Pith. sign in

REVIEW 5 major objections 6 minor 2 cited by

SOTAlign: Semi-Supervised Alignment of Unimodal Vision and Language Models via Optimal Transport

T0 review · 5 major / 6 minor · reviewed 2026-08-02 · deepseek-v4-flash

Pith's one-line read Unpaired images and text can replace most of the paired supervision needed to align vision and language encoders.

desk verdict A solid two-stage alignment method whose gains hold when unpaired data match the paired distribution, but the abstract overclaims cross-dataset robustness and the evidence lacks error bars and a random-teacher control. read the letter →

arxiv 2602.23353 v2 pith:YS4AO72Q submitted 2026-02-26 cs.LG cs.AI

classification cs.LGcs.AI MSC 68T0749Q2262H20
keywords semi-supervisedalignmentoptimaltransportvision-languagemodelsPlatonicRepresentationHypothesisfrozenunimodalencodersKLOTdivergenceimage-textretrievalzero-shotclassification
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 claims that aligning frozen pretrained image and text encoders needs far fewer paired examples than previously thought, provided large amounts of unpaired images and texts are available. The proposed two-stage method, SOTAlign, first fits a simple linear map from a few thousand pairs, then uses that map's pairwise similarity structure as a soft target to regularize training of lightweight alignment layers on unpaired data. The key is a new optimal-transport divergence, KLOT, whose gradient the authors compute in closed form, making large-batch training feasible. On retrieval and zero-shot classification benchmarks with 10k pairs and 1M unpaired samples, SOTAlign improves over supervised baselines by roughly 5 to 10 points, and matches a supervised model trained with about four times more pairs. If correct, this means unpaired unimodal data can substitute for most cross-modal supervision, which matters for domains where paired data are scarce.

What carries the argument

The central object is the KLOT divergence between entropic optimal transport plans, KL(OT_{ε*}(K*) || OT_ε(K)), which compares the entire matching structure between a batch of images and texts rather than only their nearest neighbors. OT_ε(K) is the entropic optimal transport plan of the affinity matrix K, a soft permutation that encodes relational geometry. The load-bearing result is Theorem 5.1: the gradient of KLOT with respect to K equals (OT_ε(K) − OT_{ε*}(K*))/ε*, a closed-form expression that avoids unrolling or implicit differentiation of Sinkhorn. This makes it possible to train with batches of 32k samples, which is what lets the method exploit up to 1M unpaired samples.

What would settle it

Take a pair of encoders with mutual k-NN agreement below the range plotted in Figure 9 (about 0.14), or paired data from one domain and unpaired data from a very different domain, and measure the linear teacher's top-1 cross-modal retrieval on a held-out subset of the unpaired distribution; if the teacher performs at chance, K* is uninformative and SOTAlign should collapse to the supervised baseline. Alternatively, train SOTAlign with unpaired data whose spherical sliced Wasserstein distance from the paired set exceeds the range mapped in Figure 5, and test whether the gain over the supervised

Watch

Extended reading notes

Core claim

SOTAlign aligns unimodal encoders in two stages. Stage one fits a linear alignment model (CCA, Procrustes, or contrastive) on the small paired set, recovering a coarse shared geometry that already yields strong zero-shot retrieval. Stage two uses this linear model as a teacher: on each batch of unpaired images and texts, its affinity matrix K* serves as a target geometry, and the alignment layers are trained with a supervised loss on pairs plus a regularizer that keeps the learned geometry close to K*. The regularizer is the KLOT divergence, the KL divergence between entropic optimal transport plans KL(OT_{ε*}(K*) || OT_ε(K)). The main theoretical result is an explicit gradient, ∇_K KLOT = (

Load-bearing premise

The load-bearing premise is that the linear map fit on a few thousand paired samples generalizes to arbitrary unpaired data, so that the teacher's predicted affinity on unpaired batches is semantically meaningful; if the relationship between the frozen encoders stops being approximately linear outside the paired distribution, the KLOT regularizer will amplify the teacher's errors.

Editorial extensions

If this is right

  • If the linear teacher captures the cross-modal geometry of frozen encoders, then unpaired images and texts can refine alignment without any additional paired supervision; the paper shows consistent gains up to 1M unpaired samples, with diminishing returns after about 500k.
  • The closed-form gradient for KLOT applies to any OT-based loss of the form KL(T || OT(K)), removing a scalability barrier for OT-based representation learning; the paper explicitly notes this generality for recent OT-based alignment and contrastive methods.
  • The method is robust to distribution shift: mixing unpaired images from one corpus and unpaired text from another still improves COCO and ImageNet performance, and the paper's spherical sliced Wasserstein distance predicts the size of the gain before training.
  • The paired-data requirement drops by roughly a factor of four: 10k pairs plus unpaired data match a supervised model trained on about 40k pairs, and the largest gains appear in the intermediate regime between 1k and 10k pairs.

Reading between the lines

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

  • A natural untested extension is iterative self-training: use the refined alignment as the new teacher and repeat, which would test whether the linearity of the initial teacher is a hard ceiling or just a starting point.
  • If the underlying hypothesis about convergent representation geometry holds beyond vision and language, the same two-stage recipe should work for other modality pairs with large unpaired corpora, such as audio-text or molecule-text; the paper only validates vision-language.
  • The observed link between sliced Wasserstein distance and gain suggests a practical selection rule: measure candidate unpaired corpora before training and only add data below a distance threshold, a heuristic the paper does not state explicitly.
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 / 6 minor

Summary. The paper proposes SOTAlign, a two-stage semi-supervised method for aligning frozen unimodal vision and language encoders. In stage one, a linear alignment map (Procrustes, CCA, or contrastive) is fit on a small set of paired image–text samples. In stage two, lightweight alignment layers f and g are trained with a supervised SigLIP loss on the same pairs plus an optimal-transport-based regularization (KLOT) that encourages the affinity structure of unpaired batches to match the structure induced by the linear teacher. The authors report consistent gains over supervised and semi-supervised baselines in zero-shot retrieval and classification when unpaired data are drawn from the same distribution as the paired data (e.g., CC3M), and claim robustness across datasets and encoder pairs. They also derive an explicit gradient expression for KLOT, removing the need to backpropagate through Sinkhorn iterations.

Significance. If the claims are sustained, the paper makes a useful contribution: it demonstrates that in-distribution unpaired unimodal data can substitute for a large fraction of paired supervision when aligning frozen encoders, and it provides a scalable OT-based regularizer with an explicit gradient. The code release and the careful handling of the Sinkhorn backpropagation memory bottleneck are strengths. However, the central empirical claim is currently broader than the evidence: the method's own cross-domain results show that the linear teacher's target geometry degrades under distribution shift, and the absence of error bars and of a control for the teacher's semantic content prevents the reader from attributing the gains to the proposed mechanism.

major comments (5)
  1. [Theorem 5.1 / Appendix C.2] The main-text gradient formula contradicts the appendix. Eq. (13) states ∇_K KLOT(K||K*) = (OT_ε(K) − OT_{ε*}(K*))/ε*, but KLOT is defined as KL(OT_{ε*}(K*) || OT_ε(K)). With T = OT_{ε*}(K*) fixed, Theorem C.4 (Eq. 28) gives ∇_K KL(T || OT_ε(K)) = (OT_ε(K) − T)/ε, i.e., the denominator should be ε, not ε*. Since the explicit gradient is a central claim and is used in Algorithm 1 to update the alignment layers, this discrepancy must be corrected and the implementation checked against the correct formula.
  2. [Table 7 / Section 6.2] The text states that SOTAlign 'consistently outperforms the supervised baseline' across heterogeneous unpaired sources, but Table 7 contradicts this. With CC3M pairs and ImageNet images + WikiText103 unpaired text, COCO T2I R@1 is 20.7 versus 21.0 for supervised SAIL, and I2T is 28.1 versus 27.4; with COCO images + WikiText103, T2I drops to 19.5. Thus the method does not leverage unpaired data under a substantial distribution shift. The abstract's claim of 'robust joint embeddings across datasets' and the 'consistently outperforms' wording in Section 6.2 need to be qualified to the in-distribution or mild-shift regime.
  3. [Algorithm 1 / Eq. (4)-(5)] The mechanism by which unpaired data improve performance is not isolated. The loss always contains the supervised SigLIP term on the same 10k pairs, and the KLOT term adds an extra training signal on unpaired batches. Without an ablation that replaces the teacher target K* with a random, shuffled, or identity affinity matrix on unpaired batches, the gains could in principle come from the OT regularizer's smoothing effect or from the linear teacher's initialization rather than from the semantic validity of K*. The manuscript should report such a control. Additionally, Algorithm 1 line 2 ('Initialize encoders f and g') does not specify whether the linear projections Wx, Wy are used as initialization; if they are, the comparison to SAIL (which starts from random init) is confounded.
  4. [Tables 1, 4, 5, 7] All main results are single numbers with no standard deviations or number of seeds. In the 10k-pair regime, random batch composition and optimization variability can easily account for 1–2 R@1 points, and several claimed gains are in that range (e.g., SOTAlign with ImageNet+WikiText vs SAIL, 20.7 vs 21.0). The paper should report mean ± std over at least three seeds for the central comparisons, and ideally for the hyperparameter-selected configuration.
  5. [Figure 9 / Appendix B.2] The support for the linear teacher's transferability is weak. Pearson r = 0.83 is computed over only 6 encoder pairs, and the text itself concedes that 'additional data points will be required to draw firm conclusions.' This does not establish that the linear map produces a semantically meaningful target K* on arbitrary unpaired batches; indeed, Table 7 shows it does not under distribution shift. The paper should either provide direct evidence of K* quality on held-out unpaired data (e.g., retrieval or k-NN agreement on unpaired batches) or substantially soften the claims about robustness.
minor comments (6)
  1. [Algorithm 1] Typo: 'LinearAlignement' should be 'LinearAlignment'.
  2. [Section 3.3] Typo: 'we fit a a simple linear alignment model' should read 'we fit a simple linear alignment model'.
  3. [Appendix C.2] Typo: 'permutatin' should be 'permutation' in the Monge formulation paragraph.
  4. [Table 4] The benchmark name 'Flickr30' should be 'Flickr30k' for consistency with the text and references.
  5. [Section 6 / Appendix A.1] The hyperparameters α, ε, ε*, and the CCA regularization λ are all prescribed as fixed values or selected on the CC3M validation set, but the selected α values for each experiment are not reported. The sensitivity statement 'comparable results can be obtained with alternative settings' would be more convincing with a small table or a figure showing performance versus α.
  6. [References] The reference list contains a formatting artifact: 'V ouitsis' should be 'Vouitsis'. Also, the arXiv identifier in the manuscript header 'arXiv:2602.23353v1' is not a standard format for the submission venue; this should be cleaned up.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SOTAlign's semi-supervised target is a pseudo-label from a linear teacher, but the central claims are evaluated on held-out benchmarks and the gradient result is a theorem.

full rationale

The derivation chain is not circular. The linear teacher (Eqs. 6-8) is fit on the small paired set and then used only to define the pseudo-label affinity K* on unpaired batches (Algorithm 1 line 6; Eq. 4). This is standard self-training/pseudo-labeling, not a prediction forced by definition: the reported gains are measured on held-out COCO/Flickr30k/ImageNet, which are not used to construct K*. If the linear-transfer premise fails, Table 7 shows performance degrades under distribution shift, which is exactly what an independent empirical assumption should do. The KLOT gradient (Theorem 5.1, Appendix C.2) is derived from entropic OT duality and the envelope theorem, not from fitted constants. The only self-citations are to the POT software library in implementation details and to an unrelated graph-matching paper in a list of Sinkhorn applications; neither is load-bearing for the central claim. The main limitations are empirical (small number of encoder pairs in the r=0.83 correlation; degradation under shift), not circular reasoning. The central results therefore have independent content and are not entailed by the paper's own inputs.

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

The method introduces no new physical entities. It relies on hand-set entropic regularization parameters and a validation-tuned alpha. The main load-bearing axiom is that linear maps fit on a small paired set transfer to large unpaired batches, i.e., the Platonic Representation Hypothesis as a working assumption.

free parameters (5)
  • alpha (KLOT regularization weight) = selected from {1e-3, 1e-4, 1e-5} based on CC3M validation retrieval
    Controls the strength of the unpaired-data regularization in Eq. 5; chosen by validation performance, not fixed a priori.
  • epsilon (joint-space entropic regularization) = 0.05
    Hand-set smoothness of the Sinkhorn plan OT_epsilon(K); affects gradient scale and matching softness (Appendix A.1).
  • epsilon* (reference-space entropic regularization) = 0.01
    Hand-set for the teacher plan OT_epsilon*(K*); appears as the denominator in Theorem 5.1 (with an inconsistency noted in red flags).
  • CCA regularization lambda = 0.1
    Added to eigenvalues before inversion in CCA (Appendix A.1); stabilizes the closed-form solution.
  • Sinkhorn iterations = 100
    Number of iterations in both spaces; convergence is not rigorously checked in the main experiments.
assumptions (5)
  • domain assumption Platonic Representation Hypothesis / approximate linear compatibility of frozen unimodal encoders
    Used to justify that a linear teacher fit on a few pairs transfers to large unpaired batches (Sections 1 and 3.3).
  • standard math Entropic OT dual representation: log OT_epsilon(K) = u 1^T + K/epsilon + 1 v^T
    Used in Lemma C.3 and Theorem C.4 for the gradient derivation; standard result from Peyre et al. (2019).
  • standard math Envelope theorem applies to the strongly convex entropic OT problem
    Used to compute grad_K W_epsilon(K) = -OT_epsilon(K) in Theorem C.4.
  • domain assumption Cosine affinity between batch embeddings captures semantic geometry
    The whole method compares affinity matrices built from cosine similarity; assumes cosine geometry is the right representation space for alignment.
  • domain assumption Sinkhorn algorithm converges to OT_epsilon(K) within 100 iterations
    The implementation fixes iteration count rather than a convergence tolerance; this is a computational assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SOTAlign: Semi-Supervised Alignment of Unimodal Vision and Language Models via Optimal Transport." pith.science (2026). https://pith.science/paper/YS4AO72Q

@misc{pith2026260223353,
  author       = {Pith},
  title        = {Pith review of: SOTAlign: Semi-Supervised Alignment of Unimodal Vision and Language Models via Optimal Transport},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YS4AO72Q}},
  note         = {Machine review of arXiv:2602.23353}
}
read the original abstract

The Platonic Representation Hypothesis posits that neural networks trained on different modalities converge toward a shared statistical model of the world. Recent work exploits this convergence by aligning frozen pretrained vision and language models with lightweight alignment layers, but typically relies on contrastive losses and millions of paired samples. In this work, we ask whether meaningful alignment can be achieved with substantially less supervision. We introduce a semi-supervised setting in which pretrained unimodal encoders are aligned using a small number of image-text pairs together with large amounts of unpaired data. To address this challenge, we propose SOTAlign, a two-stage framework that first recovers a coarse shared geometry from limited paired data using a linear teacher, and then refines the alignment on unpaired samples via an optimal-transport-based divergence that transfers relational structure without overconstraining the target space. SOTAlign effectively leverages unpaired images and text, learning robust joint embeddings across datasets and encoder pairs, and significantly outperforming supervised and semi-supervised baselines. Code is available at https://github.com/ExplainableML/SOTAlign.

Figures

Figures reproduced from arXiv: 2602.23353 by the authors.

Figure 1
Figure 1. Semi-Supervised Vision-Language Alignment. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. SOTAlign is a two-step method for the alignment of pretrained unimodal image and text encoders. First, we fit a linear alignment model only using the limited amount of available image-text pairs. Then, we use this linear model as a teacher to regularize the training of alignment layers f and g for a joint embedding space leveraging unimodal (unpaired) data. 4. Linear Alignment Model The first core component of the p… view at source ↗
Figure 3
Figure 3. GPU memory usage for a batchsize n = 10k when computing the gradient of the OT-based divergence with naive solver unrolling (blue) and the provided explicit gradient formula (orange). Additionnal results are reported in appendix B.1. Theorem 5.1. For any transport plan P ∈ Πn, ∇K KLOT(K || K∗ ) = OTϵ(K) − OTϵ ∗ (K∗ ) ϵ ∗ . (13) Proof is provided in Appendix C.2. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p… view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Left: Effect of the number of paired samples (while fixing 1M unpaired samples). Right: Effect of the number of unpaired samples (while fixing 10k pairs). We report the zero-shot retrieval (MeanR@1) on COCO. More metrics are reported in Appendix B. 0.00 0.01 0.02 0.03 …
Figure 5
Figure 5. Figure 5: Relationship between the total sliced Wasserstein dis￾tance between CC3M image/text dataset and unimodal datasets, and the downstream performance of SOTAlign trained on 10k CC3M image–text pairs and up to 1M samples from the corre￾sponding unimodal datasets. models, DI…
Figure 6
Figure 6. Figure 6: Comparison of memory usage, runtime, and number of Sinkhorn iterations for different gradient computation strategies. We run as many Sinkhorn iterations as required to achieve marginal convergence within a tolerance of 10−6 . B. Additional Experiments B.1. Sinkhorn Bac…
Figure 7
Figure 7. Figure 7: Effect of the number of paired samples during alignment on downstream zero-shot classification and retrieval. We fix 1M unpaired samples from CC3M and vary the number of paired samples. B.2. Robustness of SOTAlign In Section 6.2, we analyze the robustness of SOTAlign t…
Figure 8
Figure 8. Figure 8: Effect of the number of unpaired samples during alignment on downstream zero-shot classification and retrieval. We fix 10k paired samples from CC3M and vary the number of unpaired samples. 0.14 0.15 0.16 0.17 0.18 0.19 Mutual k-NN 20 22 24 26 28 30 Mean R@1 Pearson r =…
Figure 9
Figure 9. Figure 9: (R@1COCO) vs mutual k-NN. 0.000 0.005 0.010 0.015 0.020 0.025 Sliced Wasserstein distance to CC3M 24 26 28 30 32 34 COCO Mean R@1 Pearson r = -0.89 WikiText103 CC3M CC12M CC3M-s SOTAlign (Ours) Supervised SAIL 0.000 0.005 0.010 0.015 0.020 0.025 0.030 Sliced Wasserstei…
Figure 10
Figure 10. Figure 10: Performance when using CC3M as paired data, CC3M text as unpaired text, and other image datasets as unpaired images, together with a comparison to the spherical sliced Wasserstein distance between CC3M image and the other image datasets. 16 [PITH_FULL_IMAGE:figures/f…
Figure 11
Figure 11. Figure 11: Performance when using CC3M as paired data, CC3M text as unpaired text, and other image datasets as unpaired images, together with a comparison to the spherical sliced Wasserstein distance between CC3M image and the other image datasets. CC3M CC12M COCO Imagenet CC3M …
Figure 12
Figure 12. Figure 12: Spherical sliced Wasserstein distances between different image datasets (left) and text datasets (right). We report mean and std of the distances over 20 seeds. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_12.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Topology-Aware Representation Alignment for Semi-Supervised Vision-Language Learning

    cs.CV 2026-04 unverdicted novelty 6.0 of 10

    ToMA uses persistent homology on H0-death and lightweight H1-birth edges to align multimodal manifolds, delivering stable gains on remote sensing and consistent benefits on fashion retrieval.

  2. MSAlign: Aligning Molecule and Mass Spectra Foundation Models for Metabolite Identification

    cs.LG 2026-05 conditional novelty 5.0 of 10

    MSAlign aligns frozen DreaMS and ChemBERTa models with MLPs and candidate-based contrastive learning to outperform prior methods on molecule retrieval from MS/MS spectra while quantifying distribution shift in data splits.

Reference graph

Works this paper leans on

27 extracted references · 9 linked inside Pith · cited by 2 Pith papers

  1. [1]

    Adams, R. P. and Zemel, R. S. Ranking via sinkhorn propa- gation.arXiv preprint arXiv:1106.1925,

  2. [6]

    Our approach significantly outperforms both alternatives in terms of memory usage and runtime. In particular, depending on the value of ϵ, which controls the number of Sinkhorn iterations (with convergence scaling as O(1/ϵ2)), the proposed method can be up to 100× more memory efficient than unrolling and up to50×faster than implicit differentiation. 14 Se...

  3. [8]

    Jha, R., Zhang, C., Shmatikov, V ., and Morris, J. X. Har- nessing the universal geometry of embeddings.arXiv preprint arXiv:2505.12540,

  4. [11]

    Fine-grained visual classification of aircraft.arXiv preprint arXiv:1306.5151,

    Maji, S., Rahtu, E., Kannala, J., Blaschko, M., and Vedaldi, A. Fine-grained visual classification of aircraft.arXiv preprint arXiv:1306.5151,

  5. [13]

    Sim´eoni, O., V o, H. V ., Seitzer, M., Baldassarre, F., Oquab, M., Jose, C., Khalidov, V ., Szafraniec, M., Yi, S., Rama- monjisoa, M., Massa, F., Haziza, D., Wehrstedt, L., Wang, J., Darcet, T., Moutakanni, T., Sentana, L., Roberts, C., Vedaldi, A., Tolan, J., Brandt, J., Couprie, C., Mairal, J., J´egou, H., Labatut, P., and Bojanowski, P. Dinov3.arXiv ...

  6. [14]

    F., Alabdulmohsin, I., Parthasarathy, N., Evans, T., Beyer, L., Xia, Y ., Mustafa, B., H ´enaff, O., Harmsen, J., Steiner, A., and Zhai, X

    Tschannen, M., Gritsenko, A., Wang, X., Naeem, M. F., Alabdulmohsin, I., Parthasarathy, N., Evans, T., Beyer, L., Xia, Y ., Mustafa, B., H ´enaff, O., Harmsen, J., Steiner, A., and Zhai, X. SigLIP 2: Multilingual vision- language encoders with improved semantic understand- ing, localization, and dense features.arXiv preprint arXiv:2502.14786,

  7. [16]

    Experimental Setting We outline the experimental setup in Section 6.3

    11 Semi-Supervised Alignment of Unimodal Vision and Language Models via Optimal Transport A. Experimental Setting We outline the experimental setup in Section 6.3. Here, we provide further details on our implementation and baselines. A.1. Implementation Details Following Zhang et al. (2025a), we create global image representations by concatenating the [CL...

  8. [17]

    We mainly employ DINOv3 ViT-L (Sim´eoni et al.,

    with a cosine annealing learning-rate schedule, a maximum learning rate of10 −4, and a weight decay of10 −5, and train for 2000 iterations. We mainly employ DINOv3 ViT-L (Sim´eoni et al.,

Show all 27 references
  1. [18]

    as the pretrained vision and language encoders, respectively. In Section 6.2, we additionally evaluate SOTAlign with DINOv2 ViT-L (Oquab et al., 2024), Qwen3-Embedding-8B (Zhang et al., 2025b), and Llama-Embed-Nemotron-8B (Babakhin et al., 2025). All of these language models a...

  2. [19]

    loss exclusively on paired data. This method represents a series of recent supervised contrastive methods for the alignment of pretrained unimodal vision and language models (V ouitsis et al., 2024; Maniparambil et al., 2025; Huang et al., 2025). Following Zhang et al. (2025a)...

  3. [20]

    We further experiment with disjoint images and texts from CC12M (Changpinyo et al., 2021), COCO (Lin et al., 2014), ImageNet (Deng et al., 2009), and WikiText (Merity et al., 2016)

    with both raw web captions and synthetic captions generated by DreamLIP (Zheng et al., 2024). We further experiment with disjoint images and texts from CC12M (Changpinyo et al., 2021), COCO (Lin et al., 2014), ImageNet (Deng et al., 2009), and WikiText (Merity et al., 2016). W...

  4. [21]

    • Retrieval (T2I, I2T): COCO (Lin et al., 2014), Flickr30 (Plummer et al.,

  5. [24]

    Our method achieves gains of +5.5 on COCO, +28.7 on Flickr30k, and +18.2 on Polyvore I2T R@5

    In Table 11, we report the performance of SOTAlign adhering to their architectural choices. Our method achieves gains of +5.5 on COCO, +28.7 on Flickr30k, and +18.2 on Polyvore I2T R@5. However, if we lift these constraints and instead use linear alignment layers with a target...

  6. [25]

    and without these constraints. COCO Flickr30k Polyvore 100 Pairs 500 Pairs 500 Pairs MethodI2T T2I I2T T2I I2T T2I CSA 1.3 1.0 1.3 0.8 1.3 1.0 Contrastive 8.5 5.8 9.5 9.8 13.8 11.5 SUE 21.5 18.3 19.8 22.0 22.8 20.8 SOTAlign(with SUE constraints) 27.0 28.8 48.5 48.8 41.0 39.8 S...

  7. [27]

    replaces linear projections with angular projections and computes optimal transport on the circle, thereby respecting the intrinsic geometry of directional data. Total sliced Wasserstein distanceWe introduce the total spherical sliced Wasserstein distance d as a measure of the...

  8. [2008]

    Oord, A. v. d., Li, Y ., and Vinyals, O. Representation learn- ing with contrastive predictive coding.arXiv preprint arXiv:1807.03748,

  9. [2009]

    The quest for the graph level autoencoder (grale).arXiv preprint arXiv:2505.22109,

    Krzakala, P., Melo, G., Laclau, C., d’Alch ´e Buc, F., and Flamary, R. The quest for the graph level autoencoder (grale).arXiv preprint arXiv:2505.22109,

  10. [2014]

    D., Shi, K., Shahbazi, A., Land- man, B

    Liu, X., Bai, Y ., Mart´ın, R. D., Shi, K., Shahbazi, A., Land- man, B. A., Chang, C., and Kolouri, S. Linear spherical sliced optimal transport: A fast metric for comparing spherical data.arXiv preprint arXiv:2411.06055,

  11. [2015]

    We run as many Sinkhorn iterations as required to achieve marginal convergence within a tolerance of10 −6

    13 Semi-Supervised Alignment of Unimodal Vision and Language Models via Optimal Transport Figure 6.Comparison of memory usage, runtime, and number of Sinkhorn iterations for different gradient computation strategies. We run as many Sinkhorn iterations as required to achieve ma...

  12. [2017]

    Huang, W., Wu, A., Yang, Y ., Luo, X., Yang, Y ., Hu, L., Dai, Q., Wang, C., Dai, X., Chen, D., Luo, C., and Qiu, L

    ISBN 9781450349062. Huang, W., Wu, A., Yang, Y ., Luo, X., Yang, Y ., Hu, L., Dai, Q., Wang, C., Dai, X., Chen, D., Luo, C., and Qiu, L. Llm2clip: Powerful language model unlocks richer visual representation.arXiv preprint arXiv:2411.04997,

  13. [2018]

    Enevoldsen, K., Chung, I., Kerboua, I., Kardos, M., Mathur, A., Stap, D., Gala, J., Siblini, W., Krzemi´nski, D., Winata, G. I., Sturua, S., Utpala, S., Ciancone, M., Schaeffer, M., Sequeira, G., Misra, D., Dhakal, S., Rystrøm, J., Solomatin, R., ¨Omer C ¸a˘gatan, Kundu, A., B...

  14. [2019]

    for more details. Letn∈Nand denote byP n the set of permutation matrices, Pn ={P∈ {0,1} n×n |P1=1, P ⊤1=1},(Permutations Matrices) and byΠ n the set of bistochastic matrices, Πn ={T∈R n×n + |T1=1, T ⊤1=1}.(Transport Plans) We further define the (negative) entropy of a transpor...

  15. [2020]

    Reliability of cka as a similarity measure in deep learning.arXiv preprint arXiv:2210.16156,

    Davari, M., Horoi, S., Natik, A., Lajoie, G., Wolf, G., and Belilovsky, E. Reliability of cka as a similarity measure in deep learning.arXiv preprint arXiv:2210.16156,

  16. [2022]

    and Ranka, S

    Emami, P. and Ranka, S. Learning permutations with sinkhorn policy gradient.arXiv preprint arXiv:1805.07010,

  17. [2023]

    Assessing and learn- ing alignment of unimodal vision and language models

    Zhang, L., Yang, Q., and Agrawal, A. Assessing and learn- ing alignment of unimodal vision and language models. InCVPR, pp. 14604–14614, 2025a. Zhang, Y ., Li, M., Long, D., Zhang, X., Lin, H., Yang, B., Xie, P., Yang, A., Liu, D., Lin, J., et al. Qwen3 embed- ding: Advancing ...

  18. [2024]

    Babakhin, Y ., Osmulski, R., Ak, R., Moreira, G., Xu, M., Schifferer, B., Liu, B., and Oldridge, E

    URL https://huguesva.github.io/blog/2024/ inverseOT_mongegap/. Babakhin, Y ., Osmulski, R., Ak, R., Moreira, G., Xu, M., Schifferer, B., Liu, B., and Oldridge, E. Llama- embed-nemotron-8b: A universal text embedding model for multilingual and cross-lingual tasks.arXiv preprint...

  19. [2025]

    Wasserstein discriminant analysis.Machine Learning, 107(12):1923–1945,

    Flamary, R., Cuturi, M., Courty, N., and Rakotomamonjy, A. Wasserstein discriminant analysis.Machine Learning, 107(12):1923–1945,

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.