Pith. sign in

REVIEW 5 major objections 4 minor 48 references

Diffusion Disambiguation Models for Partial Label Learning

T0 review · 5 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read This paper claims that partial label learning can be recast as a diffusion-based generative denoising problem, and that the resulting model outperforms eight prior methods on most real and synthetic benchmarks.

desk verdict A serious diffusion-for-PLL paper with a real algebraic error in the reverse equation and a shallow EM proof; the empirical claims actually survive arithmetic, so it deserves major revision rather than rejection. read the letter →

arxiv 2507.00411 v2 pith:7W5A4OCO submitted 2025-07-01 cs.LG

classification cs.LG
keywords partiallabellearningdiffusionmodelsdisambiguationtransition-awarematrixpseudo-cleanlabelsweaklysupervisedgenerativeclassification
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that partial label learning, the weakly supervised setting where each instance comes with a set of candidate labels but only one is correct, can be solved by treating the label itself as a quantity to be generated and denoised. A diffusion model is trained to reconstruct a clean label vector from a noisy one, conditioned on the instance's features, and the candidate-label sets are first converted into a pseudo-clean label matrix using agreement between instance neighbors and candidate-label overlap. A transition-aware matrix is then estimated and used to iteratively update the pseudo-clean labels toward the likely ground-truth labels, with the whole loop interpreted as an EM algorithm. The paper reports that this diffusion-based disambiguation outperforms eight prior partial-label learning methods in 85% of real-world comparisons and in 97.5% of synthetic comparisons when a strong pre-trained encoder is used. If true, this would make generative label denoising a practical and competitive route for weakly supervised classification.

What carries the argument

The load-bearing objects are the pseudo-clean label matrix $S$ and the transition-aware matrix $T$. $S$ is built as $S=(P\odot J)Y$, where $P$ is a $k$-nearest-neighbor adjacency over pre-trained instance features, $J$ is a Jaccard similarity between candidate-label sets, and $Y$ is the candidate-label matrix; this masks out negative nearest neighbors that share no label support. $T$ estimates how the true label gets perturbed into candidate sets, with $T_{ij}=p(y_i\in S\mid y=y_j)$, and is re-estimated from $S$ at each iteration. The reverse diffusion model predicts the noise $\epsilon_\theta(S_t,x,f_\phi,t)$, from which a denoised label $\tilde{S}_0$ is computed, and $T$ then converts the noisy candidate-label distribution into a corrected label distribution through $[T]^{-1}$. Iterating the update $S^{e+1}=\text{Normalize}((S^e+[T^e]^{-1}\tilde{S}^e_0)S^e)$ couples the diffusion denoiser and the transition estimate, and the paper interprets this coupling as an EM algorithm that converges as the two components improve.

What would settle it

On a synthetic partial-label dataset with known ground truth, estimate $T$ via Eq. 11, check whether it is invertible, and compare $p(y|x)$ with $[T]^{-1}p(S|x)$; if the matrix is singular or the recovered posterior diverges from the true labels, the transition-aware correction is not doing the claimed work.

Watch

Extended reading notes

Core claim

The central claim is that ambiguous candidate labels are not just noise to be filtered by instance similarity alone, but samples from a stochastic label-generation process. DDMP defines a forward process in which a pseudo-clean label $S_0$ is gradually corrupted toward a latent distribution centered on the encoded instance $f_\phi(x)$, and learns a reverse process $p_\theta(S_{t-1}|S_t,x,f_\phi)$ that recovers the clean label. To start this loop, the paper constructs the initial pseudo-clean matrix as $S=(P\odot J)Y$, where $P$ is a $k$-nearest-neighbor adjacency over pre-trained instance features, $J$ is a Jaccard similarity between candidate-label sets, and $Y$ is the candidate-label matrix, so that negative nearest neighbors with no shared label support are suppressed. A transition-aware matrix $T$ with entries $T_{ij}=p(y_i\in S\mid y=y_j)$ is estimated from the current pseudo-clean labels and used to update them via $S^{e+1}=\text{Normalize}((S^e+[T^e]^{-1}\tilde{S}^e_0)S^e)$, progressively correcting the candidate-label distribution toward $p(y|x)$. The paper further argues that this update has EM structure, with the E-step computing posterior weights over the hidden ground-truth label and the M-step updating the model and the transition matrix by weighted maximum likelihood, and that the iterative refinement converges as the two components improve.

Load-bearing premise

The load-bearing premise is that one fixed, invertible transition matrix can exactly recover the true label distribution from the candidate-label distribution, and that the pseudo-clean labels are accurate enough to estimate that matrix so that inverting it recovers the ground truth.

Editorial extensions

If this is right

  • Partial label learning becomes a generative modeling problem: the same reverse-denoising machinery used for image generation can be pointed at label vectors, so improvements in diffusion models should carry over to disambiguation.
  • The Hadamard product of instance adjacency and candidate-label Jaccard similarity gives a simple way to suppress negative-nearest-neighbor interference before diffusion training, which the ablations show is worth several accuracy points.
  • The transition-aware matrix turns label refinement into an EM-style loop, so the iterative update of pseudo-clean labels has a convergence rationale instead of being a heuristic.
  • When a strong pre-trained encoder is used, the method claims 97.5% wins over prior methods on synthetic benchmarks; this makes pre-trained feature quality a direct lever on disambiguation performance.
  • The reported calibration improvement, with an expected calibration error of 0.007 on CIFAR-10, suggests that modeling label generation stochastically also yields more reliable confidence estimates.

Reading between the lines

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

  • Extension: The pseudo-clean-plus-transition loop is generic and should apply to other weak-supervision settings, such as noisy labels, complementary labels, or multiple annotators, wherever a transition between an observed label set and a hidden true label can be estimated.
  • Because the forward latent mean is fixed by a pre-trained instance encoder, DDMP's advantage may partly come from the encoder rather than the diffusion mechanism; an ablation with an untrained or random encoder would test how much of the reported margin is due to denoising.
  • The paper itself notes that the transition-aware matrix is non-identifiable without extra assumptions, so the method's robustness likely depends on the initial pseudo-clean labels; starting $S$ from random candidate-label weights rather than $(P\odot J)Y$ would expose how much the transition correction adds.
  • If the EM interpretation is correct, the update in Eq. 12 is effectively a soft-label expectation, so the diffusion loop could be replaced by any classifier that outputs soft labels, making the generative framing a template rather than a necessary component.
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

5 major / 4 minor

Summary. The paper proposes DDMP, a diffusion-based method for partial label learning. A pseudo-clean label matrix is constructed from k-NN instance adjacency and Jaccard label similarity; a conditional diffusion model is trained to denoise these labels; a transition-aware matrix T is estimated from the current pseudo-clean labels and used to recover class posteriors; the pseudo-clean labels are then iteratively updated. The authors claim state-of-the-art results on real-world and synthetic PLL benchmarks and provide an EM-style convergence argument.

Significance. If the claims held, the paper would introduce a genuinely new connection between diffusion generative models and partial label learning, and it ships several genuine strengths: an open-source implementation, comparisons with eight methods on five real-world and five synthetic datasets, an ablation study, and a calibration analysis. However, the central derivation and the headline empirical claims are not reliable as written. Equation (9) is algebraically inconsistent with the forward process in Eq. (4), the reported win rates disagree with the numbers in Tables II and III, and the EM analysis in Section IV-I does not constitute a proof. The contribution is therefore not established in the current form.

major comments (5)
  1. [Section III-C, Eq. (9)] The reverse prediction is not the algebraic inverse of the forward process in Eq. (4). The forward process is S_t = sqrt(\bar\alpha_t) S_0 + (1 - sqrt(\bar\alpha_t)) f_\phi(x) + sqrt(1 - \bar\alpha_t) \epsilon, while Eq. (9) predicts \tilde S_0 = (S_t - (1 - sqrt(\alpha_t)) f_\phi(x) - sqrt(1 - \alpha_t) \epsilon_\theta)/sqrt(\alpha_t). Solving Eq. (4) for S_0 would require dividing by sqrt(\bar\alpha_t) and using the coefficients 1 - sqrt(\bar\alpha_t) and sqrt(1 - \bar\alpha_t). Since this predicted \tilde S_0 enters the transition-matrix update in Eq. (12) and the classifier loss, the mismatch is load-bearing. Either the forward process or the reverse formula must be corrected, and the subsequent derivations re-checked.
  2. [Section IV-D, Table II] The claim that DDMP "significantly outperforms all other methods in 85.0% of the cases" is not supported by the table, and the word "significantly" is unjustified. DDMP is numerically best on Lost (74.58% vs 74.26%), MSRCv2 (53.71% vs 53.66%), Birdsong (80.36% vs 74.70%), and SoccerPlayer (62.83% vs 62.77%), but the margins on Lost, MSRCv2, and SoccerPlayer are well within the reported standard deviations, and on YahooNews DDMP (54.56%) trails PiCO, VALEN, CRDPLL, CAVL, DIRK, and PRODEN. No significance test is reported, so 34/40 pointwise wins cannot be described as "significantly" outperforming. The table also contradicts the text's statement that DDMP achieves superior performance on real-world datasets.
  3. [Section IV-D2, Table III] The stated 97.5% win rate for DDMP(CLIP) is contradicted by the table. Counting the entries, DDMP(CLIP) is below PiCO on all three Kuzushiji-MNIST settings (94.01% vs 97.68%, 93.89% vs 97.34%, 93.45% vs 97.15%) and below DIRK on Fashion-MNIST at q=0.1 (93.51% vs 93.71%), giving at least 4 losses in 120 comparisons, or 96.7% rather than 97.5%. The accompanying sentence that DDMP(SimCLR) "is 70.8% higher than comparative methods" is also unclear and appears to be a percentage-of-wins statement, not a performance gain.
  4. [Section IV-I, Eqs. (17)-(20)] The EM-style analysis does not establish convergence, and the update rules in Eqs. (11)-(12) are not derived from it. The E-step posterior in Eq. (17) is not equal to the multiplicative update in Eq. (12), which is presented as a separately motivated moving-average strategy. The M-step for T in Eq. (20) has mismatched indices relative to Eq. (11) and is not a consequence of the complete-data log-likelihood in Eq. (16) without additional distributional assumptions. The section is titled "Convergence proof" but contains no monotone-likelihood or fixed-point argument. The claim of mutual enhancement between the pseudo-clean matrix and the transition-aware matrix is therefore not theoretically proven.
  5. [Section III-C, Eqs. (11)-(12)] The transition-matrix estimation is circular and rests on an unproven identifiability assumption. The paper explicitly states that T is non-identifiable without extra assumptions and then assumes an optimal theta* such that [T]^{-1} p(S|x; theta*) = p(y|x). T is estimated from the current pseudo-clean labels S, while S is updated using [T]^{-1} of the model's own output. No conditions are given under which this alternation converges to identifiable quantities, and no regularization or constraint other than normalization is imposed. This is a central limitation of the method, not just a presentation issue.
minor comments (4)
  1. [Section III-B, Algorithm 1] In Algorithm 1, after sampling S_t, the line "convert it to a one-hot vector S_0" is ambiguous and inconsistent with the forward equations; presumably S_0 is the pseudo-clean label and S_t is its noised version. Please clarify the pseudocode.
  2. [Section III-C, Eq. (12)] The operations in Eq. (12) mix matrix and vector quantities without defining the element-wise product and normalization; please state the dimensions and the normalization rule explicitly.
  3. [Table II] The entry for DIRK on Lost reads "74.26 ± 0.58" without a percent sign, while all other entries include "%"; unify the formatting across the table.
  4. [Section IV-H] The calibration analysis would benefit from reporting sample sizes and confidence intervals for the ECE, especially because the low-consistency bins are described as having small sample size.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DDMP's contributions are an empirical method and an EM-style update; the T/S feedback loop is iterative self-training, not a definitional reduction, and the core evaluation is against external benchmarks.

full rationale

The paper does not contain a load-bearing circular step in the sense of deriving a result from itself by construction or by self-citation. The forward and reverse diffusion equations (1)-(9) follow the standard diffusion-model template with a pre-trained encoder f_phi; pseudo-clean labels S are initially constructed from instance and candidate-label affinities (Eq. 5), not from the model's own predictions. The transition-aware matrix T is estimated from S in Eq. (11), and S is updated in Eq. (12) using T and the model's reconstructed S_0. This is an iterative EM-style self-training loop: S^{e+1} is a function of S^e, the model output, and T^e, and the paper treats this as progressive refinement, not as an independent derivation of ground truth. The empirical claims are assessed against external PLL baselines on real and synthetic datasets, so the central result is not forced by the method's definitions. The paper itself acknowledges that T is non-identifiable without additional assumptions (Section III-C), which is a real theoretical limitation but not a circularity. Separately, Eq. (9) is algebraically inconsistent with the forward process in Eq. (4), and the stated 85.0%/97.5% win-rate claims are not supported by the raw tables; these are correctness and reporting concerns, outside the circularity definition used here. Self-citations in the related work are not load-bearing. Therefore, the appropriate finding is no significant circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 2 invented entities

The paper introduces two central constructs, S and T, which are updated in a mutually dependent loop. The free parameters are mostly standard diffusion hyperparameters plus the estimated transition matrix. The key axioms are standard PLL assumptions plus an unproven expressiveness condition used to justify identifiability.

free parameters (4)
  • k (nearest neighbor count) = 10
    Used to build the instance adjacency matrix P; chosen by hand, not derived.
  • Diffusion timesteps T and sampling trajectory = T=1000, trajectory=10
    Standard DDPM configuration borrowed from prior work [34]; no justification for these values.
  • Transition-aware matrix T = Estimated via Eq (11), updated each epoch
    QxQ matrix estimated from data; central to label correction but non-identifiable without extra assumptions.
  • Update rule coefficients in Eq (12) = No explicit coefficients; the mixing weight is implicit
    The pseudo-clean label update is a heuristic moving average; no derivation of the mixing factor.
assumptions (5)
  • domain assumption Ground-truth label is always in the candidate set
    Standard PLL assumption stated in Section I and used throughout.
  • domain assumption Conditional independence p(S|x,y)=p(S|y)
    Assumed in Appendix B, Eq (29), to derive the generative perspective.
  • domain assumption Similar instances tend to share the same label (smoothness)
    Used in Section III-B to justify kNN adjacency matrix P.
  • ad hoc to paper The network is expressive enough that [T]^{-1} p(S|x; theta*) equals p(y|x) everywhere
    Introduced in Section III-C to address non-identifiability; unproven.
  • domain assumption Latent distribution is centered on the pretrained encoder output, ST ~ N(f_phi(x), I)
    Modeling choice inherited from [34]; central to forward process.
invented entities (2)
  • Transition-aware matrix T
    purpose: To map true labels to candidate label distributions and correct pseudo-labels via its inverse.
    A model construct estimated from the same data; no external falsifiable prediction.
  • Pseudo-clean label matrix S
    purpose: To provide initial denoised labels for diffusion training and to be iteratively refined.
    Constructed from instance and label similarities; it is a training artifact, not an independently measured quantity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Diffusion Disambiguation Models for Partial Label Learning." pith.science (2026). https://pith.science/paper/7W5A4OCO

@misc{pith2026250700411,
  author       = {Pith},
  title        = {Pith review of: Diffusion Disambiguation Models for Partial Label Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7W5A4OCO}},
  note         = {Machine review of arXiv:2507.00411}
}
read the original abstract

Learning from ambiguous labels is a long-standing problem in practical machine learning applications. The purpose of \emph{partial label learning} (PLL) is to identify the ground-truth label from a set of candidate labels associated with a given instance. Inspired by the remarkable performance of diffusion models in various generation tasks, this paper explores their potential to denoise ambiguous labels through the reverse denoising process. Therefore, this paper reformulates the label disambiguation problem from the perspective of generative models, where labels are generated by iteratively refining initial random guesses. This perspective enables the diffusion model to learn how label information is generated stochastically. By modeling the generation uncertainty, we can use the maximum likelihood estimate of the label for classification inference. However, such ambiguous labels lead to a mismatch between instance and label, which reduces the quality of generated data. To address this issue, this paper proposes a \emph{diffusion disambiguation model for PLL} (DDMP), which first uses the potential complementary information between instances and labels to construct pseudo-clean labels for initial diffusion training. Furthermore, a transition-aware matrix is introduced to estimate the potential ground-truth labels, which are dynamically updated during the diffusion generation. During training, the ground-truth label is progressively refined, improving the classifier. Experiments show the advantage of the DDMP and its suitability for PLL.

Figures

Figures reproduced from arXiv: 2507.00411 by the authors.

Figure 1
Figure 1. An example of partial label learning with noisy labels. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The training procedure of the proposed DDMP approach. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The visualization of adjacency matrix graph. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Convergence results of DDMP on benchmark datasets. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: T-SNE visualization of MNIST dataset during DDMP reverse generation process. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Visual Representation of Disambiguation Results from the DDMP Algorithm. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Reliability diagram and expected calibration error. [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 35 canonical work pages

  1. [1]

    Deep learning from noisy image labels with quality embedding,

    J. Yao, J. Wang, I. W. Tsang, Y . Zhang, J. Sun, C. Zhang, and R. Zhang, “Deep learning from noisy image labels with quality embedding,” IEEE Transactions on Image Processing, vol. 28, no. 4, pp. 1909–1922, 2018. 12

  2. [2]

    Learning from partial labels,

    T. Cour, B. Sapp, and B. Taskar, “Learning from partial labels,” The Journal of Machine Learning Research , vol. 12, pp. 1501–1536, 2011

  3. [3]

    Partial label learning with discrimination augmentation,

    W. Wang and M.-L. Zhang, “Partial label learning with discrimination augmentation,” in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2022, pp. 1920–1928

  4. [4]

    Optimized graph learning using partial tags and multiple features for image and video annotation,

    J. Song, L. Gao, F. Nie, H. T. Shen, Y . Yan, and N. Sebe, “Optimized graph learning using partial tags and multiple features for image and video annotation,” IEEE Transactions on Image Processing , vol. 25, no. 11, pp. 4999–5011, 2016

  5. [5]

    Pseudo label association and prototype- based invariant learning for semi-supervised nir-vis face recognition,

    W. Hu, Y . Yang, and H. Hu, “Pseudo label association and prototype- based invariant learning for semi-supervised nir-vis face recognition,” IEEE Transactions on Image Processing, vol. 33, pp. 1448–1463, 2024

  6. [6]

    Who’s in the picture,

    T. Berg, A. Berg, J. Edwards, and D. Forsyth, “Who’s in the picture,” Advances in neural information processing systems , vol. 17, 2004

  7. [7]

    Learning from candidate labeling sets,

    J. Luo and F. Orabona, “Learning from candidate labeling sets,” Ad- vances in neural information processing systems , vol. 23, 2010

  8. [8]

    Large margin partial label machine,

    J. Chai, I. W. Tsang, and W. Chen, “Large margin partial label machine,” IEEE Transactions on Neural Networks and Learning Systems , vol. 31, no. 7, pp. 2594–2608, 2019

Show all 48 references
  1. [9]

    Deep discriminative cnn with temporal ensembling for ambiguously-labeled image classification,

    Y . Yao, J. Deng, X. Chen, C. Gong, J. Wu, and J. Yang, “Deep discriminative cnn with temporal ensembling for ambiguously-labeled image classification,” in Proceedings of the aaai conference on artificial intelligence, vol. 34, no. 07, 2020, pp. 12 669–12 676

  2. [10]

    Network cooperation with pro- gressive disambiguation for partial label learning,

    Y . Yao, C. Gong, J. Deng, and J. Yang, “Network cooperation with pro- gressive disambiguation for partial label learning,” in Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2020, Ghent, Belgium, September 14–18, 2020, Proceedings, Part I...

  3. [11]

    Partial label learning based on disambiguation correction net with graph representation,

    J. Fan, Y . Yu, Z. Wang, and J. Gu, “Partial label learning based on disambiguation correction net with graph representation,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 8, pp. 4953–4967, 2021

  4. [12]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems , vol. 33, pp. 6840– 6851, 2020

  5. [13]

    Nonlinear regularized reaction-diffusion filters for denoising of images with textures,

    G. Plonka and J. Ma, “Nonlinear regularized reaction-diffusion filters for denoising of images with textures,” IEEE Transactions on Image processing, vol. 17, no. 8, pp. 1283–1294, 2008

  6. [14]

    Card: Classification and regression diffusion models,

    X. Han, H. Zheng, and M. Zhou, “Card: Classification and regression diffusion models,” Advances in Neural Information Processing Systems , vol. 35, pp. 18 100–18 115, 2022

  7. [15]

    A conditional multinomial mixture model for superset label learning,

    L. Liu and T. Dietterich, “A conditional multinomial mixture model for superset label learning,” Advances in neural information processing systems, vol. 25, 2012

  8. [16]

    Partial label learning via feature-aware disambiguation,

    M.-L. Zhang, B.-B. Zhou, and X.-Y . Liu, “Partial label learning via feature-aware disambiguation,” in Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, 2016, pp. 1335–1344

  9. [17]

    A multi-class partial hinge loss for partial label learning,

    J. Fan, Z. Jiang, Y . Xian, and Z. Wang, “A multi-class partial hinge loss for partial label learning,” Applied Intelligence, vol. 53, no. 23, pp. 28 333–28 348, 2023

  10. [18]

    Addressing label ambiguity imbalance in candidate labels: Measures and disambiguation algorithm,

    J. Fan, Y . Yu, and Z. Wang, “Addressing label ambiguity imbalance in candidate labels: Measures and disambiguation algorithm,” Information Sciences, vol. 612, pp. 1–19, 2022

  11. [19]

    Dictionary learning from ambiguously labeled data,

    Y .-C. Chen, V . M. Patel, J. K. Pillai, R. Chellappa, and P. J. Phillips, “Dictionary learning from ambiguously labeled data,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2013, pp. 353–360

  12. [20]

    Partial label learning with competitive learning graph neural network,

    J. Fan, Y . Yu, and Z. Wang, “Partial label learning with competitive learning graph neural network,” Engineering Applications of Artificial Intelligence, vol. 111, p. 104779, 2022

  13. [21]

    Progressive identification of true labels for partial-label learning,

    J. Lv, M. Xu, L. Feng, G. Niu, X. Geng, and M. Sugiyama, “Progressive identification of true labels for partial-label learning,” in international conference on machine learning . PMLR, 2020, pp. 6500–6510

  14. [22]

    Leveraged weighted loss for partial label learning,

    H. Wen, J. Cui, H. Hang, J. Liu, Y . Wang, and Z. Lin, “Leveraged weighted loss for partial label learning,” in International conference on machine learning. PMLR, 2021, pp. 11 091–11 100

  15. [23]

    Graphdpi: Partial label disambiguation by graph representation learning via mutual information maximization,

    J. Fan, Y . Yu, L. Huang, and Z. Wang, “Graphdpi: Partial label disambiguation by graph representation learning via mutual information maximization,” Pattern Recognition, vol. 134, p. 109133, 2023

  16. [24]

    Pico+: Contrastive label disambiguation for robust partial label learning,

    H. Wang, R. Xiao, Y . Li, L. Feng, G. Niu, G. Chen, and J. Zhao, “Pico+: Contrastive label disambiguation for robust partial label learning,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2023

  17. [25]

    Kmt-pll: K-means cross-attention transformer for partial label learning,

    J. Fan, L. Huang, C. Gong, Y . You, M. Gan, and Z. Wang, “Kmt-pll: K-means cross-attention transformer for partial label learning,” IEEE Transactions on Neural Networks and Learning Systems , 2024

  18. [26]

    Instance-dependent partial label learning,

    N. Xu, C. Qiao, X. Geng, and M.-L. Zhang, “Instance-dependent partial label learning,” Advances in Neural Information Processing Systems , vol. 34, pp. 27 119–27 130, 2021

  19. [27]

    Variational label enhancement for instance-dependent partial label learning,

    N. Xu, C. Qiao, Y . Zhao, X. Geng, and M.-L. Zhang, “Variational label enhancement for instance-dependent partial label learning,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2024

  20. [28]

    Ambiguity-induced contrastive learning for instance-dependent partial label learning

    S. Xia, J. Lv, N. Xu, and X. Geng, “Ambiguity-induced contrastive learning for instance-dependent partial label learning.” in IJCAI, 2022, pp. 3615–3621

  21. [29]

    Distilling reliable knowledge for instance-dependent partial label learning,

    D.-D. Wu, D.-B. Wang, and M.-L. Zhang, “Distilling reliable knowledge for instance-dependent partial label learning,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 14, 2024, pp. 15 888–15 896

  22. [30]

    Dif-fusion: Toward high color fidelity in infrared and visible image fusion with diffusion models,

    J. Yue, L. Fang, S. Xia, Y . Deng, and J. Ma, “Dif-fusion: Toward high color fidelity in infrared and visible image fusion with diffusion models,” IEEE Transactions on Image Processing, vol. 32, pp. 5705–5720, 2023

  23. [31]

    Auto-encoding variational bayes,

    D. P. Kingma, “Auto-encoding variational bayes,” arXiv preprint arXiv:1312.6114, 2013

  24. [32]

    Generative adversarial networks,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial networks,” Communications of the ACM , vol. 63, no. 11, pp. 139–144, 2020

  25. [33]

    Denoising diffusion implicit models,

    J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” arXiv preprint arXiv:2010.02502 , 2020

  26. [34]

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

    J. Chen, R. Zhang, T. Yu, R. Sharma, Z. Xu, T. Sun, and C. Chen, “Label-retrieval-augmented diffusion models for learning from noisy labels,” Advances in Neural Information Processing Systems , vol. 36, 2024

  27. [35]

    Diffusevae: Efficient, controllable and high-fidelity generation from low-dimensional latents,

    K. Pandey, A. Mukherjee, P. Rai, and A. Kumar, “Diffusevae: Efficient, controllable and high-fidelity generation from low-dimensional latents,” arXiv preprint arXiv:2201.00308 , 2022

  28. [36]

    Rank-loss support instance machines for miml instance annotation,

    F. Briggs, X. Z. Fern, and R. Raich, “Rank-loss support instance machines for miml instance annotation,” in Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining, 2012, pp. 534–542

  29. [37]

    Learning by associating ambiguously labeled images,

    Z. Zeng, S. Xiao, K. Jia, T.-H. Chan, S. Gao, D. Xu, and Y . Ma, “Learning by associating ambiguously labeled images,” in Proceedings of the IEEE Conference on computer vision and pattern recognition , 2013, pp. 708–715

  30. [38]

    Multiple instance metric learning from automatically labeled bags of faces,

    M. Guillaumin, J. Verbeek, and C. Schmid, “Multiple instance metric learning from automatically labeled bags of faces,” in Computer Vision– ECCV 2010: 11th European Conference on Computer Vision, Heraklion, Crete, Greece, September 5-11, 2010, Proceedings, Part I 11. Springer,...

  31. [39]

    Gradient-based learning applied to document recognition,

    Y . LeCun, L. Bottou, Y . Bengio, and P. Haffner, “Gradient-based learning applied to document recognition,” Proceedings of the IEEE , vol. 86, no. 11, pp. 2278–2324, 1998

  32. [40]

    Deep learning for classical japanese literature,

    T. Clanuwat, M. Bober-Irizar, A. Kitamoto, A. Lamb, K. Yamamoto, and D. Ha, “Deep learning for classical japanese literature,” arXiv preprint arXiv:1812.01718, 2018

  33. [41]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky, G. Hinton et al. , “Learning multiple layers of features from tiny images,” 2009

  34. [42]

    Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms,

    H. Xiao, K. Rasul, and R. V ollgraf, “Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms,” arXiv preprint arXiv:1708.07747, 2017

  35. [43]

    Exploiting class activation value for partial-label learning,

    F. Zhang, L. Feng, B. Han, T. Liu, G. Niu, T. Qin, and M. Sugiyama, “Exploiting class activation value for partial-label learning,” in Interna- tional conference on learning representations , 2021

  36. [44]

    Revisiting consistency reg- ularization for deep partial label learning,

    D.-D. Wu, D.-B. Wang, and M.-L. Zhang, “Revisiting consistency reg- ularization for deep partial label learning,” in International conference on machine learning . PMLR, 2022, pp. 24 212–24 225

  37. [45]

    A simple framework for contrastive learning of visual representations,

    T. Chen, S. Kornblith, M. Norouzi, and G. Hinton, “A simple framework for contrastive learning of visual representations,” in International conference on machine learning . PMLR, 2020, pp. 1597–1607

  38. [46]

    Learning transferable visual models from natural language supervision,

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

  39. [47]

    Using pre-training can improve model robustness and uncertainty,

    D. Hendrycks, K. Lee, and M. Mazeika, “Using pre-training can improve model robustness and uncertainty,” in International conference on machine learning . PMLR, 2019, pp. 2712–2721

  40. [48]

    On calibration of modern neural networks,

    C. Guo, G. Pleiss, Y . Sun, and K. Q. Weinberger, “On calibration of modern neural networks,” in International conference on machine learning. PMLR, 2017, pp. 1321–1330

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.