Pith. sign in

REVIEW 4 major objections 6 minor 65 references

Guiding Noisy Label Conditional Diffusion Models with Score-based Discriminator Correction

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A discriminator trained on clean versus corrupted image-label pairs can steer a noisy conditional diffusion model back to the intended class-conditional distribution at sampling time, without retraining the model.

desk verdict A promising inference-time fix for noisy-label diffusion with a broken theoretical appendix; worth engaging with as a heuristic, not as a proven method. read the letter →

arxiv 2508.19581 v1 pith:SRV2L3YB submitted 2025-08-27 cs.CV

Dat Nguyen Cong , Hieu Tran Bao , Hoang Thanh-Tung This is my paper · ORCID
classification cs.CV
keywords diffusionmodelsnoisylabelsscore-basedguidancediscriminatorcorrectionconditionalgenerationinference-timenoisedetectionclass-conditionalquality
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 when a conditional diffusion model is trained on labels that are partly wrong, its score estimate—the gradient of the log-density that drives the reverse process—can be corrected during sampling by adding the gradient of a discriminator that separates clean image-label pairs from corrupted ones. That correction is derived as the log-likelihood ratio between the clean and noisy conditional distributions, so the discriminator acts as a learned stand-in for the missing piece of the score. The authors further claim that applying this correction only in the middle phase of the sampling trajectory, where class decisions are most unstable, improves class-conditional fidelity and diversity and keeps the added inference cost small. If true, the result is a practical way to make large pre-trained conditional generators robust to label noise without retraining them.

What carries the argument

The load-bearing identity is Eq. (8), the score-correction decomposition $\nabla_{x_t}\log p(x_t|y)=\nabla_{x_t}\log p_\theta(x_t|\tilde{y})+\nabla_{x_t}\log\frac{p(x_t|y)}{p_\theta(x_t|\tilde{y})}$, which isolates the mismatch caused by noisy labels. The mechanism that realizes it is a small time-dependent discriminator $D^t_\theta(x_t,\tilde{y})$ trained with binary cross-entropy on clean and corrupted samples, whose logit gradient approximates the likelihood-ratio term; a $\gamma$-gate restricts this guidance to the middle 'conditional phase' where class assignments are unstable, and two augmentations—Pseudo-clean Shuffle, which flips labels of clean samples to balance the corrupt set, and SiMix, a similarity-based input mixing scheme—stabilize the discriminator's training and gradients.

What would settle it

Take a diffusion model trained on clean CIFAR-10, corrupt its labels with a known transition matrix, and compare the true correction $\nabla_{x_t}\log(p(x_t|y)/p_\theta(x_t|\tilde{y}))$ against the discriminator gradient on a fixed batch of intermediate samples; if their cosine similarity is near zero, the approximation in Theorem 1 is not what the method is actually computing. A simpler behavioral test is to measure class-conditional accuracy (CAS) and FID with the gate off: if the correction is working, CAS should rise while FID stays roughly constant.

Watch

Extended reading notes

Core claim

The central claim is that the clean conditional score can be recovered from a noisy-label score model by adding a correction term: $\nabla_{x_t}\log p(x_t|y)=\nabla_{x_t}\log p_\theta(x_t|\tilde{y})+\nabla_{x_t}\log\frac{p(x_t|y)}{p_\theta(x_t|\tilde{y})}$. The paper trains a time-dependent discriminator on pseudo-clean and pseudo-corrupt pairs, filtered from the noisy training set by an off-the-shelf noise detector, and uses its logit gradient $\nabla_{x_t}\log\frac{D}{1-D}$ as the learned approximation to that intractable ratio. Theorem 1 states that for an optimally trained Lipschitz discriminator, the expected squared error between the discriminator gradient and the true correction is bounded by a constant plus the norm of the correction itself; the empirical sections then show that gating this correction to an early-to-middle interval of the reverse process improves class-wise FID, density, and coverage across symmetric, asymmetric, instance-dependent, and real-world label noise, with almost no increase in inference time.

Load-bearing premise

The method assumes the noise detector's clean/corrupt split reflects the true labels; if that split is biased, as the paper's own asymmetric-noise recall of 58.5 percent suggests, the discriminator correction steers samples toward the detector's notion of clean rather than the true clean distribution.

Editorial extensions

If this is right

  • Pre-trained conditional diffusion models can be made robust to label noise at inference time, with only a small discriminator and no retraining of the generative model.
  • Applying the discriminator correction only during the early-to-middle sampling phase improves class-conditional metrics while adding only marginal wall-clock time to generation.
  • The correction improves class-wise fidelity and diversity even at high noise rates, where the authors report the baseline and prior method produce many class-incorrect images.
  • Because the method is plug-and-play, it can be stacked on top of other noisy-label diffusion training schemes; the authors show it also improves models trained with the transition-matrix method they compare against.
  • The practical ceiling of the correction depends on the noise detector's recall, so better detection of corrupted labels should translate directly into stronger generative correction.

Reading between the lines

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

  • The same likelihood-ratio correction could be applied to text-conditioned diffusion models by training the discriminator on matched versus mismatched text-image pairs rather than clean versus corrupted labels, which the paper itself names as future work.
  • The Theorem 1 bound depends on the magnitude of the true correction, so at very high noise rates the approximation can degrade; the more informative bottleneck may be the noise detector's recall, which the paper reports dropping to 58.5 percent on asymmetric CIFAR-10.
  • The gate interval could be predicted from the paper's confidence and instability curves instead of tuned per dataset, which would make the method easier to deploy on new domains.
  • A direct test of the mechanism would be to compute the true clean-to-noisy score ratio on held-out samples with a known noise transition matrix and compare its cosine similarity with the discriminator gradient.
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

4 major / 6 minor

Summary. The paper introduces Score-based Discriminator Correction (SBDC), an inference-time guidance method for conditional diffusion models trained on noisy labels. The method trains a small discriminator to distinguish clean from corrupted (image, label) pairs, using labels from a noise detector to pseudo-label training data. During sampling, the gradient of the discriminator's logit is added to the score network, gated to a limited time interval, with the aim of steering the noisy conditional model toward the clean distribution. The paper claims a theoretical guarantee (Theorem 1) that the discriminator gradient approximates the required score correction, and reports experiments on CIFAR-10/100, Tiny-ImageNet, ImageNet-128, and real-world datasets (Food101, Clothing1M), showing improvements over the TDSM baseline on class-wise metrics.

Significance. If the theoretical claim were sound, the method would be a valuable practical contribution: it is a plug-and-play module that requires no retraining of the diffusion model, adds negligible inference overhead, and improves conditional alignment under label noise. The empirical study is extensive, covering multiple noise types and levels, and includes useful ablations that identify the limited guidance interval as a key ingredient. The proposed augmentations (Pseudo-clean Shuffle and SiMix) are interesting and appear to provide benefits beyond the main guidance. However, the central theoretical justification is not correct: the proof of Theorem 1 in Appendix A contains unjustified cancellations and yields a vacuous bound, and the discriminator trained with Eq. (9) does not estimate the log-ratio required in Eq. (8). Since the paper explicitly states that the method is 'derived' from the theory, the unsupported theory is a load-bearing weakness.

major comments (4)
  1. [Appendix A, Theorem 1] The proof is invalid. The transition from Eq. (18) to Eq. (19) drops the cross terms between the gradient of the discriminator logit and the target log-ratio without justification; the two score-matching terms that are cancelled involve different label variables (y and y~) and only coincide if the marginal distribution of x0 is identical under both, which is not established. More importantly, the resulting bound in Eq. (10) is vacuous: the constant C1 is exactly the squared norm of the target correction term, so a discriminator with zero gradient satisfies the bound trivially. Thus Theorem 1 does not establish that the discriminator gradient approximates the correction term, and the claim in Sec. 3.2 that the guidance is 'derived' from the theory is unsupported.
  2. [Sec. 3.2, Eqs. (8)-(11)] The discriminator trained with Eq. (9) estimates the log-odds of the pseudo-clean versus pseudo-corrupt status given the observed label, not the log-ratio of clean and noisy conditional densities. For symmetric noise, the discriminator logit gradient equals the gradient of log p(xt|y~) minus log sum_{y' != y~} p(xt|y')/(K-1) (up to additive constants), whereas Eq. (8) requires the gradient of log p(xt|y) minus log p_noisy(xt|y~), which subtracts the full noisy conditional. These quantities differ unless the noise rate is zero or the clean class dominates; the paper provides no argument closing this gap. Consequently, the guidance signal implemented in Eq. (11) is not the theoretically derived clean-distribution score.
  3. [Table 3] On original (clean) CIFAR-10, applying SBDC degrades FID from 1.88 (oracle) to 3.19, yet the paper neither reports nor analyzes this failure mode. This is important because it shows the method is not universally beneficial and can harm performance when the training labels are already clean. The limited-interval analysis in Sec. 4.5 does not address clean-data behavior, and the conclusion that SBDC 'consistently outperforms' baselines is not accurate for this setting.
  4. [Sec. 3.4 and Table 5] The theoretical claim assumes an optimal discriminator, but in practice the discriminator is trained on pseudo-labels from a noise detector. Table 9 shows that for asymmetric noise, CORES recall is only 58.5%, meaning many true corrupt samples are treated as clean. The paper acknowledges the importance of recall but does not analyze how imperfect pseudo-labels affect the correction term or provide any robustness bound. This disconnect between the idealized theory and the practical implementation is not addressed.
minor comments (6)
  1. [Abstract] The phrase 'limiting the usage of our guidance to the early phase' contradicts Sec. 4.5, where the guidance interval is the middle phase (steps 8-16 of 18). Please correct the abstract.
  2. [Sec. 4.5] The sentence 'we set Sclip min = 1.5 and Sclip min = 50.0' should read 'Sclip max = 50.0'.
  3. [Eq. (5)] The notation 'log p^phi_theta(x_t)' and 'log D^t_phi(x_t)/(1-D^t_phi(x_t))' is typeset inconsistently; the superscripts are misplaced. Please clarify.
  4. [Tables 1 and 5] Reported Density values are around 100, whereas the standard Density metric of Naeem et al. is typically O(1). Please clarify the scaling or the metric definition used.
  5. [Algorithm 4] In lines 5-7, the comparison 'y_t^prev != y_t^prev' is self-referential; the intended comparison is between consecutive steps. Please fix the indexing.
  6. [Introduction] The paper claims the method is 'derived' from theory, but the proof of Theorem 1 is in the supplementary material (Appendix A). Please indicate in the main text that the proof is provided in the appendix.

Circularity Check

1 steps flagged · score 6.0 of 10

Theorem 1's error bound is vacuous: the RHS contains the norm of the target correction itself, so the bound is satisfied by any Lipschitz discriminator, including a constant one; the theoretical claim that the discriminator estimates the log-likelihood ratio is not established.

  1. self definitional [Appendix A, proof of Theorem 1, Eqs. (18)-(21); stated in Sec. 3.2, Theorem 1, Eq. (10)]
    "LHS = E_{xt,˜y} [||∇xtgθ(xt, ˜y)||₂²] + C2, (19) ... C2 = C1 (20) Finally, from our assumption, gθ(·, y,t ) has Lipschitz constant L, then we have LHS ≤ L + C1 (21)"

    In Theorem 1's proof, the LHS error between the discriminator gradient and the target correction is reduced to E||∇gθ*||² + C1, where C1 = E||∇log p(xt|y)/p(xt|˜y)||² is the squared norm of the very term the discriminator is supposed to approximate. The final bound LHS ≤ L + C1 therefore adds the target norm to both sides and reduces to the Lipschitz assumption E||∇gθ*||² ≤ L. This bound holds for any discriminator with bounded gradient, including a constant discriminator with ∇gθ* = 0, so it cannot certify that the discriminator estimates the log-likelihood ratio. The theorem is tautological rather than a derivation of the claimed approximation in Eq. (8)/(11).

full rationale

The paper's empirical pipeline is not circular: the discriminator is trained on pseudo-clean/pseudo-corrupt training pairs filtered by an external noise detector, and its guidance is evaluated against held-out class-conditional quality metrics; no metric is fitted into the discriminator objective. There is no load-bearing self-citation. The circularity is confined to the theoretical justification. Theorem 1's proof reduces the approximation error to E||∇gθ*||² + C1, where C1 is the squared norm of the target log-ratio gradient, and the final bound LHS ≤ L + C1 is a tautology of the Lipschitz assumption: a constant discriminator with ∇gθ*=0 satisfies it. Thus the theorem does not establish that the discriminator estimates the log-likelihood ratio; the claimed first-principles support for Eq. (8)/(11) is, on inspection, a vacuous bound rather than a derivation. The method may still work empirically, but the theory as written is circular in this one load-bearing step.

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

The method relies on a trained discriminator and a noise detector; both are components whose behavior is assumed, not derived. The free parameters are hyperparameters selected by validation, and the core theoretical argument assumes optimal discriminator and perfect score learning, which are strong idealizations.

free parameters (4)
  • guidance strength gamma = 0.9
    Selected by FID evaluation for all experiments (Sec 4.5, Fig 5c).
  • guidance interval (Sclip_min, Sclip_max) = 1.5 and 50.0 (steps 8-16 of 18)
    Chosen by inspecting C(t)/I(t) curves and CW-FID sweeps (Sec 4.5, Fig 5a,b).
  • SiMix Beta parameter alpha = 0.2
    Set for all experiments; no sensitivity analysis provided (Sec 4.1).
  • Pseudo-clean Shuffle rate = 30%
    Applied to 20% asymmetric and 20% IDN to balance clean/corrupt sets (Sec 4.1).
assumptions (5)
  • domain assumption The score network sθ(xt, y~) perfectly learns the noisy distribution p(xt|y~).
    Stated in Sec 3.2 before Eq (8); idealization that the pre-trained model has converged to the noisy conditional.
  • domain assumption The discriminator converges to the Bayes-optimal classifier for the clean/corrupt binary problem.
    Assumed in Theorem 1 (optimal D*); not guaranteed with finite data and adversarial training.
  • domain assumption The discriminator logits g satisfy a Lipschitz condition.
    Explicit assumption in Theorem 1; not verified for the implemented network.
  • domain assumption Noise-detector pseudo-labels reflect true clean/corrupt status.
    Implicit in the training pipeline (Sec 3.4); performance depends on CORES/CL quality.
  • standard math The forward diffusion perturbation kernel p(xt|x0) is independent of the label y.
    Used in the proof (Appendix A) to cancel terms; standard property of the diffusion forward process.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Guiding Noisy Label Conditional Diffusion Models with Score-based Discriminator Correction." pith.science (2026). https://pith.science/paper/SRV2L3YB

@misc{pith2026250819581,
  author       = {Pith},
  title        = {Pith review of: Guiding Noisy Label Conditional Diffusion Models with Score-based Discriminator Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SRV2L3YB}},
  note         = {Machine review of arXiv:2508.19581}
}
read the original abstract

Diffusion models have gained prominence as state-of-the-art techniques for synthesizing images and videos, particularly due to their ability to scale effectively with large datasets. Recent studies have uncovered that these extensive datasets often contain mistakes from manual labeling processes. However, the extent to which such errors compromise the generative capabilities and controllability of diffusion models is not well studied. This paper introduces Score-based Discriminator Correction (SBDC), a guidance technique for aligning noisy pre-trained conditional diffusion models. The guidance is built on discriminator training using adversarial loss, drawing on prior noise detection techniques to assess the authenticity of each sample. We further show that limiting the usage of our guidance to the early phase of the generation process leads to better performance. Our method is computationally efficient, only marginally increases inference time, and does not require retraining diffusion models. Experiments on different noise settings demonstrate the superiority of our method over previous state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2508.19581 by the authors.

Figure 1
Figure 1. The uncurated generated images of EDM (left) and [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall training (left) and inference (bottom right) procedure of the proposed approach. The solid curve from [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. (a): Illustration of the three phases in a two-mode dis [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Ambiguous samples of CIFAR-10, filtered by noise de [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Ablation studies for several parameters, termination [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: SBDC with different checkpoints of the discriminator. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 8
Figure 8. Figure 8: Confidence score on ImageNet 20% Symmetric noise. [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: Comparison of the baseline model (first row) with our proposed method (second row). Discriminator Guidance is applied with [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Comparison of quality and diversity for two data aug [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: Confusion matrix on CIFAR10 - 40% IDN. F.10. Extended Examples [PITH_FULL_IMAGE:figures/full_fig_p016_11.png]
Figure 12
Figure 12. Figure 12: Illustration of the sampling process with correct guidance for two initial noises. The first row is the origin, and the second and [PITH_FULL_IMAGE:figures/full_fig_p017_12.png]
Figure 13
Figure 13. Figure 13: The uncurated generated images of baseline (left column), TDSM (middle column), and SBDC (right column) on the CIFAR-10 [PITH_FULL_IMAGE:figures/full_fig_p018_13.png]
Figure 14
Figure 14. Figure 14: The uncurated generated images of baseline (left column), TDSM (middle column), and SBDC (right column) on the CIFAR [PITH_FULL_IMAGE:figures/full_fig_p019_14.png]
Figure 15
Figure 15. Figure 15: The uncurated generated images of baseline (left column), TDSM (middle column), and SBDC (right column) on the FOOD101 [PITH_FULL_IMAGE:figures/full_fig_p020_15.png]
Figure 16
Figure 16. Figure 16: The uncurated generated images of baseline (left column), TDSM (middle column), and SBDC (right column) on the Cloth [PITH_FULL_IMAGE:figures/full_fig_p021_16.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

65 extracted references · 42 canonical work pages

  1. [1]

    Reverse-time diffusion equation mod- els

    Brian DO Anderson. Reverse-time diffusion equation mod- els. Stochastic Processes and their Applications, 12(3):313– 326, 1982. 2

    1982

  2. [2]

    From noisy predic- tion to true label: Noisy prediction calibration via generative model

    HeeSun Bae, Seungjae Shin, Byeonghu Na, JoonHo Jang, Kyungwoo Song, and Il-Chul Moon. From noisy predic- tion to true label: Noisy prediction calibration via generative model. In International Conference on Machine Learning , pages 1277–1297. PMLR, 2022. 1, 5

    work page 2022

  3. [3]

    Understand- ing and improving early stopping for learning with noisy la- bels

    Yingbin Bai, Erkun Yang, Bo Han, Yanhua Yang, Jiatong Li, Yinian Mao, Gang Niu, and Tongliang Liu. Understand- ing and improving early stopping for learning with noisy la- bels. Advances in Neural Information Processing Systems , 34:24392–24403, 2021. 8

    work page 2021

  4. [4]

    Frozen in time: A joint video and image encoder for end-to-end retrieval

    Max Bain, Arsha Nagrani, G ¨ul Varol, and Andrew Zisser- man. Frozen in time: A joint video and image encoder for end-to-end retrieval. In Proceedings of the IEEE/CVF inter- national conference on computer vision , pages 1728–1738,

  5. [5]

    A note on the inception score

    Shane Barratt and Rishi Sharma. A note on the inception score. arXiv preprint arXiv:1801.01973, 2018. 5

    arXiv 2018

  6. [6]

    Are we done with imagenet? arXiv preprint arXiv:2006.07159, 2020

    Lucas Beyer, Olivier J H ´enaff, Alexander Kolesnikov, Xi- aohua Zhai, and A ¨aron van den Oord. Are we done with imagenet? arXiv preprint arXiv:2006.07159, 2020. 1

    arXiv 2006

  7. [7]

    Food-101–mining discriminative components with random forests

    Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. Food-101–mining discriminative components with random forests. In Computer vision–ECCV 2014: 13th European conference, zurich, Switzerland, September 6-12, 2014, pro- ceedings, part VI 13, pages 446–461. Springer, 2014. 5, 4

    work page 2014

  8. [8]

    Denoising likelihood score match- ing for conditional score-based data generation

    Chen-Hao Chao, Wei-Fang Sun, Bo-Wun Cheng, Yi-Chen Lo, Chia-Che Chang, Yu-Lun Liu, Yu-Lin Chang, Chia-Ping Chen, and Chun-Yi Lee. Denoising likelihood score match- ing for conditional score-based data generation. ICLR, 2022. 6

    work page 2022

Show all 65 references
  1. [9]

    Slight corruption in pre-training data makes better dif- fusion models

    Hao Chen, Yujin Han, Diganta Misra, Xiang Li, Kai Hu, Difan Zou, Masashi Sugiyama, Jindong Wang, and Bhiksha Raj. Slight corruption in pre-training data makes better dif- fusion models. Advances in neural information processing systems, 2024. 2

    2024

  2. [10]

    Label-retrieval- augmented diffusion models for learning from noisy labels

    Jian Chen, Ruiyi Zhang, Tong Yu, Rohan Sharma, Zhiqiang Xu, Tong Sun, and Changyou Chen. Label-retrieval- augmented diffusion models for learning from noisy labels. Advances in Neural Information Processing Systems , 36,

  3. [11]

    Learning with instance-dependent label noise: A sample sieve approach, 2021

    Hao Cheng, Zhaowei Zhu, Xingyu Li, Yifei Gong, Xing Sun, and Yang Liu. Learning with instance-dependent label noise: A sample sieve approach, 2021. 1, 5, 2, 4

    2021

  4. [12]

    Mitigat- ing memorization of noisy labels via regularization between representations

    Hao Cheng, Zhaowei Zhu, Xing Sun, and Yang Liu. Mitigat- ing memorization of noisy labels via regularization between representations. arXiv preprint arXiv:2110.09022, 2021. 8

    2021 arXiv

  5. [13]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009. 1

    2009

  6. [14]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural informa- tion processing systems, 34:8780–8794, 2021. 4

    2021

  7. [15]

    Sharpness-aware minimization for efficiently improving generalization

    Pierre Foret, Ariel Kleiner, Hossein Mobahi, and Behnam Neyshabur. Sharpness-aware minimization for efficiently improving generalization. arXiv preprint arXiv:2010.01412,

    2010 arXiv

  8. [16]

    Generative adversarial nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In Advances in Neural Information Processing Systems. Curran Associates, Inc., 2014. 8

    2014

  9. [17]

    Noise-contrastive estimation: A new estimation principle for unnormalized statistical models

    Michael Gutmann and Aapo Hyv ¨arinen. Noise-contrastive estimation: A new estimation principle for unnormalized statistical models. In Proceedings of the thirteenth inter- national conference on artificial intelligence and statistics , pages 297–304. JMLR Workshop and Conferen...

    2010

  10. [18]

    Co- teaching: Robust training of deep neural networks with ex- tremely noisy labels

    Bo Han, Quanming Yao, Xingrui Yu, Gang Niu, Miao Xu, Weihua Hu, Ivor Tsang, and Masashi Sugiyama. Co- teaching: Robust training of deep neural networks with ex- tremely noisy labels. Advances in neural information pro- cessing systems, 31, 2018. 5, 8, 2

    2018

  11. [19]

    Improving generalization by controlling label- noise information in neural network weights

    Hrayr Harutyunyan, Kyle Reing, Greg Ver Steeg, and Aram Galstyan. Improving generalization by controlling label- noise information in neural network weights. In Interna- tional Conference on Machine Learning , pages 4071–4081. PMLR, 2020. 8

    2020

  12. [20]

    Gans trained by a two time-scale update rule converge to a local nash equilib- rium

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilib- rium. Advances in neural information processing systems , 30, 2017. 5

    2017

  13. [21]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 1

    2020

  14. [22]

    Label-noise robust generative adversarial networks

    Takuhiro Kaneko, Yoshitaka Ushiku, and Tatsuya Harada. Label-noise robust generative adversarial networks. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2467–2476, 2019. 1, 6, 8

    2019

  15. [23]

    Elucidating the design space of diffusion-based generative models

    Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-based generative models. Advances in neural information processing systems, 35:26565–26577, 2022. 6

    2022

  16. [24]

    Refining generative process with discriminator guidance in score-based diffusion mod- els

    Dongjun Kim, Yeongmin Kim, Se Jung Kwon, Wanmo Kang, and Il-Chul Moon. Refining generative process with discriminator guidance in score-based diffusion mod- els. ICML, 2023. 3, 4

    2023

  17. [25]

    Learning discriminative dynamics with label corruption for noisy label detection

    Suyeon Kim, Dongha Lee, SeongKu Kang, Sukang Chae, Sanghwan Jang, and Hwanjo Yu. Learning discriminative dynamics with label corruption for noisy label detection. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 22477–22487, 2024. 4, 8

    2024

  18. [26]

    Nlnl: Negative learning for noisy labels

    Youngdong Kim, Junho Yim, Juseung Yun, and Junmo Kim. Nlnl: Negative learning for noisy labels. In Proceedings of the IEEE/CVF international conference on computer vision, pages 101–110, 2019. 5

    2019

  19. [27]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. 2009. 3, 5, 2

    2009

  20. [28]

    Learning with noisy labels by efficient transition ma- trix estimation to combat label miscorrection

    Seong Min Kye, Kwanghee Choi, Joonyoung Yi, and Buru Chang. Learning with noisy labels by efficient transition ma- trix estimation to combat label miscorrection. In European Conference on Computer Vision , pages 717–738. Springer,

  21. [29]

    Applying guidance in a limited interval improves sample and distribution quality in diffusion models

    Tuomas Kynk ¨a¨anniemi, Miika Aittala, Tero Karras, Samuli Laine, Timo Aila, and Jaakko Lehtinen. Applying guidance in a limited interval improves sample and distribution quality in diffusion models. arXiv preprint arXiv:2404.07724, 2024. 4

    2024 arXiv

  22. [30]

    Tiny imagenet visual recognition challenge

    Yann Le and Xuan Yang. Tiny imagenet visual recognition challenge. CS 231N, 7(7):3, 2015. 5

    2015

  23. [31]

    Robust training with en- semble consensus

    Jisoo Lee and Sae-Young Chung. Robust training with en- semble consensus. arXiv preprint arXiv:1910.09792, 2019. 8

    1910 arXiv

  24. [32]

    Neighbor- hood collective estimation for noisy label identification and correction

    Jichang Li, Guanbin Li, Feng Liu, and Yizhou Yu. Neighbor- hood collective estimation for noisy label identification and correction. In European Conference on Computer Vision , pages 128–145. Springer, 2022. 8

    2022

  25. [33]

    Early-learning regularization pre- vents memorization of noisy labels

    Sheng Liu, Jonathan Niles-Weed, Narges Razavian, and Car- los Fernandez-Granda. Early-learning regularization pre- vents memorization of noisy labels. Advances in neural in- formation processing systems, 33:20331–20342, 2020. 8

    2020

  26. [34]

    Label- noise robust diffusion models

    Byeonghu Na, Yeongmin Kim, HeeSun Bae, Jung Hyun Lee, Se Jung Kwon, Wanmo Kang, and Il chul Moon. Label- noise robust diffusion models. In The Twelfth International Conference on Learning Representations, 2024. 1, 2, 5, 6, 8

    2024

  27. [35]

    Reliable fidelity and diversity metrics for generative models

    Muhammad Ferjad Naeem, Seong Joon Oh, Youngjung Uh, Yunjey Choi, and Jaejun Yoo. Reliable fidelity and diversity metrics for generative models. In International Conference on Machine Learning, pages 7176–7185. PMLR, 2020. 6

    2020

  28. [36]

    Confident learning: Estimating uncertainty in dataset labels

    Curtis Northcutt, Lu Jiang, and Isaac Chuang. Confident learning: Estimating uncertainty in dataset labels. Journal of Artificial Intelligence Research, 70:1373–1411, 2021. 4, 5

    2021

  29. [37]

    Memorization in deep neural net- works: Does the loss function matter? In Pacific-Asia Con- ference on Knowledge Discovery and Data Mining , pages 131–142

    Deep Patel and PS Sastry. Memorization in deep neural net- works: Does the loss function matter? In Pacific-Asia Con- ference on Knowledge Discovery and Data Mining , pages 131–142. Springer, 2021. 8

    2021

  30. [38]

    Identifying mislabeled data using the area under the margin ranking

    Geoff Pleiss, Tianyi Zhang, Ethan Elenberg, and Kilian Q Weinberger. Identifying mislabeled data using the area under the margin ranking. Advances in Neural Information Pro- cessing Systems, 33:17044–17056, 2020. 4, 3

    2020

  31. [39]

    Learning transferable visual models from natural language supervi- sion

    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 supervi- sion. In International conference on machine learning, ...

    2021

  32. [40]

    Classification accuracy score for conditional generative models

    Suman Ravuri and Oriol Vinyals. Classification accuracy score for conditional generative models. Advances in neural information processing systems, 32, 2019. 2

    2019

  33. [41]

    Improved techniques for training gans

    Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, Xi Chen, and Xi Chen. Improved techniques for training gans. In Advances in Neural Infor- mation Processing Systems . Curran Associates, Inc., 2016. 5

    2016

  34. [42]

    Laion-5b: An open large-scale dataset for training next generation image-text models

    Christoph Schuhmann, Romain Beaumont, Richard Vencu, Cade Gordon, Ross Wightman, Mehdi Cherti, Theo Coombes, Aarush Katta, Clayton Mullis, Mitchell Worts- man, et al. Laion-5b: An open large-scale dataset for training next generation image-text models. Advances in Neural In- f...

    2022

  35. [43]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In Proceedings of the 32nd International Conference on Machine Learning, pages 2256–2265, Lille, France, 2015. PMLR. 1

    2015

  36. [44]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502, 2020. 4

    2010 arXiv

  37. [45]

    Generative modeling by esti- mating gradients of the data distribution

    Yang Song and Stefano Ermon. Generative modeling by esti- mating gradients of the data distribution. Advances in neural information processing systems, 32, 2019. 3, 1

    2019

  38. [46]

    Score-based generative modeling through stochastic differential equa- tions

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equa- tions. arXiv preprint arXiv:2011.13456, 2020. 2

    2011 arXiv

  39. [47]

    Robustness of conditional gans to noisy la- bels

    Kiran K Thekumparampil, Ashish Khetan, Zinan Lin, and Sewoong Oh. Robustness of conditional gans to noisy la- bels. Advances in neural information processing systems , 31, 2018. 6, 8

    2018

  40. [48]

    Identifying and eliminating csam in genera- tive ml training data and models

    David Thiel. Identifying and eliminating csam in genera- tive ml training data and models. Technical report, Techni- cal Report. Stanford University, Palo Alto, CA. https://purl. stanford . . . , 2023. 1

    2023

  41. [49]

    A connection between score matching and denoising autoencoders

    Pascal Vincent. A connection between score matching and denoising autoencoders. Neural computation, 23(7):1661– 1674, 2011. 1

    2011

  42. [50]

    Diffu- siondb: A large-scale prompt gallery dataset for text-to- image generative models

    Zijie J Wang, Evan Montoya, David Munechika, Haoyang Yang, Benjamin Hoover, and Duen Horng Chau. Diffu- siondb: A large-scale prompt gallery dataset for text-to- image generative models. arXiv preprint arXiv:2210.14896,

    arXiv

  43. [51]

    Robust early-learning: Hindering the memorization of noisy labels

    Xiaobo Xia, Tongliang Liu, Bo Han, Chen Gong, Nannan Wang, Zongyuan Ge, and Yi Chang. Robust early-learning: Hindering the memorization of noisy labels. InInternational conference on learning representations, 2020. 8, 2

    2020

  44. [52]

    Part-dependent label noise: Towards instance-dependent label noise

    Xiaobo Xia, Tongliang Liu, Bo Han, Nannan Wang, Ming- ming Gong, Haifeng Liu, Gang Niu, Dacheng Tao, and Masashi Sugiyama. Part-dependent label noise: Towards instance-dependent label noise. Advances in Neural Infor- mation Processing Systems, 33:7597–7610, 2020. 5, 2

    2020

  45. [53]

    Sample selection with uncertainty of losses for learning with noisy labels

    Xiaobo Xia, Tongliang Liu, Bo Han, Mingming Gong, Jun Yu, Gang Niu, and Masashi Sugiyama. Sample selection with uncertainty of losses for learning with noisy labels. arXiv preprint arXiv:2106.00445, 2021. 5

    2021 arXiv

  46. [54]

    Learning from massive noisy labeled data for im- age classification

    Tong Xiao, Tian Xia, Yi Yang, Chang Huang, and Xiaogang Wang. Learning from massive noisy labeled data for im- age classification. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2691–2699,

  47. [55]

    Dual t: Reduc- ing estimation error for transition matrix in label-noise learn- ing

    Yu Yao, Tongliang Liu, Bo Han, Mingming Gong, Jiankang Deng, Gang Niu, and Masashi Sugiyama. Dual t: Reduc- ing estimation error for transition matrix in label-noise learn- ing. Advances in neural information processing systems, 33: 7260–7271, 2020. 1

    2020

  48. [56]

    On learning contrastive representations for learning with noisy labels

    Li Yi, Sheng Liu, Qi She, A Ian McLeod, and Boyu Wang. On learning contrastive representations for learning with noisy labels. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 16682– 16691, 2022. 8

    2022

  49. [57]

    How does disagreement help gener- alization against label corruption? In International confer- ence on machine learning, pages 7164–7173

    Xingrui Yu, Bo Han, Jiangchao Yao, Gang Niu, Ivor Tsang, and Masashi Sugiyama. How does disagreement help gener- alization against label corruption? In International confer- ence on machine learning, pages 7164–7173. PMLR, 2019. 8

    2019

  50. [58]

    Understanding deep learning (still) requires rethinking generalization

    Chiyuan Zhang, Samy Bengio, Moritz Hardt, Benjamin Recht, and Oriol Vinyals. Understanding deep learning (still) requires rethinking generalization. Communications of the ACM, 64(3):107–115, 2021. 8

    2021

  51. [59]

    mixup: Beyond empirical risk minimiza- tion

    Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Beyond empirical risk minimiza- tion. arXiv preprint arXiv:1710.09412, 2017. 5

    2017 arXiv

  52. [60]

    Learning noise transition matrix from only noisy labels via total varia- tion regularization

    Yivan Zhang, Gang Niu, and Masashi Sugiyama. Learning noise transition matrix from only noisy labels via total varia- tion regularization. In International Conference on Machine Learning, pages 12501–12512. PMLR, 2021. 1

    2021

  53. [61]

    Differentiable augmentation for data-efficient gan training

    Shengyu Zhao, Zhijian Liu, Ji Lin, Jun-Yan Zhu, and Song Han. Differentiable augmentation for data-efficient gan training. Advances in neural information processing systems, 33:7559–7570, 2020. 3

    2020

  54. [62]

    Robust cur- riculum learning: from clean label detection to noisy label self-correction

    Tianyi Zhou, Shengjie Wang, and Jeff Bilmes. Robust cur- riculum learning: from clean label detection to noisy label self-correction. In International Conference on Learning Representations, 2020. 8

    2020

  55. [63]

    Detecting cor- rupted labels without training a model to predict

    Zhaowei Zhu, Zihao Dong, and Yang Liu. Detecting cor- rupted labels without training a model to predict. In Interna- tional conference on machine learning, pages 27412–27427. PMLR, 2022. 1, 4, 8 Guiding Noisy Label Conditional Diffusion Models with Score-based Discriminator Co...

    2022

  56. [64]

    First, given the objective in Eq

    Then, given an optimally trainedDθ∗, we have E xt,y,˜y, yr,yf h ∇xt log Dθ∗(xt, yr) Dθ∗(xt, yf)−∇ xt logp(xt|y) p(xt|˜y) 2 2 i ≤L + E xt,y,˜y h ∇xt logp(xt|y) p(xt|˜y) 2 2 i (10) theorem Proof. First, given the objective in Eq. (9), the optimal dis- criminator is easily derive...

  57. [65]

    unconditional term

    as shown in Algorithm 3. Algorithm 3 Instance Dependent Noise Generation Process Require: Clean samples (xi, yi)n i=1, Noise rateτ 1: Sample instance flip rates q∈ Rn from the truncated normal distributionN (τ, 0.12, [0, 1]); 2: Independently sample w1, w2,..., wc from the sta...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.