Paired-Sampling Contrastive Framework

Updated 9 July 2026

Paired-sampling contrastive framework is a design where explicit pairing rules, rather than random augmentations, drive the construction of positive and negative sample sets.
It employs diverse mechanisms—such as annotated pairs, log-mined pairs, and graph-defined relations—to tailor contrast geometry for enhanced representation learning.
The approach improves model robustness and transferability by aligning latent features with task-relevant structures, though its success hinges on the quality of the paired samples.

Searching arXiv for the cited papers to ground the article and verify the framing. arXiv search: PairCFR / paired-sampling contrastive framework related work. A paired-sampling contrastive framework is a contrastive-learning design in which supervision depends not only on labels or augmentations, but on an explicit or induced pairing mechanism that determines which samples, sub-samples, or sample groups should co-occur during training and how they should function as positives or negatives. In the literature, this idea appears in multiple forms: explicit original–counterfactual co-batching in counterfactually augmented data, log-derived query/document or user-history pairs in personalized search, frame selection within aligned video–text pairs, graph-constructed positive and negative relations for feature extraction, weakly paired group-level multimodal supervision, and hybrid objectives that distinguish same-source paired views from broader same-label context (Qiu et al., 2024, Zhou et al., 2021, Wang et al., 2022, Zhang, 2021, Gorla et al., 3 Feb 2026, Deng et al., 1 Dec 2025). Across these formulations, the central issue is the same: the choice of paired samples governs what invariances are learned, what distinctions are preserved, and whether the model exploits meaningful structure or collapses onto shortcuts (Desai et al., 12 Feb 2025).

1. Concept and scope

A paired-sampling contrastive framework can be defined by three elements. First, it specifies an anchor and a mechanism for constructing one or more related samples. Second, it assigns those related samples a role in the contrastive objective, either as positives, negatives, or structured subsets. Third, it uses that pairing rule to shape representation geometry more deliberately than generic in-batch sampling.

The literature shows that “pairing” is broader than instance augmentation. In PairCFR, the fundamental pair is an original example $\mathbf{x}$ and its counterfactual $\mathbf{c}$ , linked by annotation in Counterfactually Augmented Data (CAD) (Qiu et al., 2024). In PSSL, the pairs are mined from query logs at several granularities: document pairs, query pairs, augmented user-history pairs, and cross-user pairs under ambiguous queries (Zhou et al., 2021). In FineCo, the global video–text pair is treated as internally heterogeneous, and the paired text is used to split video frames into relevant and irrelevant subsets within the same positive pair (Wang et al., 2022). In GROOVE, there are no exact instance-level correspondences; instead, samples are weakly paired through shared perturbation labels, and all same-label cross-modal samples form the positive set for a cross-modal anchor (Gorla et al., 3 Feb 2026).

This diversity of formulations implies that a paired-sampling contrastive framework is not a single algorithmic family but a design principle. The framework may be supervised, self-supervised, weakly supervised, or multimodal. It may rely on explicit pair annotations, induced pseudo-pairs, graph-defined relations, or adaptive pairing rules. What unifies these methods is that pair construction is a first-class modeling choice rather than a peripheral implementation detail (Desai et al., 12 Feb 2025).

2. Pair construction mechanisms

The literature distinguishes several recurrent pairing regimes.

Pairing regime	Representative mechanism	Representative paper
Explicit annotated pairs	original–counterfactual co-batching	PairCFR (Qiu et al., 2024)
Log-mined pairs	query, document, sequence, and user pair mining	PSSL (Zhou et al., 2021)
Within-pair sub-sampling	text-conditioned frame selection inside a video–text pair	FineCo (Wang et al., 2022)
Graph-defined relations	positive and negative graphs over all samples	CL-UFEF (Zhang, 2021)
Weak group pairing	cross-modal same-label group positives	GROOVE / GroupCLIP (Gorla et al., 3 Feb 2026)
Hybrid instance/class pairing	same-source paired views plus same-label context pairs	ConTeX (Deng et al., 1 Dec 2025)

In PairCFR, paired sampling is implemented through PairCAD, where each original sentence and its counterfactual are explicitly placed in the same mini-batch, as opposed to ShuffCAD, where originals and counterfactuals are randomly shuffled (Qiu et al., 2024). The pair is not created by augmentation at runtime; it is inherited from CAD annotation. This gives the framework access to a locally visible counterfactual relation during contrastive training.

In PSSL, pairing is mined from behavioral regularities in query logs. A document pair is formed when a user clicks two documents under the same query; a query pair is formed when the same user clicks the same document under two different queries; a sequence augmentation pair is formed by two random perturbations of a single user history; and a user pair is formed when two users issue the same ambiguous query and click the same document (Zhou et al., 2021). Here the pairing mechanism is not label-based in the ordinary classification sense, but behavior-based and query-conditioned.

FineCo introduces a different notion of paired sampling. The positive unit is a paired video–text segment, but only a subset of frames are semantically relevant to the text. The text is therefore used to score all frames, rank them, and split them into $\mathcal P(x)$ and $\mathcal N(x)$ , with top- $C$ frames treated as positive sub-instances and the remainder as within-video negatives (Wang et al., 2022). This suggests that in a paired-sampling framework, the pair itself may contain latent internal positives and negatives.

Graph-based feature extraction provides a deterministic alternative to minibatch sampling. CL-UFEF constructs a positive graph $G^{pos}=\{X,S^{pos}\}$ and a negative graph $G^{neg}=\{X,S^{neg}\}$ , where the nonzero entries of $S^{pos}$ and $S^{neg}$ define all active pair relations globally (Zhang, 2021). In the unsupervised variant, positives are $k$ -nearest neighbors and negatives are non-neighbors; in the supervised variants, positives are either all same-class samples or same-class $\mathbf{c}$ 0-nearest neighbors, with negatives defined by the complement (Zhang, 2021).

Weak pairing extends the same principle to multimodal data without one-to-one correspondence. GROOVE assumes only shared perturbation labels across modalities. For an anchor $\mathbf{c}$ 1, the positive set is

$\mathbf{c}$ 2

and the contrast is always cross-modal (Gorla et al., 3 Feb 2026). This transforms label co-membership into a many-to-many pairing rule.

ConTeX explicitly separates two forms of pair construction: the self positive $\mathbf{c}$ 3, which is the other augmentation from the same original image, and the context positive set $\mathbf{c}$ 4, consisting of different images with the same label (Deng et al., 1 Dec 2025). This decomposition is motivated by the claim that ordinary supervised contrastive learning can over-rely on same-label structure while underemphasizing the paired augmentations from the same source image (Deng et al., 1 Dec 2025).

3. Positive and negative geometry

The most important distinction across paired-sampling contrastive frameworks is not whether pairs exist, but how the paired relation is mapped into the contrastive geometry.

PairCFR is the clearest example of a nonstandard choice. Although it is built around explicit original–counterfactual pairs, those pairs are not used as positives. If the original is $\mathbf{c}$ 5 and the counterfactual is $\mathbf{c}$ 6, then the counterfactual belongs to the anchor’s negative set because it has the opposite label (Qiu et al., 2024). The framework defines

$\mathbf{c}$ 7

and $\mathbf{c}$ 8 by construction (Qiu et al., 2024). This is the defining twist of PairCFR: the paired sample is a structured hard negative rather than a positive.

PSSL adopts the more conventional anchor–positive–negative arrangement. For document pairing, the anchor is one clicked document and the positive is another clicked document under the same user-query context; for query pairing, the anchor is one query and the positive is another query from the same user that led to the same clicked document; for sequence augmentation pairing, the positive is the alternate augmented view of the same history; and for user pairing, the positive is another user with the same $\mathbf{c}$ 9 behavior under an ambiguous query (Zhou et al., 2021). In all four tasks, negatives are simply other instances in the minibatch (Zhou et al., 2021).

FineCo again differs by constructing negatives inside a globally positive pair. For batch index $\mathcal P(x)$ 0, it defines

$\mathcal P(x)$ 1

and optimizes

$\mathcal P(x)$ 2

in combination with a pair-level retrieval or QA loss (Wang et al., 2022). The negative set is therefore not primarily external distractors from other pairs, but semantically weak members of the same aligned pair.

Weakly paired GroupCLIP uses multiple cross-modal positives per anchor. Its per-anchor loss is

$\mathcal P(x)$ 3

where $\mathcal P(x)$ 4 contains all opposite-modality candidates (Gorla et al., 3 Feb 2026). This makes the positive set a label-conditioned cross-modal group rather than a single paired counterpart.

ConTeX formalizes a dual geometry. Its class-sensitive component contrasts same-label samples only against different-label samples,

$\mathcal P(x)$ 5

while its self-paired component explicitly enforces closeness of the two augmentations from the same image,

$\mathcal P(x)$ 6

(Deng et al., 1 Dec 2025). This separation is intended to prevent same-label context positives from overwhelming same-source paired views (Deng et al., 1 Dec 2025).

A broader theoretical implication is that positive-pair semantics need not be fixed globally. Hydra, introduced in “Rethinking Positive Pairs in Contrastive Learning,” allows arbitrary class pairs to serve as positives, but only in a learned pair-conditioned subspace defined by a gate vector $\mathcal P(x)$ 7 (Wu et al., 2024). This suggests that a paired-sampling framework can decouple the existence of a pair from the geometry in which the pair is judged similar.

4. Objectives, theory, and representations

The mathematical forms used in paired-sampling contrastive frameworks vary, but most remain variations of InfoNCE-style or supervised contrastive objectives.

PairCFR combines cross-entropy and supervised contrastive regularization,

$\mathcal P(x)$ 8

with sentence embeddings $\mathcal P(x)$ 9 taken from the transformer $\mathcal N(x)$ 0 representation (Qiu et al., 2024). Its theoretical analysis begins with a toy binary classifier on $\mathcal N(x)$ 1 and counterfactual $\mathcal N(x)$ 2, arguing that CAD plus cross-entropy alone can overemphasize the revised feature $\mathcal N(x)$ 3. The derived gradient for the negative branch of the contrastive loss involves

$\mathcal N(x)$ 4

and the paper interprets this as evidence that contrastive learning reflects the entire feature space rather than only edited dimensions (Qiu et al., 2024). This is presented as the mathematical basis for the claim that contrastive regularization promotes broader feature usage.

PSSL uses four auxiliary contrastive losses— $\mathcal N(x)$ 5, $\mathcal N(x)$ 6, $\mathcal N(x)$ 7, and $\mathcal N(x)$ 8—to pretrain the sentence encoder $\mathcal N(x)$ 9 and the sequence encoder $C$ 0, before fine-tuning the personalized ranker with supervised pairwise ranking loss (Zhou et al., 2021). The paper does not write a single combined pretraining equation, but states that the four tasks are jointly optimized and reports task weights $C$ 1, $C$ 2, $C$ 3, and $C$ 4 (Zhou et al., 2021). A plausible implication is that paired sampling here functions primarily as representation pretraining rather than direct task supervision.

FineCo supplements a pair-level objective with a finer-grained within-pair contrastive term,

$C$ 5

where $C$ 6 is the frame-level contrastive loss and $C$ 7 is the usual pair-level retrieval loss or cross-entropy for VideoQA (Wang et al., 2022). This multi-granularity design implies that pair-level alignment and sub-instance denoising can be trained jointly rather than treated as competing alternatives.

CL-UFEF uses a graph-weighted contrastive loss to learn a linear projection matrix $C$ 8,

$C$ 9

with

$G^{pos}=\{X,S^{pos}\}$ 0

(Zhang, 2021). Here the pair structure is global and deterministic rather than batch-local.

An important theoretical generalization is the NCA reinterpretation of contrastive learning. “Revisiting Contrastive Learning through the Lens of Neighborhood Component Analysis” shows that contrastive learning can be written as maximizing neighborhood mass over multiple positives,

$G^{pos}=\{X,S^{pos}\}$ 1

which recovers SimCLR as the case $G^{pos}=\{X,S^{pos}\}$ 2 (Ko et al., 2021). The resulting NaCl and IntNaCl losses generalize pairwise contrast into positive neighborhoods and robustness-aware pairwise objectives (Ko et al., 2021). This suggests that paired sampling can naturally broaden into neighborhood sampling when multiple semantically meaningful positives are available.

A more recent theoretical treatment shifts attention from the loss to the pair-sampling law itself. “The Loss Is Not Enough” formalizes positive-pair generation as a conditional law $G^{pos}=\{X,S^{pos}\}$ 3 on latent space and defines the diversity condition

$G^{pos}=\{X,S^{pos}\}$ 4

as a necessary support condition for isometric latent recovery (Zaliaduonis et al., 2 Jun 2026). Under full-support von Mises–Fisher sampling, asymptotic InfoNCE minimizers recover latent geometry up to orthogonal transformation; under restricted support, standard InfoNCE can prefer non-orthogonal, geometry-distorting maps (Zaliaduonis et al., 2 Jun 2026). A support-corrected InfoNCE variant restores orthogonal maps as minimizers but not uniquely so (Zaliaduonis et al., 2 Jun 2026). This implies that a paired-sampling contrastive framework cannot be understood only through the algebra of the loss; the support of the positive-pair sampling mechanism is itself an identifiability condition.

5. Empirical behavior across domains

The empirical literature indicates that paired sampling is most effective when it injects structure unavailable to generic augmentation or naive label supervision.

PairCFR reports that on human-edited CAD datasets, explicit pair-aware batching plus contrastive regularization improves out-of-distribution performance over both non-CAD baselines and prior CAD-specific methods (Qiu et al., 2024). For example, with BERT on sentiment, PairCFR reaches average OOD accuracy $G^{pos}=\{X,S^{pos}\}$ 5, compared with $G^{pos}=\{X,S^{pos}\}$ 6 for HCAD, $G^{pos}=\{X,S^{pos}\}$ 7 for CFGSL, and $G^{pos}=\{X,S^{pos}\}$ 8 for ECF; with BERT on NLI, it reaches $G^{pos}=\{X,S^{pos}\}$ 9 average OOD, compared with $G^{neg}=\{X,S^{neg}\}$ 0 for HCAD, $G^{neg}=\{X,S^{neg}\}$ 1 for CFGSL, and $G^{neg}=\{X,S^{neg}\}$ 2 for ECF (Qiu et al., 2024). The ablation comparing ShuffCAD + CE, PairCAD + CE, ShuffCAD + CE+CL, and PairCAD + CE+CL finds that PairCAD + CE+CL performs best, indicating that CAD availability alone is insufficient; preserving pairwise batch co-occurrence matters (Qiu et al., 2024).

PSSL likewise shows that log-mined pair construction is not a cosmetic addition. On AOL, PSSL improves over PEPS by $G^{neg}=\{X,S^{neg}\}$ 3 MAP, $G^{neg}=\{X,S^{neg}\}$ 4 MRR, and $G^{neg}=\{X,S^{neg}\}$ 5 P@1; on the commercial dataset, it improves over RPMN by $G^{neg}=\{X,S^{neg}\}$ 6 MAP, $G^{neg}=\{X,S^{neg}\}$ 7 MRR, $G^{neg}=\{X,S^{neg}\}$ 8 P@1, and $G^{neg}=\{X,S^{neg}\}$ 9 P-improve (Zhou et al., 2021). The ablations report that removing any one of DP, QP, SAP, or UP hurts performance, with SAP having the largest impact (Zhou et al., 2021).

FineCo’s gains are strongest when the global pair contains substantial internal noise. On YouCookII, FineCo improves VideoCLIP from $S^{pos}$ 0 to $S^{pos}$ 1 R@1 without Dual Softmax, and to $S^{pos}$ 2 R@1 with Dual Softmax (Wang et al., 2022). On shorter-video benchmarks the gains are smaller, which is consistent with the paper’s claim that paired sub-sampling is most beneficial when correspondence is sparse and videos are long (Wang et al., 2022). The fixed- $S^{pos}$ 3 versus random frame-selection comparison is especially direct: on YouCookII, fixed $S^{pos}$ 4 gives $S^{pos}$ 5 R@1, whereas random $S^{pos}$ 6 gives $S^{pos}$ 7 R@1, supporting the claim that useful positives must be sampled conditionally on the paired text rather than chosen arbitrarily (Wang et al., 2022).

ConVIRT shows that naturally paired multimodal data can be more effective than image-only contrastive learning in medical imaging. On RSNA linear evaluation with $S^{pos}$ 8 labels, ConVIRT reaches $S^{pos}$ 9 AUC, compared with $S^{neg}$ 0 for ImageNet initialization, $S^{neg}$ 1 for SimCLR, and $S^{neg}$ 2 for MoCo v2; on CheXpert linear evaluation with $S^{neg}$ 3 labels, it reaches $S^{neg}$ 4 AUC, compared with $S^{neg}$ 5, $S^{neg}$ 6, and $S^{neg}$ 7, respectively (Zhang et al., 2020). The image–text retrieval-style bidirectional objective therefore appears to provide a stronger paired-sampling signal than same-modality instance discrimination in a domain with high inter-class visual similarity (Zhang et al., 2020).

ConTeX provides evidence that separating same-source paired views from same-label context positives can improve both convergence and robustness to spurious correlations. On CIFAR10, CIFAR100, and ImageNet linear evaluation, it reports $S^{neg}$ 8, $S^{neg}$ 9, and $k$ 0, compared with $k$ 1, $k$ 2, and $k$ 3 for SupCon (Deng et al., 1 Dec 2025). More strikingly, on BiasedMNIST the abstract reports a $k$ 4 improvement compared to original contrastive loss functions, and the full tables show large gains at severe target-bias correlation settings (Deng et al., 1 Dec 2025). This suggests that when label-level positives encode shortcuts, enforcing same-source pair fidelity can act as a regularizer against distortion.

The survey literature generalizes these findings. “A Survey on Data Curation for Visual Contrastive Learning” treats pair construction as a determinant of representation quality, semantic fidelity, convergence speed, robustness, and computational cost (Desai et al., 12 Feb 2025). Its taxonomy of positive-pair curation and negative-pair curation implies that paired-sampling frameworks should be evaluated not only by loss design but by the semantic validity, diversity, hardness, and computational feasibility of the curated pairs (Desai et al., 12 Feb 2025).

6. Limitations, controversies, and theoretical cautions

The main limitation recurring across paired-sampling contrastive frameworks is that pair quality is task-critical and failure-prone. The methods differ, but the failure modes are structurally similar.

PairCFR is tested only on relatively small, high-quality human-edited CAD datasets, and the paper states that it remains unclear how robust the paired-sampling contrastive setup is under noisier LLM-generated counterfactuals (Qiu et al., 2024). It also notes that gains are strongest on OOD generalization and can come with lower in-domain performance than non-CAD baselines, reflecting the distribution shift introduced by CAD (Qiu et al., 2024). The NLI neutral-class analysis further shows that not every class is equally suitable for contrastive computation, since excluding neutral samples improved OOD performance (Qiu et al., 2024).

PSSL depends on heuristics for pair mining from search logs. Same-query co-clicks, same-document reformulations, and ambiguous-query user matches are plausible proxies for shared intent, but they are not guaranteed semantic equivalences (Zhou et al., 2021). A plausible implication is that the framework’s success depends on the degree to which log-derived pair rules track true latent relevance rather than incidental behavioral correlations.

FineCo uses current model similarities to define top- $k$ 5 frame assignments, so the positive/negative split is bootstrapped from imperfect similarity estimates early in training (Wang et al., 2022). The paper does not explicitly analyze this training-dynamics issue. This suggests that pair-conditioned sub-sampling may be sensitive to initialization and to the hyperparameter controlling the positive-set size.

Weakly paired group-level contrast has its own risk: same-label cross-modal samples may share only coarse structure. GROOVE assumes that a shared perturbation label is a valid weak pairing signal across modalities (Gorla et al., 3 Feb 2026). If labels are noisy or if shared versus modality-specific variation is weak, the positive sets are contaminated by design (Gorla et al., 3 Feb 2026). The paper also stresses that no optimal transport aligner uniformly dominates across settings, which implies that good weak-pair representation learning does not eliminate the need for careful downstream alignment selection (Gorla et al., 3 Feb 2026).

ConTeX raises a different controversy: whether same-label positives should be treated as an undifferentiated positive pool in supervised contrastive learning. The paper argues that this can create conflict and distortion, but its mathematical presentation contains notation and typesetting inconsistencies, especially in the combined loss and some gradients (Deng et al., 1 Dec 2025). The empirical claim is clear, but the exact formalism is less cleanly presented than in some earlier contrastive frameworks.

The most general theoretical caution comes from sampling-support analysis. “The Loss Is Not Enough” argues that restricted positive-pair support can make standard InfoNCE prefer geometry-distorting encoders, even at the population optimum (Zaliaduonis et al., 2 Jun 2026). This suggests that paired-sampling contrastive design cannot be reduced to choosing “better positives” in an intuitive sense; the support of the positive-pair law determines what geometry is even identifiable. A related, more formal perspective from active learning shows that the usefulness of contrastive pairs depends strongly on the concept class and metric, not on pair minimality alone (Mansouri et al., 18 Jun 2025). In that framework, nearest opposite-label pairs can reduce exact learning of monotone monomials from $k$ 6 queries to $k$ 7, yet remain essentially unhelpful for parity functions (Mansouri et al., 18 Jun 2025). This suggests that pair informativeness is representation-dependent and class-dependent rather than universal.

Taken together, these results undermine a common misconception: that any explicit pairing automatically strengthens contrastive learning. The literature instead shows that pairing helps only when the pairing rule is semantically aligned with the structure that the objective should recover.

7. General principles and conceptual synthesis

Several principles recur across the literature and define the modern notion of a paired-sampling contrastive framework.

The first is that pair construction is a modeling choice, not merely a data-loader detail. PairCFR’s difference from generic supervised contrastive learning is the co-batching of original and counterfactual so that the counterfactual becomes a structured hard negative (Qiu et al., 2024). PSSL’s novelty lies in mining four specific forms of log-derived pairs and using them to pretrain the exact encoders later used for ranking (Zhou et al., 2021). FineCo’s central move is to mine true positive substructure inside a globally positive pair (Wang et al., 2022). These methods would be qualitatively different if the same losses were applied to randomly shuffled data.

The second is that positive-pair semantics are not uniform across domains. Sometimes the paired sample should be a positive, as in PSSL’s same-document or same-history views (Zhou et al., 2021). Sometimes it should be a negative, as in PairCFR’s original–counterfactual geometry (Qiu et al., 2024). Sometimes the pair determines a conditioning relation that defines latent positive and negative subsets, as in FineCo (Wang et al., 2022). Sometimes the positive relation is many-to-many and group-level, as in GROOVE (Gorla et al., 3 Feb 2026). A plausible implication is that the most informative paired-sampling framework for a domain is the one whose pair semantics match the task’s causal or structural relations rather than defaulting to augmentation identity.

The third is that pair sampling and batch geometry interact. PairCFR reports that performance rises and then plateaus or declines with batch size, interpreting this as a balance between local paired hard negatives and global batch negatives (Qiu et al., 2024). PSSL uses in-batch negatives without memory structures, so batch composition defines the effective hardness of negatives (Zhou et al., 2021). GROOVE explicitly uses balanced under-sampling because GroupCLIP is sensitive to batch composition (Gorla et al., 3 Feb 2026). These findings suggest that a paired-sampling framework is often partly a sampler-design problem and partly a loss-design problem.

The fourth is that richer pairing can improve transfer and robustness, but only if it preserves semantic fidelity. The survey literature emphasizes that positives should be semantically aligned yet diverse, while negatives should be informative but trustworthy (Desai et al., 12 Feb 2025). ConTeX’s split between same-source paired views and same-label context positives can be read as one practical attempt to maintain that balance (Deng et al., 1 Dec 2025). The support-based identifiability results of (Zaliaduonis et al., 2 Jun 2026) imply that this balance also has a formal side: too-restricted positive support can make the objective itself misidentify the latent geometry.

The fifth is that paired-sampling contrastive learning increasingly includes weak, induced, or structured pairings rather than exact one-to-one pairs. GroupCLIP shows that group labels can stand in for exact multimodal correspondences (Gorla et al., 3 Feb 2026). scBeacon shows that cluster-level matching across conditions can induce pseudo-positive cell pairs for a contrastive siamese network (Liu et al., 2023). Hydra shows that even semantically distinct class pairs can be treated as positives in a learned pair-conditioned subspace (Wu et al., 2024). This suggests a broadening of the term “pair” from literal object identity to relation-specific compatibility.

In summary, a paired-sampling contrastive framework is best understood as a contrastive learning architecture whose supervision is organized around an explicit relation between samples and whose success depends on how faithfully that relation encodes task-relevant structure. In some cases, the pair is a positive anchor. In others, it is a hard negative, a group-level set, a sub-instance selector, or a theoretically constrained conditional law. The common lesson is that contrastive learning is shaped not only by the loss, but by the semantics, support, and geometry of the pairing mechanism itself (Qiu et al., 2024, Desai et al., 12 Feb 2025, Zaliaduonis et al., 2 Jun 2026).