Inference Time Debiasing Concepts in Diffusion Models

Jordan Kopper; Lucas S. Kupssinsk\"u; Marco N. Bochernitsan; Ot\'avio Parraga; Rodrigo C. Barros

REVIEW 3 major objections 4 minor 24 references

Inference Time Debiasing Concepts in Diffusion Models

T0 review · 3 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 DeCoDi debiases text-to-image diffusion models at inference time by subtracting a biased-concept guidance term from the classifier-free guidance equation, shifting gender, ethnicity, and age distributions with no retraining and negligible c

desk verdict The central debiasing claim collapses on the paper's own CEO-age data: it swaps an elderly majority for a young one, and Section 4.4 admits the result is 'equaly biased' — yet the conclusion still claims success. read the letter →

arxiv 2508.14933 v1 pith:R4BUPKD4 submitted 2025-08-19 cs.GR cs.AIcs.LG

Lucas S. Kupssinskü , Marco N. Bochernitsan , Jordan Kopper , Otávio Parraga , Rodrigo C. Barros This is my paper

classification cs.GRcs.AIcs.LG

keywords inference-timedebiasingdiffusionmodelsclassifier-freeguidancetext-to-imagegenerationfairnessandbiasprotectedattributesvision-languagemodelevaluationSDXL

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

DeCoDi is a debiasing procedure that works entirely at inference time for classifier-free guided diffusion models: rather than retraining or fine-tuning, it subtracts a "biased concept" guidance term from the noise prediction during denoising, steering generation away from the protected-attribute value a model over-represents. Tested on gender for nurses, ethnicity for firefighters, and age for CEOs, the paper claims the procedure shifts the demographic distribution of generated images while leaving measured image quality essentially unchanged and adding negligible compute. Two human evaluators and GPT-4o annotated 1,200 generated images, and the results showed reliable agreement between evaluators and clear diversity gains in the debiased outputs. The paper also argues that a vision-language model can substitute for human bias evaluation, since GPT-4o's annotations were statistically indistinguishable from a human evaluator in most tested settings. The practical payoff, if the claims hold, is that any practitioner with a frozen diffusion model can debias it without touching the training pipeline.

What carries the argument

The load-bearing object is the adjusted denoising prediction ε_DeCoDi(z_t,c,b) = ε_θ(z_t) + s_g[ε_θ(z_t,c) − ε_θ(z_t) − γ(z_t,c,b)]: standard classifier-free guidance minus a bias guidance term γ(z_t,c,b) = s_b[ε_θ(z_t,b) − ε_θ(z_t)] + s_m v_t. Here b is a text prompt describing the over-represented attribute value (feminine-coded words for nurses, Eurocentric features for firefighters, elderly markers for CEOs), and v_t = βv_t + (1−β)γ_t is a momentum accumulator that stabilizes steering across denoising steps. The difference ε_θ(z_t,b) − ε_θ(z_t) is the latent direction of the biased concept; subtracting it is meant to change only that attribute. A warm-up hyperparameter δ=7 delays bias gu

What would settle it

Using the paper's own data, the decisive test is whether DeCoDi can balance rather than invert an attribute distribution: the CEO-age debias produced 87.5% young images (up from 1.25%), so sweeping s_b and the biased-concept prompt should reveal whether any setting brings the debiased age distribution near uniform while CLIP-Score stays flat. A second test targets the linear-separability premise directly: compute ε_θ(z_t,b) − ε_θ(z_t) at several timesteps and measure its cosine alignment with the direction that actually flips the protected attribute (using an attribute classifier on decoded im

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that classifier-free guidance can be edited at generation time to push a diffusion model away from a biased concept. DeCoDi combines three noise estimates—unconditioned, prompt-conditioned, and biased-concept-conditioned—as ε_DeCoDi(z_t,c,b) = ε_θ(z_t) + s_g[ε_θ(z_t,c) − ε_θ(z_t) − γ(z_t,c,b)], where γ subtracts the latent direction of the over-represented attribute value plus a momentum term, starting at a warm-up timestep. Empirically the protected attribute shifts while pose, expression, and prompt semantics persist: nurse images went from 99.5% to 13% female, firefighter images from 88.5% to 11.25% white, CEO images from 55.75% to 0.5% elderly.

Load-bearing premise

The load-bearing premise is that the noise-prediction difference between a prompt describing the biased attribute and the empty prompt isolates exactly that protected attribute in latent space, so subtracting it during denoising changes only that attribute and leaves the rest of the image intact; the paper's own side-effect and overshoot results show this isolation is only approximate.

Editorial extensions

If this is right

Any classifier-free guided diffusion model, not just SDXL, can be debiased by practitioners who lack training compute, because the change lives entirely in the inference-time noise-prediction equation.
Bias monitoring for image generators can be largely automated: GPT-4o annotations were statistically indistinguishable from a human evaluator in most tested settings, so vision-language judges could work as a cheap screening step.
The steering can pass beyond balance: the CEO-age debias replaced an elderly skew with a young skew (87.5% young), and the firefighter debias flipped a white majority into a Black/Indian majority — the method shifts distributions rather than guaranteeing balance.
Debiasing one attribute tends to shift others: the paper documents co-debiasing of ethnicity when debiasing CEO age, and of ethnicity and age when debiasing nurse gender, which it attributes to entanglement of attributes in latent space.
Image quality as measured by CLIP-Score and normalized KL-divergence stays roughly flat between original and debiased outputs, supporting the no-significant-visual-degradation claim.

Reading between the lines

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

Because the same fixed bias scale is applied regardless of how strongly the protected attribute is entangled with other semantics, a natural extension is attribute-aware or timestep-aware scaling; the paper's overcorrection cases suggest DeCoDi could land closer to a uniform distribution if the subtraction strength were calibrated per attribute.
The GPT-4o-as-judge result holds up best for gender and worst for age, where the human annotators themselves disagreed in about half of the CEO-age images; an automated bias-monitoring pipeline should therefore weight vision-language verdicts by attribute and keep humans in the loop for age.
The side-effect co-debiasing the paper documents points toward a multi-attribute extension: subtracting several biased-concept directions simultaneously, with per-direction scales, could handle intersectional bias in one inference pass instead of one attribute at a time.
Since the biased-concept direction is defined by a text prompt rather than by model internals, the method's ceiling depends on prompt quality; comparing positive bias descriptions (as used here) against negative-prompt formulations would be a cheap, testable way to probe how well the prompt isolates the attribute.

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.

Desk Editor's Note

read the letter

Quick take: the headline claim — that DeCoDi debiases age, gender, and ethnicity at inference with negligible overhead — does not survive the paper's own numbers. The CEO-age result is a bias flip, not a mitigation: original 55.75% elderly becomes 87.5% young and 0.5% elderly. The authors admit this in Section 4.4 ("equaly biased"), yet the conclusion still says the method "substantially lower[s] the bias." That is an internal contradiction, not a minor caveat.

What is genuinely here: a simple, inference-only guidance update (SLD plus a momentum term), a concrete biased-concept prompt, and real evaluation effort — 1,200 images, two human annotators, plus a GPT-4o-as-judge comparison. The method is easy to reproduce and the code is promised. The GPT-4o agreement results, where they hold, are a useful data point for cheap bias evaluation. That is real work, and it deserves credit.

The soft spots are substantial. First, the same flip appears for nurse gender (99.5% female to 87% male). Overcorrection is structurally expected because the update subtracts the b-conditioned direction with no balancing or parity term. Second, there are no baselines: no SEGA, no SLD ablation, no comparison to Jiang et al., so you cannot tell what the momentum or warm-up is adding. Third, the chi-square "equivalence" argument reads failure-to-reject as equivalence, and the paper openly skips multiple-test correction, which worsens the problem. Fourth, the targeted-attribute reduction is partly circular: the guidance explicitly removes the biased attribute, so observing that the attribute is reduced is not evidence of fairness. A proper diversity metric over all categories, or an explicit side-effect measure, is needed. Fifth, reproducibility is thin: hyperparameters are listed, but the seed list and annotation instructions are not in the text, and the code link sits in a footnote.

This is an incremental but potentially useful technique, honestly described in some places and overclaimed in others. The paper is not a sound demonstration of debiasing as advertised. I would not cite it as evidence that inference-time debiasing works; I would cite it as an example of an overcorrection failure mode.

If this crossed my desk, I would not desk-reject it outright — the evaluation protocol and the GPT-4o comparison are worth a referee's time — but I would send it out with a strong warning that the central claim is internally contradicted. As it stands, reject after major revision, or reject outright if the venue runs on first-glance soundness.

Referee Report

3 major / 4 minor

Summary. The paper proposes DeCoDi, an inference-time debiasing method for classifier-free guided diffusion models. The adjusted prediction is ε_DeCoDi(z_t,c,b) = ε_θ(z_t) + s_g[ε_θ(z_t,c) − ε_θ(z_t) − γ(z_t,c,b)], where γ subtracts a biased-concept conditional difference ε_θ(z_t,b) − ε_θ(z_t) plus a momentum term. The method is evaluated on SDXL for three concepts—CEO with age bias, firefighter with ethnicity bias, and nurse with gender bias—using two human annotators and GPT-4o on 1,200 generated images. The paper claims that DeCoDi mitigates gender, ethnicity, and age biases, preserves image quality, adds negligible compute overhead, and that GPT-4o evaluation is not statistically distinct from human evaluation.

Significance. If the claims held, an inference-only debiasing method for diffusion models would be a practically valuable contribution, especially because it requires no retraining or fine-tuning. The authors provide a public code repository, a reasonably large human annotation effort, and a comparison with GPT-4o as an automated judge. These are concrete strengths. However, the central claim of bias mitigation is contradicted by the paper's own quantitative results: in two of the three concept/attribute pairs, the procedure does not move the generated distribution toward parity but flips it to the opposite extreme. The acknowledged CEO and nurse results are not a minor side effect; they are failures of the headline claim. The statistical-equivalence argument for GPT-4o also rests on an invalid use of chi-square tests. As a result, the contribution as stated is not established.

major comments (3)

[Table 2, §4.1, §4.4] The CEO age result directly contradicts the abstract and conclusion. Original CEO images are 55.75% elderly and 1.25% young; debiased CEO images are 0.5% elderly and 87.5% young. This is not mitigation but a bias flip to a more extreme distribution. Section 4.4 explicitly admits the CEO images are 'equaly biased, but this time toward a distinct valuation of the protected attribute.' A procedure that replaces one skew with the opposite skew cannot support the claim of 'substantially lower the bias' or 'effective in mitigating biases based on age.'
[Sec. 3.1, Table 2] The update rule has no balancing term toward the non-biased categories. γ(z_t,c,b) subtracts only the b-conditioned direction, where b is the biased concept prompt (e.g., 'female' for nurse). Consequently, overcorrection is structural. The nurse gender result confirms this: original images are 99.5% female and 0.5% male; debiased images are 13% female and 87% male. Reducing the targeted attribute value is not the same as debiasing, and the paper's own evaluation criterion—diversity across protected attribute values—is not achieved. The observed reduction in the b-attribute is essentially constructed by the subtraction in γ.
[Sec. 3.3, §4.2, §4.3] The statistical claim that GPT-4o is 'not significantly statistically distinct' from human evaluators is based on failing to reject the null hypothesis in a chi-square test. Failure to reject H0 is not evidence of equivalence, especially without any power analysis. The paper even states that the 'positive' result is to fail to reject the null and that no multiple-testing correction was applied. This is an invalid inference and undermines the abstract's claim that GPT-4o evaluation is 'not significantly statistically distinct from a human evaluation.' The same issue affects the conclusion that VLM judges are a viable substitute for human evaluation.

minor comments (4)

[Sec. 3.1] The biased-concept noise estimate is written as ε_θ(z_t,c) in one sentence and as ε_θ(z_t,b) in the equation. Please use consistent notation, e.g., ε_θ(z_t,b) for the biased-concept conditioning.
[Table 2] The CEO middle-age row reads '43.%' instead of '43.00%'. Also, the table caption and text would benefit from clarifying that all percentages are over the 400 generated images per condition (200 per evaluator).
[Fig. 3 caption] The caption says 'Hyperparametrization two reffers to the original model without debiasing,' while the text in §4.4 says the original model is 'marked with the number one.' This is inconsistent and should be corrected.
[§4.1, §4.4] Typos: 'oberseved' should be 'observed'; 'equaly' should be 'equally'; 'hiperparametrization' should be 'hyperparameterization'; 'reffers' should be 'refers.'

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity; the only self-citation is a taxonomy label, and the target-attribute outcome is a mechanism check rather than a fitted prediction.

full rationale

DeCoDi is an explicit inference-time intervention: the adjusted noise estimate subtracts a biased-concept guidance term gamma(z_t,c,b)=s_b(eps_theta(z_t,b)-eps_theta(z_t))+s_m v_t, where b is a text description of the protected attribute to be reduced. The paper then measures the frequency of that same protected attribute in generated images. This alignment is not circular in the prohibited sense: the biased-concept prompt is a fixed natural-language description (generated once by GPT4o), not a parameter fitted to the evaluation set; the guidance scale and momentum are fixed hyperparameters; and the human annotators are blind to model origin. The direction of change is by design, but the magnitude and side effects are empirical, as shown by the CEO age result in Sec. 4.4, where the paper admits the debiased images are 'equaly biased, but this time toward a distinct valuation of the protected attribute.' That outcome contradicts the headline claim and demonstrates the result was not forced by construction. No prediction is derived from a fitted parameter, and no load-bearing result is imported from the authors' prior work. The only self-citation ([15], Parraga et al.) supplies the 'Inferential Vector-Space Manipulation' taxonomy label and is not used to justify the method's correctness. The quality-preservation and GPT4o-agreement results rest on external, blind evaluations. Thus the derivation chain is self-contained; circularity score is low.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim depends on hand-chosen guidance scales and the untested premise that the biased-concept noise difference linearly isolates protected attributes. No parameters are fitted to data, but the s_b scale is missing entirely, and the biased prompts are free textual choices that determine the outcome direction.

free parameters (6)

delta (warm-up timestep) = 7
Hand-chosen hyperparameter controlling when biased guidance starts (Sec. 3.1).
s_g (guidance scale) = 7.5
Standard CFG guidance scale adopted without tuning.
beta (momentum coefficient) = 0.7
Hand-chosen momentum term in v_t update.
s_m (momentum scale) = 0.5
Hand-chosen scale for momentum contribution.
s_b (bias guidance scale) = not reported
Appears in Eq. (2) but its value is never stated; essential to reproduce the method.
biased concept prompts = three GPT-4o-generated prompts (Sec. 3.2)
Defines the direction of debiasing; generated by GPT-4o and selected by authors; changing the prompt changes the outcome.

assumptions (5)

domain assumption Classifier-free guidance's linear interpolation of noise estimates steers semantics linearly in latent space.
Invoked in Sec. 3.1, Eq. (1), the core mechanism assumes the noise prediction difference is a usable semantic direction.
ad hoc to paper Subtracting the biased-concept conditional difference epsilon_theta(z_t,b) - epsilon_theta(z_t) removes the attribute without corrupting other semantics.
This is the paper's core premise; it is untested directly and contradicted by observed side effects and overcorrection.
ad hoc to paper Failing to reject the null hypothesis in a chi-square test indicates equivalence between annotators or models.
Used throughout Sec. 4.2 and 4.3; failure to reject can occur due to low power, so equivalence is not established.
domain assumption Human annotators can infer gender, ethnicity, and age from single generated face images.
The entire evaluation depends on this (Sec. 3.3); the paper itself notes age and race are subjectively hard.
domain assumption SDXL is representative of diffusion-based text-to-image models.
Stated in Sec. 3; used to generalize the claimed applicability to 'any diffusion-based image generation model'.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Inference Time Debiasing Concepts in Diffusion Models." pith.science (2026). https://pith.science/paper/R4BUPKD4

@misc{pith2026250814933,
  author       = {Pith},
  title        = {Pith review of: Inference Time Debiasing Concepts in Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R4BUPKD4}},
  note         = {Machine review of arXiv:2508.14933}
}

read the original abstract

We propose DeCoDi, a debiasing procedure for text-to-image diffusion-based models that changes the inference procedure, does not significantly change image quality, has negligible compute overhead, and can be applied in any diffusion-based image generation model. DeCoDi changes the diffusion process to avoid latent dimension regions of biased concepts. While most deep learning debiasing methods require complex or compute-intensive interventions, our method is designed to change only the inference procedure. Therefore, it is more accessible to a wide range of practitioners. We show the effectiveness of the method by debiasing for gender, ethnicity, and age for the concepts of nurse, firefighter, and CEO. Two distinct human evaluators manually inspect 1,200 generated images. Their evaluation results provide evidence that our method is effective in mitigating biases based on gender, ethnicity, and age. We also show that an automatic bias evaluation performed by the GPT4o is not significantly statistically distinct from a human evaluation. Our evaluation shows promising results, with reliable levels of agreement between evaluators and more coverage of protected attributes. Our method has the potential to significantly improve the diversity of images it generates by diffusion-based text-to-image generative models.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

24 extracted references · 10 canonical work pages

[1]

arXiv preprint arXiv:2406.19097 (2024)

Adewumi, T., Alkhaled, L., Gurung, N., van Boven, G., Pagliai, I.: Fairness and bias in multimodal ai: A survey. arXiv preprint arXiv:2406.19097 (2024)

arXiv 2024
[2]

Barocas, S., Selbst, A.D.: Big data’s disparate impact. Calif. L. Rev.104, 671 (2016)

2016
[3]

Besiroglu, T., Bergerson, S.A., Michael, A., Heim, L., Luo, X., Thompson, N.: The compute divide in machine learning: A threat to academic contribution and scrutiny? arXiv preprint arXiv:2401.02452 (2024)

arXiv 2024
[4]

In: The 2024 ACM Conference on Fairness, Accountability, and Transparency

Birhane, A., Dehdashtian, S., Prabhu, V., Boddeti, V.: The dark side of dataset scaling: Evaluating racial classification in multimodal models. In: The 2024 ACM Conference on Fairness, Accountability, and Transparency. pp. 1229–1244 (2024)

work page 2024
[5]

Advances in Neural Information Processing Systems36, 25365–25389 (2023)

Brack, M., Friedrich, F., Hintersdorf, D., Struppek, L., Schramowski, P., Kersting, K.: Sega: Instructing text-to-image models using semantic guidance. Advances in Neural Information Processing Systems36, 25365–25389 (2023)

work page 2023
[6]

In: Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision

Cho, J., Zala, A., Bansal, M.: Dall-eval: Probing the reasoning skills and social biases of text-to-image generation models. In: Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision. pp. 3043–3054 (2023)

work page 2023
[7]

In: International Conference on Machine Learning

Choi, K., Grover, A., Singh, T., Shu, R., Ermon, S.: Fair generative modeling via weak supervision. In: International Conference on Machine Learning. pp. 1887–

work page
[8]

Hessel, J., Holtzman, A., Forbes, M., Bras, R.L., Choi, Y.: Clipscore: A reference- freeevaluationmetricforimagecaptioning.arXivpreprintarXiv:2104.08718(2021)

arXiv 2021

Show all 24 references

[9]

arXiv preprint arXiv:2207.12598 (2022) Inference Time Debiasing Concepts in Diffusion Models 15

Ho, J., Salimans, T.: Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598 (2022) Inference Time Debiasing Concepts in Diffusion Models 15

2022 arXiv
[10]

In: ACM Multi- media 2024 (2023)

Jiang, Y., Lyu, Y., He, Z., Peng, B., Dong, J.: Mitigating social biases in text-to- image diffusion models via linguistic-aligned attention guidance. In: ACM Multi- media 2024 (2023)

2024
[11]

arXiv preprint arXiv:2501.12612 (2025)

Li, L., Shi, Z., Hu, X., Dong, B., Qin, Y., Liu, X., Sheng, L., Shao, J.: T2isafety: Benchmark for assessing fairness, toxicity, and privacy in image generation. arXiv preprint arXiv:2501.12612 (2025)

2025 arXiv
[12]

In: Proceedings of International Conference on Computer Vision (ICCV) (December 2015)

Liu, Z., Luo, P., Wang, X., Tang, X.: Deep learning face attributes in the wild. In: Proceedings of International Conference on Computer Vision (ICCV) (December 2015)

2015
[13]

ACM computing surveys (CSUR)54(6), 1–35 (2021)

Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., Galstyan, A.: A survey on bias and fairness in machine learning. ACM computing surveys (CSUR)54(6), 1–35 (2021)

2021
[14]

Science366(6464), 447–453 (2019)

Obermeyer, Z., Powers, B., Vogeli, C., Mullainathan, S.: Dissecting racial bias in an algorithm used to manage the health of populations. Science366(6464), 447–453 (2019)

2019
[15]

ACM Comput

Parraga, O., More, M.D., Oliveira, C.M., Gavenski, N.S., Kupssinskü, L.S., Medronha, A., Moura, L.V., Simões, G.S., Barros, R.C.: Fairness in deep learning: A survey on vision and language research. ACM Comput. Surv.57(6) (Feb 2025). https://doi.org/10.1145/3637549

2025 doi
[16]

arXiv preprint arXiv:2307.01952 (2023)

Podell, D., English, Z., Lacey, K., Blattmann, A., Dockhorn, T., Müller, J., Penna, J., Rombach, R.: Sdxl: Improving latent diffusion models for high-resolution image synthesis. arXiv preprint arXiv:2307.01952 (2023)

2023 arXiv
[17]

In: Proceedings of the 2020 conference on fairness, accountability, and transparency

Raghavan, M., Barocas, S., Kleinberg, J., Levy, K.: Mitigating bias in algorithmic hiring: Evaluating claims and practices. In: Proceedings of the 2020 conference on fairness, accountability, and transparency. pp. 469–481 (2020)

2020
[18]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Schramowski, P., Brack, M., Deiseroth, B., Kersting, K.: Safe latent diffusion: Mitigating inappropriate degeneration in diffusion models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 22522– 22531 (2023)

2023
[19]

arXiv preprint arXiv:2308.00755 (2023)

Seshadri, P., Singh, S., Elazar, Y.: The bias amplification paradox in text-to-image generation. arXiv preprint arXiv:2308.00755 (2023)

2023 arXiv
[20]

arXiv preprint arXiv:1909.01326 (2019)

Sheng, E., Chang, K.W., Natarajan, P., Peng, N.: The woman worked as a babysit- ter: On biases in language generation. arXiv preprint arXiv:1909.01326 (2019)

1909 arXiv
[21]

Proceedings of the IEEE112(1), 4–11 (2024)

Togelius, J., Yannakakis, G.N.: Choose your weapon: Survival strategies for de- pressed ai academics [point of view]. Proceedings of the IEEE112(1), 4–11 (2024)

2024
[22]

Mind 59(236), 433–460 (1950), http://www.jstor.org/stable/2251299

Turing, A.M.: Computing machinery and intelligence. Mind 59(236), 433–460 (1950), http://www.jstor.org/stable/2251299

1950
[23]

arXiv preprint arXiv:2112.07804 (2021)

Xiao, Z., Kreis, K., Vahdat, A.: Tackling the generative learning trilemma with denoising diffusion gans. arXiv preprint arXiv:2112.07804 (2021)

2021 arXiv
[24]

arXiv preprint arXiv:2403.19738 (2024)

Yesiltepe, H., Akdemir, K., Yanardag, P.: Mist: Mitigating intersectional bias with disentangled cross-attention editing in text-to-image diffusion models. arXiv preprint arXiv:2403.19738 (2024)

2024 arXiv

Pith tools

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

[1] [1]

arXiv preprint arXiv:2406.19097 (2024)

Adewumi, T., Alkhaled, L., Gurung, N., van Boven, G., Pagliai, I.: Fairness and bias in multimodal ai: A survey. arXiv preprint arXiv:2406.19097 (2024)

arXiv 2024

[2] [2]

Barocas, S., Selbst, A.D.: Big data’s disparate impact. Calif. L. Rev.104, 671 (2016)

2016

[3] [3]

Besiroglu, T., Bergerson, S.A., Michael, A., Heim, L., Luo, X., Thompson, N.: The compute divide in machine learning: A threat to academic contribution and scrutiny? arXiv preprint arXiv:2401.02452 (2024)

arXiv 2024

[4] [4]

In: The 2024 ACM Conference on Fairness, Accountability, and Transparency

Birhane, A., Dehdashtian, S., Prabhu, V., Boddeti, V.: The dark side of dataset scaling: Evaluating racial classification in multimodal models. In: The 2024 ACM Conference on Fairness, Accountability, and Transparency. pp. 1229–1244 (2024)

work page 2024

[5] [5]

Advances in Neural Information Processing Systems36, 25365–25389 (2023)

Brack, M., Friedrich, F., Hintersdorf, D., Struppek, L., Schramowski, P., Kersting, K.: Sega: Instructing text-to-image models using semantic guidance. Advances in Neural Information Processing Systems36, 25365–25389 (2023)

work page 2023

[6] [6]

In: Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision

Cho, J., Zala, A., Bansal, M.: Dall-eval: Probing the reasoning skills and social biases of text-to-image generation models. In: Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision. pp. 3043–3054 (2023)

work page 2023

[7] [7]

In: International Conference on Machine Learning

Choi, K., Grover, A., Singh, T., Shu, R., Ermon, S.: Fair generative modeling via weak supervision. In: International Conference on Machine Learning. pp. 1887–

work page

[8] [8]

Hessel, J., Holtzman, A., Forbes, M., Bras, R.L., Choi, Y.: Clipscore: A reference- freeevaluationmetricforimagecaptioning.arXivpreprintarXiv:2104.08718(2021)

arXiv 2021