definitions are flawed. As a result, the significance of the contribution as stated is not established; the paper needs a corrected set of experiments and metric definitions before its claims can be assessed.

major comments (4)
  1. [§4.2, Table 1] The abstract claims 'improvements in both fairness and accuracy across all four attributes,' but Table 1 contradicts this. For ViT-L14 under Language, Fair-MoE/l14 reports DPD 15.67±2.99 and EOD 23.63±14.40 versus FairCLIP/l14's DPD 8.21±1.99 and EOD 11.00±1.25; for Gender, Fair-MoE/l14 reports DPD 2.94±1.60 and EOD 7.33±2.55 versus FairCLIP/l14's DPD 2.11±1.81 and EOD 5.24±1.46. Fair-MoE is therefore worse on both fairness metrics for two of four attributes in the larger architecture. The sentence in §4.2 stating that 'results of DPD show that Fair-MoE achieves better fairness than baselines in all attributes' is false for Language under ViT-L14, and the same paragraph already concedes that EOD gains hold only for race, gender, and ethnicity. The paper's central empirical claim is internally contradicted by its own numbers.
  2. [§4.1, metric definitions] The Demographic Parity Difference is defined as DPD_s = |max_a P(ŷ=1|G=a, y=1) − min_b P(ŷ=1|G=b, y=1)|, which conditions on the true label y=1. This is not demographic parity; it is the true-positive-rate gap, which is the first term of the Equalized Odds Difference defined immediately afterwards. Thus DPD and EOD are not independent fairness metrics as reported. All fairness claims in the abstract, §4.2, and the conclusions that refer to 'demographic parity' are therefore unsubstantiated, and the numerical DPD values would need to be recomputed under the correct conditional probability P(ŷ=1|G=a).
  3. [§4.3, Tables 3–6] The ablation study does not support the claim that FOL is indispensable for fairness. Table 3 shows several cases where removing FOL improves DPD or EOD: FairMoE/b16 race DPD worsens from 3.19 (without FOL) to 7.25 (with FOL); FairMoE/b16 ethnicity DPD worsens from 6.94 to 8.52; FairMoE/l14 language DPD worsens from 10.43 to 15.67 and EOD from 9.65 to 23.63. The text states that removing FOL makes 'metrics that measure effectiveness and fairness deteriorate significantly,' which is not uniformly true. Moreover, Tables 4–6, used to argue the necessity of the four variance-based loss components, report only ES-AUC and never report DPD or EOD, so they cannot establish that these components improve outcome-level fairness metrics.
  4. [§3.1–3.2] The proposed mechanism is under-validated. Section 3.1 asserts that the capacity parameter C 'filter[s] out biased path embeddings,' but no analysis is provided of which patches are actually dropped or whether the top-k capacity preferentially discards biased rather than arbitrary low-weight embeddings. Section 3.2 treats the variance of gate weights as a proxy for fairness, but the paper gives no empirical or theoretical connection between gate-weight variance and the outcome-level metrics DPD/EOD. Because the FOL losses are computed on gate weights rather than on predictions, the claimed fairness gains are not attributable to the proposed mechanism without such a link.
minor comments (8)
  1. [§3.1] The notation 'T opc' and 'T opr' should be typeset as proper functions, and the capacity formula α = C(N+1)k1/M1 needs a clearer derivation; as printed, the expression is difficult to parse.
  2. [§4.1] The experimental setup does not report the key hyperparameters: number of experts M1 and M2, top-k values k1 and k2, capacity α, the sample size N for variance estimation, or the loss weights of FEI, FET, FFI, FFT, and Ldistance. Without these, the experiments cannot be reproduced.
  3. [§4.3] Several sentences need rewriting for clarity, for example 'Utilizing FO-MoE achieves higher AUC for all attributes demonstrates and gains 1.1% improvements in race' is not grammatical and obscures the intended claim.
  4. [Figure 1 and Figure 2] The figures are referenced but the caption text does not fully explain the color coding or the data-flow steps in FO-MoE; please ensure the captions are self-contained so the reader can interpret the proposed architecture without the main text.
  5. [Table 1] The caption says 'green text highlights our method,' but the table as rendered does not use green highlighting; please fix the formatting or remove the caption statement.
  6. [Abstract/Section 1] There is a typo: 'textural data' should be 'textual data'.
  7. [§4.2] The statement that Fair-MoE 'outperforms baselines 5.00% in ES AUC' for race is ambiguous: please specify the baseline and whether the number is an absolute difference in ES-AUC points.
  8. [Abstract] The statement 'Code will be publicly available' should include a URL or a planned release identifier, as the current wording is not actionable.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claim is an empirical benchmark comparison against external baselines on Harvard-FairVLMed, and no equation reduces a prediction to a fitted constant or to a self-citation.

full rationale

Fair-MoE's central claim is empirical: it reports AUC, DPD, EOD, and ES-AUC on a held-out test set from Harvard-FairVLMed, comparing against CLIP and FairCLIP. The proposed method combines an MoE architecture with a fairness-oriented loss FOL, defined as the sum of gate-weight variance-difference terms and the existing Sinkhorn distance loss from FairCLIP. There is no derivation chain in which the reported fairness metrics are fitted parameters or are identical by construction to the optimized loss. FOL regularizes the variance of gating weights across protected groups; whether this proxy improves outcome-parity metrics is an empirical modeling assumption, not a circular identity. The ablation in Table 4 isolates the variance components and reports only ES-AUC, so the paper's stronger claim that the dispersion term is 'indispensable' for DPD/EOD improvements is under-supported, but this is a validation gap, not circularity. The numerous self-citations to the authors' earlier works are background and are not load-bearing for the benchmark claim. The internal inconsistency between the abstract's 'all four attributes' statement and Table 1's language/gender results under ViT-L14 is a correctness or reporting issue, not a circularity. Because the paper is self-contained against external benchmarks and does not reduce its central claim to its own inputs, the circularity score is 0.

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

The method rests on two unvalidated bridges: (1) FOL minimizes differences in the variance of gating weights across protected groups, but the evaluation metrics DPD/EOD measure outcome parity; no link is established between gate statistics and outcome fairness; (2) the top-k capacity filter in embedding-based MoE is claimed to remove biased patches, but the router is learned from data without bias labels and no analysis supports this. The hyperparameters of the MoE (expert counts, top-k, capacity, sample size, loss weights) are unreported, so the model cannot be reproduced or audited for tuning on the test set. No new entities are introduced.

free parameters (5)
  • M1, M2 (number of experts in embedding/feature MoE)
    Architecture hyperparameters governing expert capacity; values not reported in the paper.
  • k1, k2 (top-k gates)
    Sparsity level for routing; values not reported.
  • capacity C (alpha)
    Capacity limit for filtering embeddings; value not reported.
  • N (sample size for variance estimation)
    Number of data pairs sampled for FOL variance terms; value not reported.
  • loss weights of FEI, FET, FFI, FFT, Ldistance
    Relative weighting of FOL components; no values or tuning procedure given.
assumptions (5)
  • domain assumption CLIP's pretrained encoders provide suitable visual and textual feature spaces for medical images and text; the proposed MoE layers are inserted into these encoders.
    Section 3.1 modifies the last attention block MLP in CLIP encoders; CLIP is never cited and its validity for medical images is assumed.
  • domain assumption Harvard-FairVLMed's protected attribute labels (race, gender, ethnicity, language) and glaucoma labels are correct, and the train/val/test split follows FairCLIP.
    Section 4.1 relies on the dataset without verification; the paper does not analyze label noise or data quality.
  • ad hoc to paper The variance of per-expert gating weights across protected attribute groups is a valid proxy for outcome fairness (DPD/EOD).
    Section 3.2 defines FOL using variance of gate weights but provides no theoretical or empirical link to DPD/EOD beyond the final benchmark results; this is the bridge between the loss and the fairness metrics.
  • ad hoc to paper The top-k capacity filtering in embedding-based MoE removes biased patch embeddings rather than arbitrary low-weight patches.
    Section 3.1 states 'To filter out biased path embeddings, a capacity C is introduced' but the gating network is trained end-to-end without bias labels; no analysis shows the discarded patches correlate with protected attributes.
  • domain assumption The fairness metrics as defined (DPD, EOD, ES-AUC) capture the intended notion of fairness.
    Section 4.1 defines DPD conditional on y=1, which is a TPR gap rather than demographic parity; the paper assumes these metrics are standard, but DPD as defined overlaps with the TPR portion of EOD.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fair-MoE: Fairness-Oriented Mixture of Experts in Vision-Language Models." pith.science (2026). https://pith.science/paper/3CQDEB2B

@misc{pith2026250206094,
  author       = {Pith},
  title        = {Pith review of: Fair-MoE: Fairness-Oriented Mixture of Experts in Vision-Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3CQDEB2B}},
  note         = {Machine review of arXiv:2502.06094}
}
read the original abstract

Fairness is a fundamental principle in medical ethics. Vision Language Models (VLMs) have shown significant potential in the medical field due to their ability to leverage both visual and linguistic contexts, reducing the need for large datasets and enabling the performance of complex tasks. However, the exploration of fairness within VLM applications remains limited. Applying VLMs without a comprehensive analysis of fairness could lead to concerns about equal treatment opportunities and diminish public trust in medical deep learning models. To build trust in medical VLMs, we propose Fair-MoE, a model specifically designed to ensure both fairness and effectiveness. Fair-MoE comprises two key components: \textit{the Fairness-Oriented Mixture of Experts (FO-MoE)} and \textit{the Fairness-Oriented Loss (FOL)}. FO-MoE is designed to leverage the expertise of various specialists to filter out biased patch embeddings and use an ensemble approach to extract more equitable information relevant to specific tasks. FOL is a novel fairness-oriented loss function that not only minimizes the distances between different attributes but also optimizes the differences in the dispersion of various attributes' distributions. Extended experiments demonstrate the effectiveness and fairness of Fair-MoE. Tested on the Harvard-FairVLMed dataset, Fair-MoE showed improvements in both fairness and accuracy across all four attributes. Code will be publicly available.

Figures

Figures reproduced from arXiv: 2502.06094 by the authors.

Figure 1
Figure 1. An illustration of Fair-MoE architecture. MoE-based ar [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. (a) is an illustration of embedding-based MoE. Blue [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 32 canonical work pages

  1. [1]

    An image is worth 16x16 words: Trans- formers for image recognition at scale

    [Dosovitskiy et al., 2021] Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Min- derer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Trans- formers for image recognition at scale. ICLR,

    work page 2021

  2. [4]

    Med- think: Explaining medical visual question answering via multimodal decision-making rationale

    [Gai et al., 2024] Xiaotang Gai, Chenyi Zhou, Jiaxiang Liu, Yang Feng, Jian Wu, and Zuozhu Liu. Med- think: Explaining medical visual question answering via multimodal decision-making rationale. arXiv preprint arXiv:2404.12372,

    arXiv 2024

  3. [6]

    Beyond bias and discrimination: re- defining the ai ethics principle of fairness in healthcare machine-learning algorithms

    [Giovanola and Tiribelli, 2023] Benedetta Giovanola and Si- mona Tiribelli. Beyond bias and discrimination: re- defining the ai ethics principle of fairness in healthcare machine-learning algorithms. AI & society , 38(2):549– 563,

    work page 2023

  4. [8]

    Eclb: Efficient contrastive learn- ing on bi-level for noisy labels

    [Guan et al., 2024] Juwei Guan, Jiaxiang Liu, Shuying Huang, and Yong Yang. Eclb: Efficient contrastive learn- ing on bi-level for noisy labels. Knowledge-Based Sys- tems, page 112128,

    work page 2024

  5. [9]

    A visual–language foundation model for pathology im- age analysis using medical twitter

    [Huang et al., 2023] Zhi Huang, Federico Bianchi, Mert Yuksekgonul, Thomas J Montine, and James Zou. A visual–language foundation model for pathology im- age analysis using medical twitter. Nature medicine , 29(9):2307–2316,

    work page 2023

  6. [10]

    Jacobs, Michael I

    [Jacobs et al., 1991] Robert A. Jacobs, Michael I. Jordan, Steven J. Nowlan, and Geoffrey E. Hinton. Adaptive mix- tures of local experts. Neural Computation, 3(1):79–87,

    work page 1991

  7. [12]

    Visiongpt: Vision- language understanding agent using generalized multi- modal framework

    [Kelly et al., 2024b] Chris Kelly, Luhui Hu, Bang Yang, Yu Tian, Deshun Yang, Cindy Yang, Zaoshan Huang, Zi- hao Li, Jiayin Hu, and Yuexian Zou. Visiongpt: Vision- language understanding agent using generalized multi- modal framework. arXiv preprint arXiv:2403.09027 ,

    arXiv

  8. [13]

    How fair are medical imaging foundation models? In Machine Learning for Health (ML4H) , pages 217–231

    [Khan et al., 2023] Muhammad Osama Khan, Muham- mad Muneeb Afzal, Shujaat Mirza, and Yi Fang. How fair are medical imaging foundation models? In Machine Learning for Health (ML4H) , pages 217–231. PMLR,

    work page 2023

Show all 41 references
  1. [14]

    A chatgpt aided explainable framework for zero-shot medical image diagnosis

    [Liu et al., 2023c] Jiaxiang Liu, Tianxiang Hu, Yan Zhang, Xiaotang Gai, Yang Feng, and Zuozhu Liu. A chatgpt aided explainable framework for zero-shot medical image diagnosis. arXiv preprint arXiv:2307.01981,

    arXiv

  2. [15]

    Vpl: Visual proxy learning framework for zero-shot medical image diagnosis

    [Liu et al., 2024a] Jiaxiang Liu, Tianxiang Hu, Huimin Xiong, Jiawei Du, Yang Feng, Jian Wu, Joey Zhou, and Zuozhu Liu. Vpl: Visual proxy learning framework for zero-shot medical image diagnosis. In Findings of the As- sociation for Computational Linguistics: EMNLP 2024 , page...

    2024

  3. [16]

    Medcot: Medical chain of thought via hierarchical expert

    [Liu et al., 2024b] Jiaxiang Liu, Yuan Wang, Jiawei Du, Joey Zhou, and Zuozhu Liu. Medcot: Medical chain of thought via hierarchical expert. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 17371–17389,

    2024

  4. [17]

    Kpl: Training-free medical knowledge mining of vision- language models

    [Liu et al., 2025] Jiaxiang Liu, Tianxiang Hu, Jiawei Du, Ruiyuan Zhang, Joey Tianyi Zhou, and Zuozhu Liu. Kpl: Training-free medical knowledge mining of vision- language models. arXiv preprint arXiv:2501.11231,

    2025 arXiv

  5. [19]

    Biogpt: generative pre-trained transformer for biomedical text generation and mining

    [Luo et al., 2022] Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon, and Tie-Yan Liu. Biogpt: generative pre-trained transformer for biomedical text generation and mining. Briefings in bioinformatics , 23(6):bbac409,

    2022

  6. [20]

    Fairclip: Harnessing fairness in vision-language learning

    [Luo et al., 2024] Yan Luo, Min Shi, Muhammad Osama Khan, Muhammad Muneeb Afzal, Hao Huang, Shuaihang Yuan, Yu Tian, Luo Song, Ava Kouhana, Tobias Elze, et al. Fairclip: Harnessing fairness in vision-language learning. In Proceedings of the IEEE/CVF Conference on Computer Visi...

    2024

  7. [21]

    A survey on bias and fairness in machine learning

    [Mehrabi et al., 2021] Ninareh Mehrabi, Fred Morstatter, Nripsuta Saxena, Kristina Lerman, and Aram Galstyan. A survey on bias and fairness in machine learning. ACM computing surveys (CSUR), 54(6):1–35,

    2021

  8. [22]

    Fairness in deep learning: A survey on vision and language research

    [Parraga et al., 2023] Otavio Parraga, Martin D More, Chris- tian M Oliveira, Nathan S Gavenski, Lucas S Kupssinsk¨u, Adilson Medronha, Luis V Moura, Gabriel S Sim ˜oes, and Rodrigo C Barros. Fairness in deep learning: A survey on vision and language research. ACM Computing Surveys,

    2023

  9. [23]

    Com- putational optimal transport: With applications to data sci- ence

    [Peyr´e et al., 2019] Gabriel Peyr´e, Marco Cuturi, et al. Com- putational optimal transport: With applications to data sci- ence. Foundations and Trends® in Machine Learning , 11(5-6):355–607,

    2019

  10. [26]

    Medical image understanding with pretrained vision language models: A comprehensive study

    [Qin et al., 2022] Ziyuan Qin, Huahui Yi, Qicheng Lao, and Kang Li. Medical image understanding with pretrained vision language models: A comprehensive study. arXiv preprint arXiv:2209.15517,

    2022 arXiv

  11. [27]

    Scaling vision with sparse mixture of experts

    [Riquelme et al., 2021] Carlos Riquelme, Joan Puigcerver, Basil Mustafa, Maxim Neumann, Rodolphe Jenatton, Andr´e Susano Pinto, Daniel Keysers, and Neil Houlsby. Scaling vision with sparse mixture of experts. Advances in Neural Information Processing Systems, 34:8583–8595,

    2021

  12. [28]

    Dr-fairness: Dynamic data ratio adjustment for fair training on real and generated data

    [Roh et al., 2023] Yuji Roh, Weili Nie, De-An Huang, Steven Euijong Whang, Arash Vahdat, and Anima Anand- kumar. Dr-fairness: Dynamic data ratio adjustment for fair training on real and generated data. Transactions on Ma- chine Learning Research,

    2023

  13. [29]

    Feamoe: fair, explainable and adap- tive mixture of experts

    [Sharma et al., 2022] Shubham Sharma, Jette Henderson, and Joydeep Ghosh. Feamoe: fair, explainable and adap- tive mixture of experts. arXiv preprint arXiv:2210.04995,

    2022 arXiv

  14. [30]

    Conceptualising fairness: three pillars for medical algorithms and health equity

    [Sikstrom et al., 2022] Laura Sikstrom, Marta M Maslej, Katrina Hui, Zoe Findlay, Daniel Z Buchman, and Sean L Hill. Conceptualising fairness: three pillars for medical algorithms and health equity. BMJ health & care infor- matics, 29(1),

    2022

  15. [31]

    Fairness-related performance and explainability effects in deep learning models for brain image analysis.Journal of Medical Imag- ing, 9(6):061102–061102,

    [Stanley et al., 2022] Emma AM Stanley, Matthias Wilms, Pauline Mouches, and Nils D Forkert. Fairness-related performance and explainability effects in deep learning models for brain image analysis.Journal of Medical Imag- ing, 9(6):061102–061102,

    2022

  16. [32]

    Fairvit: Fair vision transformer via adaptive mask- ing

    [Tian et al., 2024] Bowei Tian, Ruijie Du, and Yanning Shen. Fairvit: Fair vision transformer via adaptive mask- ing. arXiv preprint arXiv:2407.14799,

    2024 arXiv

  17. [33]

    Tsnet: Integrating dental position prior and symptoms for tooth segmentation from cbct images

    [Tong et al., 2023] Linjie Tong, Jiaxiang Liu, Yang Feng, Tianxiang Hu, and Zuozhu Liu. Tsnet: Integrating dental position prior and symptoms for tooth segmentation from cbct images. In Medical Imaging with Deep Learning, short paper track,

    2023

  18. [34]

    Principles of clinical ethics and their application to practice

    [Varkey, 2021] Basil Varkey. Principles of clinical ethics and their application to practice. Medical Principles and Prac- tice, 30(1):17–28,

    2021

  19. [35]

    Medclip: Contrastive learn- ing from unpaired medical images and text

    [Wang et al., 2022] Zifeng Wang, Zhenbang Wu, Dinesh Agarwal, and Jimeng Sun. Medclip: Contrastive learn- ing from unpaired medical images and text. arXiv preprint arXiv:2210.10163,

    2022 arXiv

  20. [36]

    Multidisciplinary considerations of fairness in medical ai: A scoping review

    [Wang et al., 2023] Yue Wang, Yaxin Song, Zhuo Ma, and Xiaoxue Han. Multidisciplinary considerations of fairness in medical ai: A scoping review. International Journal of Medical Informatics, page 105175,

    2023

  21. [37]

    Elixr: Towards a general purpose x-ray artificial intelligence system through alignment of large language models and radiology vision encoders

    [Xu et al., 2023] Shawn Xu, Lin Yang, Christopher Kelly, Marcin Sieniek, Timo Kohlberger, Martin Ma, Wei- Hung Weng, Atilla Kiraly, Sahar Kazemzadeh, Zakkai Melamed, et al. Elixr: Towards a general purpose x-ray artificial intelligence system through alignment of large languag...

    2023 arXiv

  22. [38]

    Addressing fairness issues in deep learning-based medical image analysis: a systematic review

    [Xu et al., 2024] Zikang Xu, Jun Li, Qingsong Yao, Han Li, Mingyue Zhao, and S Kevin Zhou. Addressing fairness issues in deep learning-based medical image analysis: a systematic review. npj Digital Medicine, 7(1):286,

    2024

  23. [39]

    Infrared and visible image fusion via texture conditional generative adversarial network

    [Yang et al., 2021] Yong Yang, Jiaxiang Liu, Shuying Huang, Weiguo Wan, Wenying Wen, and Juwei Guan. Infrared and visible image fusion via texture conditional generative adversarial network. IEEE Transactions on Circuits and Systems for Video Technology, 31(12):4771– 4783,

    2021

  24. [40]

    Worldgpt: a sora-inspired video ai agent as rich world models from text and image inputs

    [Yang et al., 2024] Deshun Yang, Luhui Hu, Yu Tian, Zi- hao Li, Chris Kelly, Bang Yang, Cindy Yang, and Yuex- ian Zou. Worldgpt: a sora-inspired video ai agent as rich world models from text and image inputs. arXiv preprint arXiv:2403.07944,

    2024 arXiv

  25. [41]

    St-moe: Designing stable and transferable sparse expert models

    [Zoph et al., 2022] Barret Zoph, Irwan Bello, Sameer Ku- mar, Nan Du, Yanping Huang, Jeff Dean, Noam Shazeer, and William Fedus. St-moe: Designing stable and transferable sparse expert models. arXiv preprint arXiv:2202.08906, 2022

    2022 arXiv

  26. [1991]

    Visiongpt-3d: A generalized multimodal agent for enhanced 3d vision un- derstanding

    [Kelly et al., 2024a] Chris Kelly, Luhui Hu, Jiayin Hu, Yu Tian, Deshun Yang, Bang Yang, Cindy Yang, Zihao Li, Zaoshan Huang, and Yuexian Zou. Visiongpt-3d: A generalized multimodal agent for enhanced 3d vision un- derstanding. arXiv preprint arXiv:2403.09530,

    arXiv

  27. [2019]

    Justice: a key consideration in health policy and systems research ethics

    [Pratt et al., 2020] Bridget Pratt, Verina Wild, Edwine Barasa, Dorcas Kamuya, Lucy Gilson, Tereza Hendl, and Sassy Molyneux. Justice: a key consideration in health policy and systems research ethics. BMJ Global Health , 5(4):e001942,

    2020

  28. [2020]

    Fairness-aware vision transformer via debiased self-attention

    [Qiang et al., 2023] Yao Qiang, Chengyin Li, Prashant Khanduri, and Dongxiao Zhu. Fairness-aware vision transformer via debiased self-attention. arXiv preprint arXiv:2301.13803,

    2023 arXiv

  29. [2021]

    Switch transformers: Scaling to trillion param- eter models with simple and efficient sparsity

    [Fedus et al., 2022] William Fedus, Barret Zoph, and Noam Shazeer. Switch transformers: Scaling to trillion param- eter models with simple and efficient sparsity. Journal of Machine Learning Research, 23(120):1–39,

    2022

  30. [2022]

    E ad- dressing fairness in artificial intelligence for medical imag- ing nat

    [Ferrante, 2022] Ricci Lara MA Echeveste R Ferrante. E ad- dressing fairness in artificial intelligence for medical imag- ing nat. Commun, 13(1):4581,

    2022

  31. [2023]

    Algorithmic encoding of protected characteristics in chest x-ray disease detection models

    [Glocker et al., 2023] Ben Glocker, Charles Jones, M ´elanie Bernhardt, and Stefan Winzeck. Algorithmic encoding of protected characteristics in chest x-ray disease detection models. EBioMedicine, 89,

    2023

  32. [2024]

    Fairmoe: counterfactually-fair mixture of experts with levels of interpretability

    [Germino et al., 2024] Joe Germino, Nuno Moniz, and Nitesh V Chawla. Fairmoe: counterfactually-fair mixture of experts with levels of interpretability. Machine Learn- ing, pages 1–21,

    2024

  33. [2025]

    Cross-token modeling with con- ditional computation

    [Lou et al., 2021] Yuxuan Lou, Fuzhao Xue, Zangwei Zheng, and Yang You. Cross-token modeling with con- ditional computation. arXiv preprint arXiv:2109.02008 ,

    2021 arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.