Self-Supervised Learning

Many self-supervised methods for training foundation models at the Large Hadron Collider (LHC) rely on data augmentations to encourage the model to embed events into a representation space invariant to certain physical or detector symmetries. A common challenge arises from the large freedom in choosing a proper set of augmentations on which downstream performance depends. The implementation of augmentations involves either modifying existing events, potentially breaking the event fidelity, or simulating more event variants, which is computationally intensive. In this work, we present a data-driven method of pairing events by their similarity via the energy mover's distance (EMD), which measures how similar two events are in terms of the work required to transform one into the other. With this approach, distinct events are sampled and matched by their similarity to serve as views for learning invariance, keeping the physics content of each event intact without handcrafted distortions. We demonstrate this augmentation-free pairing method by pre-training on QCD jets via self-distillation and show that it can yield semantic jet embeddings with downstream discrimination power comparable to or better than an augmentation-based baseline.

Reconstructing the mass of a heavy resonance from its decay products with missing energy is one of the central tasks that directly determine the sensitivity in new physics searches at collider experiments. Supervised learning approaches to this problem often struggle to generalize well due to the presence of various systematic uncertainties and distribution shifts. Exhausting all possible variations in the labeled data can be very compute-intensive, while a failure of the model to generalize can corrupt the reconstructed resonance widths that are critical in peak-hunting analyses. In this work, following the foundation model paradigm, we use a self-supervised approach to pre-train a transformer encoder with VICReg to learn an embedding invariant to various corruptions, then fine-tune it for mass regression on a heavy resonance with masses ranging from 2.5 to 6.5 TeV and a SUSY-like cascade decay into an eleven-body final state. We show that the pre-trained model reconstructs sharper resonance peaks and has a more stable performance under various realistic corruptions, compared to a supervised model of the same architecture trained on the same augmented data from scratch.

Self-supervised learning (SSL) can reduce the need for labelled medical images, but the choice of pretext objective remains unclear for lung ultrasound (LUS). Contrastive learning, masked reconstruction, and joint-embedding predictive architectures (JEPA) differ in the space in which their targets are defined, yet existing ultrasound studies compare them under different corpora, backbones, and evaluation protocols. We compare these three objective families using the same encoder backbone, pretraining corpus, optimisation schedule, and frozen-evaluation protocol. Encoders are pretrained on COVID-BLUeS LUS videos and evaluated with linear, $k$NN, and attentive probes at 5%, 10%, 50%, and 100% label budgets. Evaluation is performed on POCUS using patient-level five-fold cross-validation and on the independently acquired Mendeley-Uganda dataset, which is excluded from both pretraining and probe fitting. At the full label budget under linear probing, VideoMAE and V-JEPA achieve $66.5 \pm 13.1$ and $65.4 \pm 11.7$ balanced accuracy on POCUS, while MoCo achieves $42.1 \pm 1.2$. On Mendeley-Uganda, the ranking reverses: MoCo performs best at $62.7 \pm 1.0$, followed by VideoMAE at $53.8 \pm 2.8$, while V-JEPA falls near chance at $35.1 \pm 4.9$. These results show that POCUS probe accuracy alone does not identify the objective that transfers best across datasets. We also outline planned representation-level analyses to examine this reversal. Code is publicly available at https://github.com/moeinheidari7829/LUSVideoSSL.

Passive acoustic monitoring can measure biodiversity at larger scales, but time--frequency annotation of animal vocalizations is expensive, site-specific, and difficult to sustain at scale. We present a label-efficient sound detection framework that combines masked audio pretraining with a lightweight detector on mel spectrograms, then further improves robustness through iterative self-training on unlabeled audio. We first pretrain a ViT-based encoder on unlabeled recordings via masked reconstruction and transfer the encoder to a detection backbone. To better separate animal sounds from confounding background, we add a box-level contrastive loss that pulls matched event regions together while pushing noisy negatives apart. We then apply a two-stage pseudo-labeling curriculum to exploit large unlabeled pools without additional annotation. We evaluate the performance on two ecologically distinct domains: tropical rainforest soundscapes (Indonesia) and bird vocalizations in Mediterranean habitats (Spain). On both domains, masked audio pretraining and contrastive learning consistently improve time--frequency detection under temporal and cross-site distribution shift, and self-training yields further gains in out-of-distribution performance. On the rainforest domain, MAST with self-training achieves +0.22 mAP and +0.24 F1 over the strongest baseline under cross-site shift. On the bird domain, self-training achieves +0.12 mAP and +0.10 F1 over the strongest baseline under cross-site shift. Overall, our results show that MAST can effectively extend self-supervised audio representations from clip-level tasks to robust box-level localization across diverse bioacoustic settings, providing a practical path for biodiversity monitoring with limited labels.

Object-Centric Learning (OCL) aims to decompose images into objects without human annotations. A major family of mainstream methods uses Slot Attention to aggregate image features into object-level representations and then from them reconstructs masked image content, i.e., Random Masking (RM), to provide self-supervision. The recent method DIAS simply masks image patches at uniform randomness yet achieves competitive object discovery accuracy. Since attention during aggregation already possesses object discovery ability, we explore using it to develop a better image patch masking strategy, i.e., Attention Guided Masking (AGM), thereby providing better self-supervision. Results on six recognized datasets show that AGM does not always outperform RM. Under unconditional slot initialization, AGM substantially improves background segmentation on datasets with realistic textures (COCO and VOC); Regardless of conditional or unconditional slot initialization and across datasets, foreground object discovery remains comparable or decreases. We suggest peer researchers in the OCL community that attempts to exploit internal attention semantics to improve OCL with masked decoding are risky. Our source code, model checkpoints and evaluation logs will be released upon acceptance.

Modelling how a disease progresses over time requires longitudinal imaging cohorts, which are scarce and small, whereas cross-sectional data -- one image per participant -- is abundant. Self-supervised pre-training on such data offers a way to bridge this gap, but it is unclear which strategy best supports progression modelling, or how that answer depends on the amount of labelled longitudinal data. We study this for age-related macular degeneration (AMD), pre-training encoders on the large cross-sectional NAKO cohort and predicting time to late AMD on the longitudinal AREDS dataset. We compare in-house self-supervised encoders against a general-purpose (DINOv2) and a domain-specific (RETFound) foundation model, across contrastive, masked-autoencoding, and self-distillation objectives, under frozen and fine-tuned protocols, and across labelled training sets from 100 to 32,250 examples. Which model performs best depends on how the encoder is used. When the encoder is frozen and labels are few -- the regime typical of longitudinal cohorts -- pre-trained representations reach clinically reasonable discrimination from a few hundred labelled samples, while models trained from scratch do not; this advantage fades under fine-tuning. Transfer is governed by the self-supervision objective rather than corpus scale or domain match, so that an encoder pre-trained on a modest cross-sectional cohort matches or exceeds a far larger in-domain foundation model. Together, these results offer a practical recipe for building progression models where longitudinal data is scarce: a frozen self-supervised encoder with a lightweight survival head.

Mean Opinion Score (MOS) is the gold standard for evaluating synthesized speech naturalness. However, current automatic MOS predictors are dominated by self-supervised learning (SSL) models that prioritize high-level semantics, potentially compromising their ability to capture critical acoustic details. In this paper, we systematically investigate representations from three paradigms: SSLs, acoustic-only neural audio codecs (NACs), and unified NACs that integrate semantics into reconstruction-based architectures. Extensive benchmarking on the standard BVCC and multiple out-of-domain (OOD) datasets demonstrates that features synergizing semantic understanding with fine-grained acoustic modeling achieve a higher performance upper bound in speech quality assessment. Ultimately, our findings highlight that semantics alone are not enough; a dual focus on semantic content and acoustic fidelity is essential for robust MOS prediction.

Dense self-supervised learning (SSL) is a powerful paradigm for learning without annotations the local descriptors required to solve dense medical imaging tasks. We present Pix2Rep-v2, a framework for SSL of pixel- and voxel-level representations suitable for few-shot downstream applications. Pix2Rep-v2 addresses the main challenges of dense SSL by leveraging a redundancy reduction objective at the pixel-level with a principle of equivariance of dense representations, that scales efficiently to 3D or wide field-of-view applications. We evaluate our method on four datasets, across multiple tasks, multiple modalities and anatomical structures using multiple backbones in 2D and 3D, and under various data regimes. As an alternative to linear probing or full fine-tuning on the downstream task, we also propose an in-context variant, without downstream training, based on a dense prototype approach. Pix2Rep-v2 shows substantially higher data-efficiency in few-shot scenarios compared to fully supervised baselines, and is competitive with the state-of-the-art e.g., +9.3 Dice points in one-shot segmentation on the M&Ms-2 dataset. Our code and pre-trained models are publicly available at https://github.com/BioMedTP/pix2rep-v2.

Low-dose computed tomography (LDCT) measurements contain mixed Poisson-Gaussian noise. However, most self-supervised methods rely on generic image statistics and do not explicitly model this noise, which may limit their ability to effectively suppress realistic LDCT noise. To address this issue, we propose a physics-driven framework with cross-domain iteration for self-supervised LDCT denoising. The proposed framework proceeds in three main steps. First, a learned sinogram prior and the LDCT noise model guide posterior inference of photon counts, enabling separation of the Poisson and Gaussian components. Second, the separated Poisson and Gaussian components are respectively processed by binomial thinning and Gaussian data thinning to construct two branches, and residual scaling matches each branch's noise level to that of the observation, yielding a training pair with approximately independent noise realizations from one low-dose measurement. Finally, the pair is used to train an image-domain network whose forward-projected outputs update the prior. Through cross-domain iteration, the prior and the training pair are progressively refined while maintaining consistency with CT acquisition physics. Experiments on simulated data from AAPM, LIDC-IDRI, and LoDoPaB-CT and on real LDCT data show consistent gains over the evaluated self-supervised baselines across dose levels, with performance comparable to the evaluated supervised baseline.

While self-supervised approaches to reinforcement learning have achieved strong results by learning representations of states and actions, a key open question is the time scale over which actions should be modeled. Departing from the standard formulation relying on single-step actions, we extend contrastive reinforcement learning (CRL), a prototypical self-supervised method, to operate over action chunks, and find that this results in large, pervasive gains across established offline and online benchmarks: +31.7% and +93.1% across 18 and 11 environments respectively. While action-chunking-driven gains are generally explained through the ability to model non-Markovian, temporally extended policies, and to propagate unbiased multi-step returns, interestingly, we find that these arguments only partially apply to CRL. Our empirical studies suggest that, in the context of CRL, an action chunk carries more information about the goal than a single action, measurably improving the critic's representations, and rendering the algorithm significantly more effective.

We compare six self-supervised pretext tasks for infant cry analysis under a fixed budget, meaning the same compact encoder of 1.17M parameters, the same 115 hours of license-verified public pretraining audio, and the same evaluation protocol for every candidate. On cry detection the reconstructive objectives dominate, and a linear probe over a masked-spectrogram encoder reaches 0.988 AUC with subject-wise splits even though the encoder never observed a cry during pretraining. On cry-reason classification over donateacry, the de facto public benchmark for cry reasons, every encoder performs at chance (0.38 to 0.54 macro AUC over 5 classes), and neither domain adaptation on 1.8 hours of real cries nor end-to-end fine-tuning moves the result. Since a frozen HuBERT-base with 80 times more parameters shows the same pattern, the bottleneck must sit in the labels and not in model capacity. We then reproduce the 90%+ accuracies of the donateacry literature on our own system by changing nothing but the evaluation protocol: clip-wise splits raise accuracy to 85.2% (barely above the 83.8% majority-class baseline), and applying augmentation before splitting raises it to 97.9%, matching the reported state of the art, from the same model that measures 0.49 macro AUC under subject-wise splits. Under leakage-free splits, a twentyfold augmentation of the labeled set (vocoder speaker perturbation and noise mixing, 21 hours) leaves cross-subject AUC unchanged: for this task the effective sample size is the number of infants. We release code, seeds and per-clip license manifests.

The proliferation of distributed energy resources (DERs) in distribution grids enables the active coordination of these assets to reduce costs and enable cleaner operations. Realizing this potential requires solving multiphase AC optimal power flow (AC-OPF) quickly across varying loads, DER availabilities, and topology reconfigurations, at much greater speed and scale than conventional nonlinear solvers. Learning-based surrogates can offer millisecond inference, yet existing methods target largely balanced transmission systems and do not scale to the multiphase, unbalanced, and reconfigurable nature of distribution feeders at utility scale. We present the Penalty + Sequential Linearized Feasibility Seeking (SLFS) algorithm, a self-supervised learning framework for multiphase distribution AC-OPF under switch-induced topology changes. Penalty+SLFS requires no labeled optimal solutions and trains directly from the AC-OPF objective and constraints through a differentiable fixed-point power flow solver, avoiding expensive label generation and admitting robust training procedures. Topology changes are handled efficiently using Sherman-Morrison-Woodbury updates of the admittance-matrix inverse, while an M-step Jacobian approximation accelerates differentiation through the power flow solver. At inference, SLFS repairs any infeasible predictions, providing feasibility guarantees with low computational overhead. On IEEE feeders ranging from 13 to 8,500 nodes, Penalty+SLFS achieves negligible optimality gaps and near-zero constraint violations, delivers up to three orders of magnitude speedups over IPOPT, and remains robust under large distributional shifts, demonstrating a viable path toward real-time, topology-aware AC-OPF for large-scale distribution grids.

Visual foundation models are a cornerstone of image and video understanding but typically require large amounts of data and computation. The current scale required for pretraining visual foundation models may be unsustainable or unnecessary, and significant benefits arise when effective models can be obtained with fewer resources. To better understand how self-supervised learning (SSL) objectives behave under resource constraints, we conduct a controlled study of image and video SSL objectives under matched data, architecture, and compute budgets. We compare contrastive, reconstruction, feature-prediction, and diffusion objectives and evaluate both standalone and jointly trained image-video SSL formulations across a diverse set of image and video understanding tasks. Our results show that DINOv2-style pretraining consistently provides the strongest overall performance under limited resources. Furthermore, combining DINOv2 with video SSL objectives such as VideoMAE substantially improves image classification and segmentation performance, but degrades video tracking and camera-pose estimation performance, revealing an important tradeoff between semantic and geometric representation learning. These findings suggest that combining image and video SSL objectives can be beneficial in resource-limited settings, while highlighting the need for improved methods that better balance semantic, temporal, and geometric supervision.

Generalized Category Discovery aims to recognize known categories while identifying novel ones within unlabeled data. Existing methods, typically based on self-supervision and contrastive learning, often struggle to capture fine-grained distinctions, relying on superficial visual cues rather than the intrinsic attributes humans use for categorization. We introduce Fourier Self-Supervision, that leverages the Fourier transform of images to enhance the discrimination of subtle differences and support the discovery of new categories. Our method employs a dual frequency filtering strategy: a low-pass filter first extracts broad, abstract attributes that capture high-level category information, while a high-pass filter emphasizes fine details such as edges and textures that are essential for fine-grained recognition. Each operates on a dedicated latent space, and their overlapping representations together yield a richer, more complete feature space. This dual-frequency approach not only refines feature extraction to identify novel categories, but also strengthens the model's discriminative power in fine-grained category discovery. Experiments on multiple fine-grained datasets show that incorporating Fourier Self-Supervision outperforms state-of-the-art methods, even when the number of classes is unknown, demonstrating its effectiveness for Generalized Category Discovery. Our code is available at: https://github.com/SarahRastegar/FourEx.

We argue that learning visual representations without labels requires a training signal jointly complete across three non-overlapping objectives: semantic invariance across augmented views, patch-level spatial prediction, and representational non-degeneracy. We formalize these as the observation, prediction, and regularization principles and prove (i) that combining observation and prediction without regularization admits the constant encoder as a global minimizer under negative-free alignment; (ii) that the two objectives are gradient-complementary and structurally non-conflicting at the encoder output; and (iii) that the momentum encoder converges to the same fixed point as the online encoder and provides no collapse guarantee at convergence. Contrastive alignment provides only self-limiting collapse resistance, formalized via an explicit gradient-decay argument. Dropping prediction withholds the spatial training signal by construction; dropping observation forfeits cross-view semantic invariance by construction; at the scale we study, no pair substitutes for the third. Every major self-supervised method is a special case of a single unified energy decomposition. We pair every theoretical claim with a controlled experiment, including a patch-retrieval evaluation for the spatial consequence of prediction.

LiDAR serves as a primary sensing modality for robots operating in outdoor environments. However, the performance of deep learning models in this domain is severely limited by the scarcity of labeled data, a direct result of the high cost of 3D annotation. Self-supervised learning addresses this scarcity by learning general-purpose features from unlabeled data. In this work, we present a multi-modal, multi-teacher distillation framework for self-supervised learning on outdoor LiDAR point clouds. Building upon the Sonata architecture, we introduce Vernata, consisting of three extensions: sparse view augmentation to improve robustness against varying point densities, a memory bank mechanism to stabilize resource-constrained training, and cross-modal distillation utilizing dense, high-resolution 2D image features to enable fine-grained semantic guidance. We evaluate our method on the GrandTour, TartanGround, and Waymo datasets, as well as data collected from our own robotic platforms. Our experiments demonstrate a significant performance improvement over Sonata baselines, yielding mIoU scores of 54.7 on TartanGround (+5.9 points, +12.1%) and 57.1 on Waymo (+7.3 points, +14.7%). Finally, we show that the self-supervised approach maintains strong performance even in reduced-modality settings (lacking color or normals), achieving competitive mIoU scores of 49.4 and 50.2 on the respective datasets.

Inspired by recent evidence that transformer architectures benefit from Self-PreTraining (SPT) on long-context benchmarks, we investigate whether similar gains extend to multimodal, multivariate, and even simple univariate medical time series. Our objective is to assess the impact of SPT on the performance and scalability of transformer-based models across diverse medical applications, particularly under limited data conditions. We evaluate transformer architectures on three representative medical time-series tasks: rehabilitation robotics (Camargo dataset), stress detection (Non-EEG Stress), and Parkinson's disease detection (Gait Parkinson's Disease). Models are trained either from scratch or through SPT using four masking-based objectives designed to promote temporal and cross-modal representation learning, and we systematically vary model depth to examine how capacity interacts with pre-training benefits. Across datasets and configurations, SPT consistently improves classification accuracy by 0-6 percentage points depending on masking strategy, dataset and architecture, with gains observed not only in multivariate settings but also when models are restricted to simple univariate inputs. The improvements increase for deeper models that can better exploit the enriched temporal representations learned during pre-training. These findings indicate that SPT is a simple and general strategy that enhances transformer performance on medical time-series tasks without requiring task-specific architectural changes, supporting its potential to improve robustness and accuracy in data-limited clinical settings.

Deep learning is a new way for machinery fault diagnosis but requires extensive labeled data, a scarce resource in industrial settings. We propose Spectral Aliasing Pretext (SAP), a self-supervised learning method that pretrains models on unlabeled vibration data by exploiting spectral aliasing. We deliberately undersample signals to create folded spectrum, then train a Transformer to reconstruct the original unfolded spectrum. This pretext task forces the model to learn frequency-domain invariants characteristic of mechanical faults, without potentially destructive augmentations. Experiments on the CWRU dataset show that SAP learns stable and highly discriminative representations. In a linear probing setting, SAP quickly achieves very high classification performance with only a small fraction of labeled data and low variance. In contrast, full fine-tuning, including fully supervised training, does not lead to more stable or better results. Overall, these findings suggest that SAP combined with linear probing can be more effective and reliable than fully supervised training for fault diagnosis with limited labeled data.

Self-supervised pretraining objectives are spatially uniform: the teacher temperature and the per-patch loss weight are identical everywhere in the image, so a lesion a few patches wide contributes no more to the training signal than the surrounding parenchyma. Prior work biases the views toward annotated regions, which changes what the model sees but adds no pressure on the objective. We instead condition the targets of self-distillation, a method we call SALT (Spatially Adaptive Label-guided Temperature). Weak, box-derived labels, available only during pretraining, define a compact region on the encoder's patch grid, inside which the teacher's softmax temperature is sharpened and the masked-patch loss is up-weighted. The objectives, the masking policy and the centering statistics are otherwise unchanged, and at every downstream use the encoder is a plain feature extractor with no labels and no conditioning. We evaluate by freezing the encoder and training only a lightweight multi-depth CenterNet-style head, detecting lesions in 3D on four CT cohorts, and we isolate the mechanism against a backbone identical in architecture, pretraining data, schedule and label-guided cropping but with no target conditioning. We report patch-level separability, 3D detection stratified by cohort and by lesion size, box quality, and a detector-free probe in which a single frozen patch embedding re-identifies a lesion in a follow-up scan without registration, masks or fine-tuning. Because the conditioning is expressed through a spatial indicator rather than through label semantics, the formulation admits any weak spatial annotation; we instantiate and validate it for lesions.

Self-supervised learning on graphs is largely shaped by contrastive methods that depend on carefully designed augmentations, and by generative methods that reconstruct node attributes in the input space. Both paradigms can entangle representations with low-level input statistics rather than with relational structure. Joint-embedding predictive architectures (JEPA) instead learn by predicting latent targets rather than reconstructing inputs. Recent work has explored this idea for graph-level representation learning, but how to design JEPA-style objectives for node-level tasks, and which structural signals the predictor should condition on, remains less clear. We present NodeJEPA, a joint-embedding predictive architecture for node-level graph self-supervised learning. NodeJEPA masks structure-aware k-hop ego-subgraphs and trains a context encoder to predict the latent representations of the masked nodes. These targets come from an EMA-updated target encoder with stop-gradient. A structure-conditioned predictor integrates spectral and centrality descriptors through cross-attention. Variance, covariance, and Laplacian spectral regularizers help stabilize the embedding geometry, and an optional curriculum gradually increases masking difficulty during training. Because prediction occurs in latent space, NodeJEPA does not rely on input reconstruction or hand-crafted graph augmentations. We evaluate NodeJEPA on standard node classification benchmarks under linear probing and fine-tuning protocols, and conduct ablations on masking, prediction, and regularization design choices. Our study offers a practical recipe for node-level JEPA-style latent prediction on graphs, and clarifies when structural conditioning helps representation learning. Code, configurations, and evaluation scripts are publicly available at https://github.com/OliverZ-dot/Node-Jepa.

Existing studies on self-supervised learning for white-box networks typically decouple the derivation of white-box networks via optimization algorithms from self-supervised learning paradigms. In this work, we instead revisit the two components from a joint perspective. The LeJEPA-based self-supervised framework assumes an isotropic Gaussian distribution as the optimal embedding distribution for downstream tasks, which is conceptually equivalent to the expansion term $R(Z)$ in the sparse rate reduction objective guiding white-box Transformer optimization. Building on this observation, we use the LeJEPA self-supervised paradigm to optimize $R(Z)$, and derive the remaining terms $R^{c}(Z\mid U_{[K]})+λ\lVert Z\rVert_{0}$ via the alternating direction method of multipliers (ADMM) into an attention-only Transformer that dispenses with the ISTA structure or MLP layers of the original design. Experimental results demonstrate that our attention-only white-box Transformer achieves classification accuracies of $88.88\%$ on CIFAR-10 and $63.54\%$ on CIFAR-100 at the Base scale under the LeJEPA self-supervised paradigm, while the original white-box Transformer CRATE achieves classification accuracies of $89.18\%$ on CIFAR-10 and $63.56\%$ on CIFAR-100. Our model achieves competitive performance while reducing the parameter count by roughly $31\%$. Beyond the white-box setting, we further investigate standard ViTs and find that replacing all MLP blocks with ReLU activations under knowledge distillation removes approximately 66% of the parameters while preserving competitive accuracy, motivating further investigation into the potential redundancy of MLP modules in standard ViT architectures.

Pretraining has shown strong potential for learning transferable representations, yet it remains underexplored for electron-density-based molecular learning. Electron density provides a continuous three-dimensional description of molecular electronic structure, capturing both local spatial patterns and global physical quantities. This raises a key question: can electron-density fields be used for self-supervised pretraining to learn a shared representation that transfers across diverse electronic-structure-related tasks? We propose ED-DiT, a physics-guided Diffusion Transformer for self-supervised pretraining on electron-density point clouds. ED-DiT learns reusable representations by reconstructing corrupted and partially masked log-density fields across diffusion noise levels. An electron-number consistency constraint is further introduced to preserve the total electronic mass. The pretrained encoder can be adapted to property prediction, open-/closed-shell classification, molecule-electron-density retrieval, and molecule-conditioned electron-density prediction. Experiments on six EDBench tasks show that ED-DiT consistently outperforms the same architecture trained from scratch, especially under limited supervision. For molecule-conditioned electron-density prediction, it reduces RMSE from 2.2474 to 1.3753 and surpasses the available baseline. With only 10% labels, it improves orbital energy prediction RMSE from 0.0293 to 0.0138. These results demonstrate the effectiveness of physics-guided electron-density pretraining for learning transferable molecular representations.

We study a self-supervised generation task for single-cell gene expression vectors: given a set of vectors from a cell type, we aim to generate additional gene expression vectors of that cell type. For this task we characterize both the biological fidelity of the generated gene expression vectors and the scaling behavior of the pretraining loss. The model is a causal transformer paired with a learned quantized VAE tokenizer, trained with a cross-entropy loss. To evaluate the model, we condition it on held-out gene expression vectors of a cell type and generate vectors of gene expression, comparing the resulting distribution over gene expression vectors to the ground truth distribution of that cell type. We study the scaling properties of the proposed architecture by varying the number of trained parameters and the amount of training data. To our knowledge, we find the first jointly-fit two-exponent scaling law and compute-optimal frontier for a single-cell foundation model. Finally, we discuss how this pretrained model could be finetuned for perturbation response prediction.

Malware clustering is a critical task in cybersecurity that helps discover threats and analyze evolving malware families. While self-supervised learning (SSL) and tabular representation learning (TRL) have achieved breakthroughs in other domains, their application to binary program clustering (the task of clustering all incoming samples regardless of label) remains largely unexplored. This study presents the first systematic investigation of SSL and TRL methods for binary program clustering, conducted in two phases on the public Ember and Bodmas datasets. In Phase 1, we establish a performance ceiling by adapting prominent vision-based SSL models (BYOL, SimSiam, Barlow Twins, VICReg) for tabular data with supervised pair generation, finding that BYOL and SimSiam achieve performance comparable to fully supervised models, while Barlow Twins and VICReg significantly underperform. In Phase 2, we evaluate purely unsupervised TRL methods against strong baselines (PCA, Autoencoder, UMAP), demonstrating that VIME establishes a new state of the art for binary program clustering. Informed by these findings, we propose VIME-R, a retrieval-augmented extension of VIME that replaces random marginal-distribution corruption with retrieval-based augmentation to generate more informative training pairs. VIME-R further improves upon VIME, achieving 2.7%-5.8% higher Homogeneity on both datasets. Our results highlight retrieval-augmented tabular representation learning as a promising direction for enhancing automated malware analysis. Code will be made available.

This paper describes the participation of team "Go-To-Germany" in the ImageCLEF 2026 Audio Deepfake Detection and Generation task. Our detection system, built on a four-backbone self-supervised learning (SSL) ensemble combining WavLM-Large, Wav2Vec2-XLS-R-300M, ECAPA-TDNN, and x-vector representations, achieved a final score of 0.9522 on the official ImageCLEF 2026 evaluation, with perfect accuracy (1.0000) on participant-generated deepfakes and 0.8875 on the held-out organizer ground-truth real data. For the Generation sub-task, our official team submission, an F5-TTS v1 baseline processed with a uniform reverberation pass and submitted as a deliberate anti-forensic probe, ranked first with a final score of 0.4304 (word error rate (WER) 4.99%, character error rate (CER) 2.07%); details of our four-model program (GLM-TTS, F5-TTS, XTTS v2, CosyVoice3), from which the official entry was drawn, appear in the paper. We present a cross-track analysis revealing a pronounced asymmetry: our detection system identifies 100% of participant-generated deepfakes, while our official generation entry, despite ranking first in the Audio Generation sub-task and evading 61.4% and 56.2% of participant and organizer detectors, attains a Final Score of 0.4304 against 0.9522 on the Detection side. We further report falsification-based ablation experiments (LOSO 56-speaker cross-validation, three-region backbone geometry, bootstrap confidence intervals, and PCA analysis) that motivate our architectural-insurance hypothesis for multi-backbone SSL ensembling. We complement these results with five cross-track insights and five pre-registered falsification experiments connecting generation-side evasion to detection-side design decisions, and we openly report an 11.25% false-positive gap on held-out organizer real recordings as the principal open challenge for deployment.

Modern autonomous-driving fleets record far more video than human reviewers can inspect. This motivates the need for an automatic clip triage mechanism, to surface rare and review-worthy clips, so that driving models can be fine-tuned to better handle unideal circumstances. We test a label-free approach that scores clips by the prediction-error "novelty" of a self-supervised joint-embedding predictive architecture (JEPA); a frozen V-JEPA video encoder is paired with a lightweight predictor head to reconstruct masked clip embeddings, and clips whose embeddings are hard to predict are flagged as interesting. Evaluated under a realistic protocol that trains on one dataset and tests against footage from others, this approach appears highly effective. We show that this apparent success is actually a domain-shift consequence: on a fair benchmark drawn from a single dataset, this mechanism collapses to chance and is on par with simple no-training baselines. A lightly supervised probe on the same frozen embeddings results in almost double the average precision, indicating that the bottleneck is indeed the self-supervised objective, rather than the representation. We present this as a study for evaluating the effectiveness of self-supervised learning, where cross-dataset protocols can silently reward domain separation over novelty.

Robotic manipulation often requires object specific programming, manual data annotation, or calibrated perception pipelines, which limits rapid deployment in practical settings. Learning from demonstration offers a more direct alternative, but collecting demonstrations can still demand human teleoperation or kinesthetic teaching. This paper presents a self supervised visual manipulation method in which a robot automatically generates demonstrations around a target pose and learns relative pose corrections directly from wrist mounted RGB images. The proposed pipeline uses ROS~2 and Isaac Sim to collect labeled image-pose pairs without requiring explicit camera to robot extrinsic calibration. Separate datasets are generated for planar refinement and coarse three dimensional approach, and a convolutional network is trained to regress relative translation and rotation from single frame RGB observations. During execution, a coarse to fine controller first approaches the object using models trained with height variation and then refines the final alignment using planar data. The method is evaluated both in simulation and on a real UR5e collaborative robot equipped with a gripper and a monocular camera. In simulation, the refinement stage reduces the final planar dispersion from 9.69 mm to 5.38 mm. In real world experiments, the system performs end to end grasp attempts on three physical objects and reaches success rates of 66.6% and 63.6% for two objects without object rotation, while still maintaining partial robustness under rotated conditions. These results show that automatically generated demonstrations can support practical visual manipulation with limited setup effort, while also exposing remaining challenges in depth prediction and object dependent generalization.

The increasing realism of speech generated by text-to-speech and voice conversion systems poses growing challenges to media integrity and voice authentication. Self-supervised learning (SSL) has substantially advanced speech deepfake detection, where downstream backbones conventionally process SSL representations through a single forward pass. This work investigates the practical effectiveness of recurrent hierarchical reasoning for this task. We term this controlled study REIMU and systematically compare conventional single-pass backbones, weight-shared recurrence, homogeneous HRM, and heterogeneous HRM across four Base-scale SSL frontends. We further examine heterogeneous high- and low-level modules that combine self-attention with linear attention. Experiments on the ASVspoof 2019 and 2021 evaluation sets show that recurrence and hierarchical decomposition do not inherently improve detection, whereas heterogeneous operator assignment provides a more competitive configuration. Notably, the heterogeneous design remains competitive while using 10.8% fewer downstream parameters than the matched baseline, demonstrating its potential for parameter-efficient speech deepfake detection.

Graph self-supervised learning aims to learn transferable representations from large-scale unlabeled graph data. Joint-embedding predictive architectures (JEPAs) avoid explicit negative-pair construction and raw-input reconstruction by predicting masked targets directly in latent space. However, existing graph JEPAs typically rely on a single predefined graph partition, biasing the learned representations toward one structural granularity and limiting their ability to capture complementary patterns at different graph scales. To address this limitation, we propose HP-JEPA, a hierarchical partitioning framework for multi-resolution graph joint-embedding prediction. HP-JEPA organizes each graph into an ordered bank of coarse-to-fine partition resolutions and performs context-target latent prediction separately at each resolution using an online encoder, an exponential-moving-average target encoder, and a latent predictor. The resulting resolution-specific graph representations are subsequently integrated through concatenation or task-specific resolution weighting, allowing downstream models to combine complementary local, regional, and global structural information. Experiments on seven graph classification benchmarks and one graph regression benchmark show that HP-JEPA outperforms the fixed-resolution Graph-JEPA baseline on 6 of 8 tasks, improving upon Graph-JEPA on most evaluated benchmarks. Size-stratified analyses further show that HP-JEPA achieves higher accuracy than Graph-JEPA in most evaluated graph-size quartiles on three representative datasets. These results highlight the effectiveness of hierarchical multi-resolution partitioning for transferable graph representation learning.

Satellite constellations require orbital transfers that are both fuel efficient and collision avoidant. Yet, the computational cost of optimization methods traditionally used to plan their trajectories scales poorly with both the number of satellites as well as the number of obstacles to avoid, due to the pairwise safety constraints. In this work, we introduce a permutation-equivariant neural operator for trajectory planning of spacecraft swarms. This neural operator maps distributions of spacecraft initial states, target states, and obstacle initial states to trajectories which avoid collision and conserve fuel. This neural operator output is then paired with a batched Gauss-Newton finish to enforce exact orbital dynamics, and further reduce fuel use. The operator is self-supervised, trained without optimal trajectory labels. When trained on ten spacecraft, the proposed method generalized zero-shot to swarms of 1,000 spacecraft and 11,000 obstacles. The generated trajectories matched a per-agent optimal control solver's accuracy while retaining collision avoidance. Operator learning grounded in physics may offer a fast, scalable alternative to trajectory optimization in the increasingly crowded orbits of the future.

Flexible manufacturing requires rapid deployment of solutions and minimal setup time to remain competitive. An essential attribute is the ability to control error levels, as failures can range from minor performance degradation to severe equipment damage. However, conventional deployment often involves extensive setup, data collection, model training or parameter tuning, and system testing, resulting in significant delays that hinder commercial feasibility. We propose a data engine which gathers data and improves its performance while executing the task. The data engine consists of two classifiers, a fast model prediction and expensive verification. First, a model prediction is performed and based on the confidence level of the prediction, the expensive verification can be used. By adjusting the confidence level, users can control the level of tolerable error. Our method is implemented on a real-world robotic insertion task, which uses force data for the model prediction. The system applies UMAP dimensionality reduction and uses Wilson-Score to compute the confidence bounds of the prediction. Results demonstrate the ability to learn and reduce the need for expensive verifications over time, while staying within the set error-rate. The results highlight the potential of confidence bounds in self-improving models to enhance reliability in robotic classification task.

With growing privacy and portability concerns, source-free domain adaptation requires only a source pre-trained model and an unlabeled target domain, allowing for effective adaptation to the target data. Most existing self-training methods focus on selecting and exploiting samples with reliable predictions, often neglecting others. Inspired by the finding that deep models learn clean samples faster than noisy ones, we propose a domain-division based progressive learning method named DPL. Specifically, our approach consists of two alternating stages, each beginning with the division of the target domain into easy-to-adapt and hard-to-adapt subdomains based on adaptation difficulty, followed by neighborhood-based pseudo label assignment. In stage one, we enhance classification accuracy through uncertainty-aware self-training and alignment of corresponding classes between subdomains. Stage two then applies tailored learning strategies to each subdomain, starting with consistency learning on the easy-to-adapt samples and progressing to utilizing local structural information for the more challenging ones, thereby mining the intrinsic properties of the target data. Extensive experiments on several widely used benchmarks validate the effectiveness of our approach, demonstrating superior performance compared to state-of-the-art methods. Our code is available at https://github.com/iamjingli/DPL.

Data augmentations define the invariances learned by self-supervised learning (SSL). Standard augmentation pipelines were designed for natural images, yet scientific imaging modalities are governed by physical measurement processes with distinct symmetry and acquisition constraints. Enforcing invariances that contradict these constraints can distort learned representations and limit downstream performance, but practitioners moving from machine learning into a new scientific modality currently have little guidance beyond transferring natural-image pipelines unexamined. We address this gap with a principled, reproducible procedure for augmentation design in scientific SSL: we formalise the physics-aligned augmentation set as a union of measurement-consistent symmetries and acquisition-driven perturbations, and we give a concrete, largely label-free workflow---enumerate candidates, label each by the measurement operator, validate with representation-geometry diagnostics, and confirm by single-factor ablation---for selecting them. We instantiate the procedure for real-space electron microscopy and reciprocal-space 4D-STEM diffraction, and evaluate it across five SSL paradigms (DINOv2, SimCLR, MAE, VICRegL, I-JEPA) on classification and crystal-orientation regression. Physics-aligned augmentations substantially improve downstream performance for objectives relying on cross-view consistency, reduce geodesic error and improve robustness under realistic acquisition variability (detector gain, resolution loss), and systematically reshape representation geometry. While our experiments use electron microscopy, the procedure is modality-agnostic and applies to other measurement-driven domains such as medical and remote-sensing imaging. These results position augmentation design as a primary, and controllable, source of inductive bias in scientific self-supervised learning.

Effective flood monitoring is critical for minimizing the impacts of flood disasters on populations and infrastructure. Yet reliable remote sensing across extensive and environmentally diverse regions remains challenging, as most segmentation algorithms lack the generalisation capacity required for large-scale application, while annotated flood data are scarce and unevenly distributed. This study presents an end-to-end multimodal framework for Russian Federation territories sustainable flood monitoring and damage assessment based on synthetic aperture radar data, multispectral imagery, and digital elevation models with their derivatives, forming a 21-channel input. Using a self-collected multimodal dataset covering seven Russian regions, two strategies for water surface detection under limited data conditions were compared: a supervised U-Net++ model and the self-supervised AnySat architecture pre-trained and fine-tuned for the segmentation task. Under the data conditions of this study, supervised learning proved more effective, while the AnySat-based approach offered greater stability and retains advantages for settings where larger unlabelled data or missing modalities at inference are expected. The best flood area predictions were used to estimate flood impact in urban areas in terms of the area affected, material damage, casualties, and ecological and agricultural impact. The estimations were conducted following the official methodology of the Russian Ministry of Emergency Situations. Applied to the 2019 Tulun flood, the obtained results closely matched official assessments, except for material damage, due to the open-source databases usage. The results demonstrate the potential of deep learning and multimodal satellite data integration for scalable, reliable flood monitoring across diverse environmental and data-limited conditions.

Neural audio codecs (NACs) have become popular for obtaining speech representations as discrete tokens. Beyond compression, discrete tokens can be used to train self-supervised learning (SSL) models. Such models, referred to as codec-based SSL models, reduce data storage and computational cost, enabling scalable SSL pre-training. However, their language sensitivity remains unclear. When the language changes, codec-based SSL models may require retraining, which undermines their efficiency. In this paper, we present a systematic analysis of language sensitivity by varying either the NAC training language or the SSL pre-training language while keeping the other fixed. Experimental results show that downstream performance is insensitive to the NAC training language but strongly dependent on the SSL pre-training language. These findings suggest that a single NAC can be reused across languages, while aligning the SSL pre-training language with the target language is crucial.

Latent Acoustic Mapping (LAM) is a self-supervised learning method that generates high-resolution spherical acoustic maps from multichannel recordings without labelled data, matching supervised baselines on direction-of-arrival benchmarks. However, LAM degrades significantly with sparse 4-channel arrays, as the low-resolution cross-spectral matrix captures far less spatial information than the 32-channel inputs LAM was designed for. We benchmark a diverse set of upsampling architectures, spanning lightweight convolutional networks, iterative back-projection models, physics-informed networks, and generative adversarial approaches. We also study whether aligning these upsamplers with LAM by training them jointly or in different stages helps preserve the spatial structure that LAM depends on. Results show that the original full-resolution LAM is the strongest, that separately trained lightweight models are the most competitive learned approaches, and that representation alignment between the upsampler and LAM matters more than model complexity.

Single-image self-supervised denoising replaces unavailable clean targets with surrogate targets constructed from noisy observations. Its effectiveness therefore depends on how closely the surrogate objective remains aligned with supervised denoising, especially when noise is correlated, spatially nonstationary, or unknown. We express the discrepancy between a broad class of MSE-based self-supervised objectives and supervised MSE as a parameter-independent constant and a trace interaction between the surrogate-target residual and the prediction error. The corresponding gradient discrepancy is determined by the gradient of this interaction. This formulation provides a common view of paired-noise, blind-spot, weak-noise, re-corruption, and sub-image methods, while revealing that a small global interaction may conceal substantial positive and negative regional interactions through spatial cancellation. Building on these observations, we propose LoTA-N2N, a two-stage zero-shot adaptation framework. Stage 1 trains a denoiser on complementary sub-image pairs and freezes it to construct detached clean-sub-image proxies. Stage 2 estimates the residual--prediction interaction using these proxies and suppresses its patch-wise absolute magnitude. We show that the local construction prevents spatial cancellation and upper-bounds the magnitude of the corresponding global interaction. Experiments across natural, confocal, and X-ray images, complemented by iteration-matched controls, controlled noise shifts, and gradient diagnostics, show consistent gains over MSE-only adaptation under IID, spatially varying, and mixed noise. Overall, LoTA-N2N demonstrates that estimated local interaction and spatial cancellation control provide effective design principles for single-image self-supervised denoising without paired clean targets, repeated acquisitions, or a predefined re-corruption model.

Neurological disorders are a leading cause of global disability and are increasingly linked to environmental chemical exposures. Yet neurotoxicity assessment still relies on hand-scored morphological readouts that are subjective and poorly predictive of behavioral outcomes. Caenorhabditis elegans provides a genetically tractable, 3R-compliant alternative, but quantifying neuronal phenotypes from confocal microscopy at scale remains computationally challenging: existing vision foundation models, trained on natural or radiological images, cannot resolve the sparse signals and multi-scale lesions of neuronal imaging. Here, we introduce a dedicated self-supervised vision model for C. elegans dopaminergic neurons, together with CeNeuMorph, a multi-grained confocal benchmark of 27,117 annotated images. Specifically, moving beyond standard Masked Autoencoders, we propose a scale-adaptive masked image modeling strategy that jointly learns representations across resolutions and patch sizes under a fixed token budget. By decoupling structural semantic learning from rigid grid constraints, the model effectively resolves the full spectrum of neurodegenerative lesions - ranging from fine dendritic beading to gross soma shrinkage - within a tractable computational framework. Finally, our model surpasses both generalist and biomedical foundation models across classification, segmentation and detection tasks. Fusing visual features with morphological descriptors enables prediction of dopamine-dependent behavioral deficits ($R^2=0.498$). Screening 180 agrochemicals, we identify the benzimidazole moiety as a previously unrecognized determinant of dopaminergic neurotoxicity. Together, the work demonstrates how scale-adaptive self-supervised learning can connect morphology to function for a scalable alternative to mammalian in vivo models for neurotoxicity assessment and drug discovery.

The speed of sound in tissue is a prerequisite for well-focused imaging and has diagnostic value, but recovering it from raw pulse-echo channel data is fundamentally a nonlinear inverse problem. Learned solvers are fast yet label hungry. Simulated sound-speed labels are expensive, while abundant real channel data is unlabeled. We propose IQ-JEPA to exploit both data types. An encoder is pretrained without labels to predict the latent representation of masked in-phase and quadrature (IQ) regions from visible context, then fine-tuned on simulated maps. Sound speed appears in the IQ signal as a phase difference, invariant to the constant phase offset. The encoder is a Hermitian vision transformer that operates on the complex signal directly. Its attention is equivariant to that phase and its conjugate-product feed-forward is invariant to it, so the encoder reads a quantity analogous to the one classical coherence methods use. On 79,293 Fullwave 2.5 simulations at 2.5 MHz, pretraining on the 63,435 unlabeled acquisitions reaches 15.60 m/s at 10,000 labels. This is a roughly threefold gain in label efficiency over supervised training, growing to over fourfold at 1,000 labels. It is about 2.2x below an InversionNet baseline, and 8.71 m/s at full labels. The gain still grows with more unlabeled pretraining data. Our comparisons point to self-supervision as the dominant factor. The same encoder transfers. Its frozen features expose sound speed and attenuation, and cross-distribution pretraining between layered and abdominal phantoms costs little accuracy. We see this as a first step toward a foundation model for quantitative ultrasound.

Despite recent advances, self-supervised learning (SSL) models and Joint-Embedding Predictive Architectures (JEPAs) remain susceptible to learning spurious biases in the dataset. These techniques rely on regularization, which prevents representation collapse by enforcing a global target distribution such as a multivariate Gaussian or a uniform distribution on the sphere. However, these global constraints are insufficient to prevent bias entanglement, as task-irrelevant features can still segregate the latent space into distinct sub-regions. While recent approaches like Entangling and Disentangling (EnD) and Fair Supervised Contrastive Learning (FSCL) empirically debias the latent space, we show that they act as partial approximations of conditional distribution matching. To enforce this matching explicitly, we propose Unbiased Open World Regularization (UOWReg), an encoder-only framework. We show that this shift from a global to a conditional objective guarantees statistical independence between the learned representations and the targeted attributes, regardless of the chosen target distribution. We empirically validate this framework across both Gaussian and spherical latent spaces, using statistical measures to enforce these target distributions. While conditional matching successfully mitigates bias with both distributions, we demonstrate that enforcing conditional uniformity on the sphere yields a lower linearprobing classification error. Empirically, UOWReg reduces Equalized Odds violations on the CelebA benchmark while maintaining competitive classification accuracy compared to existing encoder-only baselines. Furthermore, we introduce the Synthetic Engraving Task-a novel setting in which a dominant macro-structure masks a fine-grained micro-signature. We show that UOWReg effectively prevents the subpopulation collapse observed in standard SSL, successfully isolating micro-signatures even when heavily entangled with the global structure.

Understanding motion in video is a fundamental challenge for visual learning, as frame-to-frame change entangles two sources of dynamics: camera motion and object motion. This decomposition has remained underexplored in representation learning, partly because these factors are tightly coupled in natural videos and difficult to supervise separately. Yet recovering it is important for learning robust motion representations that separate meaningful object dynamics from camera-induced variation. We study whether such structured motion representations can be recovered from frozen features of a pretrained image vision transformer. We propose the Structured Dynamics Model (SDM), which explicitly separates the dominant source of temporal change from residual dynamics through future-feature prediction, rather than representing video change with a single entangled latent or with unstructured, spatially dense transition tokens. Training combines self-supervised learning on real video with weak supervision of scene dynamics on synthetic Kubric data. We evaluate SDM on ProbeMotion, a new evaluation suite spanning synthetic and real videos with camera motion, object motion, and combined dynamics. SDM outperforms backbone baselines using global CLS or average-pooled features, and compares favorably to strongly supervised representations such as VGGT on several probes, despite using substantially weaker supervision. These results suggest that pretrained image models can be readily repurposed into structured video-dynamics representations, providing a useful inductive bias for learning and analyzing latent video dynamics.

Self-supervised foundation models have recently shown strong potential for electroencephalogram (EEG)-based analysis. However, existing approaches struggle to capture the inherently multi-scale temporal structure of EEG signals, where local neural patterns and long-range dependencies jointly encode task-relevant information. This limitation hampers cross-scale representation learning and generalization across diverse downstream tasks. To address this challenge, we propose MSBraM, a Multi-Scale self-supervised Brain foundation Model designed to learn hierarchical EEG representations. MSBraM follows a two-stage pretraining framework. First, a multi-scale neural tokenizer discretizes raw EEG signals into semantic codes at different temporal resolutions via vector-quantized reconstruction. Second, the model is pretrained to predict masked codes using a curriculum multi-scale masking strategy, progressively integrating fine-grained local patterns with global temporal context. We pretrain MSBraM on over 2,400 hours of EEG data and evaluate it across 10 downstream tasks on 12 public datasets. Extensive experiments show that MSBraM achieves superior performance on other state-of-the-art pretrained models, demonstrating strong generalization and transferability. These results indicate that explicitly modeling multi-scale temporal dynamics is critical for effective EEG foundation models.

Computer aided design (CAD) is ubiquitous: virtually any modern object was designed using editable CAD tools. However, with the shortage of available CAD datasets in its native editable and parametric format, boundary representation (B-Rep), it is ever more important to develop data-efficient methods for this domain. We present a new self-supervised pretraining task, Masked Topology Modeling (MTM), that leverages the face-adjacency graph, an induced structure unique to B-reps that the encoder can be asked to reconstruct. MTM masks a fraction of edges and trains a small head to predict each masked edge's convexity and curve type from the encoder's post-message-passing face features. We combine MTM with a MoCo-style momentum-queue contrastive learning over B-rep-aware augmentations, a BFS-connected face-region masked-reconstruction objective, and pretraining on the ABC dataset and our new procedurally generated dataset to show strong performance on a number of benchmarks.

Medical image encoders from different groups are increasingly treated as interchangeable, on the assumption that scale and clinical supervision concentrate their representations onto a shared structure. Whether this convergence is real, what produces it, and whether it is clinically usable are untested, and the similarity measures behind such claims are fragile. We present a controlled dissection across 18 image and 7 text encoders, all open-weight and run locally, spanning 7M to 27B parameters and five imaging modalities, including 650,982 chest radiographs from six datasets. To isolate cause, we train encoders that vary only the objective under fixed data, architecture, and scale, and reproduce the effect in a synthetic model. Convergence is modest but above a random floor, driven by the self-supervised objective, not clinical supervision: matched self-supervised encoders aligned most (40.4% on chest radiography), with label-supervised (21.1%) and image-text (3.3%) far lower, and did not grow with size (Spearman 0.302, p=0.223) or capability. It is within-modality, does not reach clinical language, and does not reproduce how radiologists judge case similarity. Yet a linear classifier transfers across encoders and to five held-out hospitals, retaining about 85% of within-encoder performance. Convergence in medical imaging is therefore set by the pretraining objective, not inherited from scale or clinical supervision. Interoperability is accordingly something to design for through that objective, and to validate where the shared geometry is weakest, across patient subgroups and against clinical judgment.

The training of learned inertial odometry depends on dense, high-precision position ground truth from motion capture, visual-inertial odometry or SLAM, which is costly and hard to acquire at scale. We propose DINS-IO, which learns inertial odometry directly from raw IMU streams without position labels. Our key insight is that the strapdown INS velocity recursion is a strong, fully differentiable consistency prior: the predicted velocity, rotated into the navigation frame, must agree with the integrated specific force up to an unknown initial velocity and a constant accelerometer bias. We cast this constraint as a sliding-window least-squares problem with a globally shared bias, solve it in closed form, and use the solver residual as a self-supervised loss whose gradient flows back to the network through the analytic solution. To supply this per-sample constraint, we design a high-frequency network that emits dense body-frame velocity at the IMU rate. Since the self-supervised network learns consistent motion but its velocity is not yet metrically calibrated, we calibrate it to true metric velocity from a few labeled trajectories by directly supervising the predicted body-frame velocity and adapting only low-rank (LoRA) patches. On standard benchmarks, DINS-IO pretrained self-supervised and fine-tuned with a small fraction of labels matches or surpasses fully supervised baselines.

This paper introduces Sentence Splitter, a self-supervised framework built upon a T5-based encoder--decoder architecture for uncovering the latent factual structure of natural language sentences. The proposed method identifies the semantic boundary between a descriptive prefix (head) and its factual completion (tail) by formulating sentence splitting as a discrete segmentation problem, where a sentence of length $N$ admits $N$ possible split points but only one recovers the intended head--tail structure. Rather than explicitly searching over all candidate boundaries, the model learns to recover the factual completion through probabilistic sequence generation. To eliminate the need for manual annotation, symbolic head--tail pairs are first verbalized into natural-language templates that provide supervision for training the Sentence Splitter. The trained splitter is then applied to raw text to extract aligned prefix--tail pairs, which are subsequently used to train a generative model that proposes additional plausible completions through a lightweight bootstrapping process. This unified pipeline provides a scalable and structure-aware approach to constructing self-supervised training data while bridging symbolic knowledge and natural language. Experiments on both structured and naturally occurring text demonstrate that the proposed splitter generalizes beyond synthetic templates and that the resulting structure-aware supervision consistently improves downstream performance on knowledge graph completion and commonsense question answering, highlighting the effectiveness of recovering latent factual structure for knowledge-centric NLP.

Deep neural networks have become a promising approach for IMU-based sensing, but their scalability is fundamentally limited by costly labeled data and poor robustness to heterogeneous devices, placements, and users. Existing unsupervised and self-supervised methods reduce but do not remove this dependence, still requiring labeled data for domain adaptation and largely ignoring known physical structure. We propose physical self-supervised learning, an autoencoder-style paradigm for label-free IMU sensing. We replace the conventional neural decoder with an auto-adaptive physics decoder, a learnable family of kinematic equations that enforces explicit physical structure while adapting across environments, and adopt a hybrid two-stage IMU encoder with reconstruction in a structured latent space to mitigate sensor noise. Our framework further introduces probabilistic frequency-spatial constraints to disentangle sensor and object motion, a multi-view kinematic tree to exploit sparse physical self-supervised signals, and an uncertainty-aware formulation to handle the inherent ambiguity of IMU inference. Evaluated on inertial tracking and full-body motion capture over public datasets and realistic deployments, physical self-supervised learning reduces errors by up to 5x for tracking and 4x for motion capture in challenging generalization scenarios, consistently outperforming state-of-the-art supervised and self-supervised baselines without any labels. Our code is available at https://github.com/YuyangLeng/physical-ssl-imu-label-free

Foundation models pretrained using self-supervised learning have transformed computer vision by learning transferable representations from large-scale unlabeled data. However, existing foundation models for neuroimaging remain limited by task-specific training, slice-based learning strategies, or relatively small pretraining datasets, restricting their generalizability across diverse brain MRI applications. In this work, we present BrainNext, a general-purpose self-supervised foundation model for volumetric brain MRI analysis. BrainNext combines masked autoencoder (MAE) pretraining with a native three-dimensional Bi-Directional xLSTM-UNet architecture to learn rich anatomical representations from 60,551 unlabeled brain MRI examinations spanning multiple MRI modalities. The pretrained model is subsequently adapted to downstream tasks through lightweight task-specific fine-tuning. We evaluate BrainNext on the Foundation Models for Medical Imaging (FOMO) 2025 Method Track, encompassing classification, segmentation, and brain-age estimation, where it achieved second place overall and ranked first in the meningioma segmentation task on the official FOMO 2025 challenge leaderboard, demonstrating strong transferability across heterogeneous neuroimaging tasks. These results highlight the potential of large-scale self-supervised pretraining to learn robust and transferable volumetric representations, establishing BrainNext as a scalable foundation model for diverse brain MRI applications.

Early sepsis prediction from electronic health records is challenged by irregular sampling, high missingness, and class imbalance. We systematically compare four modeling paradigms -- self-supervised Joint Embedding Predictive Architecture (JEPA) via masked latent prediction, self-supervised VICReg (variance-invariance-covariance regularization) with two-view augmentation, semi-supervised fine-tuning of a VICReg-pretrained encoder, and supervised Temporal Convolutional Network (TCN) -- alongside raw-feature baselines. All models share a common preprocessing pipeline of hourly binning with forward-fill imputation applied to 7 biomarkers selected via sparsity analysis from the MIMIC-III dataset. Our best model (JEPA + XGBoost + mean pooling) achieves AUPRC 0.636 at the time of onset (H0), approaching the SupMix benchmark (0.667) while using 83% fewer biomarkers. The Tier 1 pipeline -- VICReg pretraining followed by semi-supervised fine-tuning and XGBoost -- achieves AUPRC 0.510 at H0, a 3.1$\times$ improvement over the raw-feature baseline (0.165) and a 7.6% improvement over the end-to-end supervised TCN (0.474). Crucially, the fine-tuned VICReg encoder exhibits the most temporally persistent representations, degrading only 16.8% from H0 to H10 compared to 47.5% for supervised TCN and 65.3% for JEPA, demonstrating that self-supervised pretraining with task-aware fine-tuning yields features that are both sharp near onset and robust across prediction horizons.

Cross-modal learning, i.e., learning to predict one modality from another, is a fundamental mechanism for self-supervision via leveraging multimodality. Many practical applications, e.g., deploying a household robot, involve devices that are equipped with a rich set of sensors that enable multimodal sensing in their test environment. This presents an opportunity to apply cross-modal learning to the multimodal data sensed by these devices to learn representations. Findings in developmental psychology also suggest that biological agents leverage it to build an effective representation of their surroundings. To study this, we propose a controlled setup, where we restrict a user device to just a given test environment. It results in a specialization setup where we attempt to develop a performant model for this specific test environment. Under this setup, we develop Test-Space Training (TST), which performs multimodal data collection in the test environment and performs self-supervised pre-training on it. We evaluate these models on various downstream tasks in the same environment. Under this setup, we find various interesting insights, such as collecting rich multimodal data only from the test environment and leveraging cross-modal learning, we can achieve competitive results with generalist models (e.g., DINOv2 and CLIP) pre-trained on large-scale internet datasets. This enables an alternative scenario where the need for external Internet-scale datasets for pre-training models is reduced. We also present a set of analyses and ablations that raise intriguing points on substituting data with (multi)modality, and how varying pre-training data enables a tradeoff between a model's abilities to specialise to a test environment, and generalize to held-out spaces.

Robust and accurate neural decoders are integral to neurotechnologies such as brain-computer interfaces and closed-loop experiments. Recent work has shown that tokenizing neural data at the spike level facilitates multi-session pretraining and delivers state-of-the-art decoding performance. However, current spike-based models are restricted to supervised learning (SL), limiting training to datasets with paired behavioural labels. To address this limitation, we introduce MOJO (Masked autOencoder-based JOint training), a training framework for spike-tokenizing models that jointly leverages self-supervised learning (SSL) via masked autoencoding and SL objectives. We evaluate MOJO on three spiking datasets spanning monkey motor cortex during reaching tasks and multi-regional mouse recordings during vision and decision making tasks, demonstrating superior performance over purely SL-trained models. This improvement is especially pronounced when training with limited labelled data, particularly in few-shot finetuning, where only a small amount of labelled data from a new session is available. Incorporating SSL also yields more interpretable neuronal representations, improving performance on brain region classification and spike-statistics prediction without explicit optimization for these tasks. We further show that MOJO generalizes beyond spiking data to human electrocorticography during speech, where it continues to outperform purely SL-trained models and achieves performance comparable to neuro-foundation models (NFMs) designed specifically for continuous signals. Overall, augmenting spike-tokenizing models with SSL improves performance in label-impoverished settings and enables the use of unlabelled data across various tasks and species, while generalizing to other neural modalities. These results suggest a path towards more flexible and scalable data usage when training NFMs.

Self-supervision is a powerful technique for learning visual representations from unlabeled data. Existing techniques primarily adopt a two-stage approach for self-supervised learning (SSL): a pretraining stage on unlabeled data followed by a finetuning stage on labeled data. While this pipeline has demonstrated extreme effectiveness, the interaction between self-supervised and supervised learning objectives remains insufficiently understood. In this work, we systematically investigate whether jointly optimizing the self-supervised and supervised objectives during training provides a better alternative. We compare two training paradigms: (1) the aforementioned pretraining followed by finetuning (PFT) and (2) joint training (JT), where self-supervised and supervised losses are optimized simultaneously in the same network. Across eight representative SSL methods and diverse computer vision tasks on natural, medical, crisis response, and remote sensing data, we evaluate performance under varying percentages of labeled data. Our results reveal that the relative effectiveness of PFT and JT depends strongly on the task at hand, the availability of labeled data, and the complexity of the domain. We find that JT consistently improves data and training efficiency while being robust in low-label settings, while PFT is more reliable in more specialized domains. We further analyze representation quality, robustness, and cross-domain generalization, providing new insights into how self-supervised and supervised objectives interact during optimization. We establish a comprehensive empirical benchmark for hybrid SSL-based semi-supervised learning and offer practical guidance for selecting appropriate training strategies across diverse vision applications.

Although self-supervised learning (SSL) offers a promising way to reduce annotation effort in close-range remote sensing, its effectiveness for high-resolution multispectral unmanned aerial vehicle (UAV) imagery remains underexplored due to limited data. This study evaluated SSL pretraining for precision agriculture using cm-scale multispectral drone imagery collected across multiple sensors, years, and regions. Transformer-based encoders were pretrained with Momentum Contrast v3 (MoCo-v3) and Masked Autoencoders on a harmonized dataset combining msuav500K with newly collected multi-year UAV imagery from agricultural fields in Finland. Pretraining used four spectral bands (Green, Red, Red-Edge, Near-Infrared) for cross-sensor compatibility. The models were evaluated on crop-weed semantic segmentation using the WeedMap dataset with 5--100% training data. The following two subsets served as downstream tasks: Task A (Germany, RedEdge-M), where all pretrained models were compared under partial and full fine-tuning, and Task B (Switzerland, Sequoia), where the best encoder from Task A was assessed. Our Swin Transformer pretrained with MoCo-v3 achieved the strongest performance on both tasks, surpassing the Swin Transformer model of Doornbos et al. pretrained on a pre-release of msuav500K. Our pretrained Swin Transformer further demonstrated cross-sensor and cross-region generalization. We additionally provide a public multi-year multispectral UAV dataset from Finland to support future research.

Self-supervised learning offers a compelling approach for medical imaging, where labeled data are scarce and acquisition costs are high. We present COJEPA, a self-supervised framework for volumetric brain MRI that combines a joint-embedding predictive architecture (JEPA) with a contrastive loss (CO), targeting two complementary properties: local predictivity and global discriminability. The model is trained without labels on T1-weighted structural MRI from two cohorts (HCP-YA and AABC, $N{=}2286$, ages 22 to 90), extending I-JEPA to 3D with foreground-aware block masking, a hierarchical convolutional patch embedding, and world-space sinusoidal positional encodings. We evaluate all three objectives across zero-shot twin retrieval, brain tumor segmentation (BraTS 2024), and age regression (OpenBHB). COJEPA achieves the best monozygotic twin recall at rank@1 (0.84), the best finetuning age MAE (2.55 years on OpenBHB 3.0T), and matches CO on BraTS whole-tumor Dice, demonstrating that the combined objective yields representations that are simultaneously discriminative and locally structured.

High-quality alpha mattes are notoriously expensive to annotate, creating a fundamental data bottleneck for deep image matting. While prior work attempts to reduce annotation cost using coarser labels like trimaps or masks, they remain reliant on costly per-pixel supervision, limiting scalability and generalization. In this work, we push the boundary further and ask: can we train an automatic matting model using only RGB images, with no manual annotation at all? We answer this by presenting SSMatte, a self-supervised framework that for the first time achieves performance on par with fully-supervised automatic matting. Our key insight is to decompose the problem into semantic anchoring and detail matting. SSMatte first generates a semantic matting prompt from frozen self-supervised ViT features by propagating class-token seeds via a novel, training-efficient semantic anchoring loss based on a generalized Rayleigh quotient. This prompt then anchors a detail matting network, which is optimized via a fixed-point-based loss that enforces alpha-RGB consistency. Extensive experiments show SSMatte outperforms prior weakly-supervised methods, matches the performance of fully-supervised models on portrait benchmarks, and demonstrates favorable scaling and generalization behaviors with additional data. Our work pushes automatic matting to an fresh, fully annotation-free paradigm. Code will be available.

Melody skeleton extraction aims to derive a shorter melody that preserves structural notes while removing ornaments. Prior methods rely on hand-crafted reduction rules or note-wise salience classifiers trained with heuristically or procedurally generated pseudo-labels. Such supervision can inherit generator bias and does not explicitly optimize a coherent reduced melody. We introduce MeloBottleneck, a self-supervised framework that represents a skeleton as a length-controlled, order-preserving latent subsequence. A hard-bottleneck extractor selects note events, a rhythmic-closure operator produces a self-consistent skeleton, and a re-ornamentation decoder reconstructs the input melody. Training combines reconstruction, a frozen autoregressive melody prior, ornament-invariant consistency across procedurally ornamented views, and ornament exclusion. We evaluate three regimes: synthetic out-of-distribution ornament-to-skeleton, TAVERN variation-to-theme, and Jiugong ornamented-to-gongche. A matched pseudo-label classifier excels on the synthetic benchmark, while MeloBottleneck transfers better, achieving competitive selection quality on TAVERN and Jiugong. Skeletonized melodies also improve BM25-based fragment retrieval, boosting Recall@K and MRR while reducing query time. Overall, the results suggest that learning skeletons as latent subsequences yields more robust transfer than pseudo-label imitation.

In time-lapse fluorescence imaging, single-particle-tracking is a powerful tool to monitor the dynamics of objects of interest, and extract information about biological processes. However, tracked particles can be subject to occlusion and intermittent detectability. When these phenomena persist over a few frames, tracking algorithms tend to produce multiple tracklets for the same particle. In this work, we introduce self-supervised learning of visual features to compare tracked particles, and we exploit both visual and positional distances to robustly stitch tracklets representing the same particle. We demonstrate the performance of our stitching framework on time-lapse fluorescence sequences of Hydra vulgaris neurons. Results show high stitching precision, and reduction of errors made by previous algorithms on the same data by a factor of two.

Traditionally, continual learning has assumed access to labeled data, yet many real-world applications -- such as lifelong robotics -- require models to adapt continuously from unlabeled streams. This has led to the development of continual self-supervised learning (CSSL), a rapidly growing area that lacks a dedicated, systematic review. In this work, we present a comprehensive survey of CSSL for vision, with connections to emerging vision-language settings. First, we analyze existing evaluation protocols and highlight inconsistencies that hinder fair comparison. We then examine why self-supervised objectives exhibit improved robustness to catastrophic forgetting, relating this to task-agnostic representations and smoother loss landscapes. Next, we organize existing methods into a unified taxonomy based on their forgetting-mitigation strategies, including distillation, replay, regularization, architectural approaches, model merging, and objective-level adaptation. Finally, we identify open challenges such as scalability and the need for fast adaptability. We argue that advancing CSSL requires moving beyond small-scale benchmarks towards continual pre-training paradigms for large-scale systems.

Video object segmentation (VOS) is a fundamental task in video understanding, requiring accurate delineation and consistent tracking of objects across frames. While supervised methods achieve strong performance, they rely on densely annotated datasets that are costly to obtain and have limited domain coverage. Self-supervised learning offers a promising alternative by removing the need for manual labels; however, existing approaches often struggle to jointly maintain spatial accuracy and temporal coherence, particularly in unconstrained multi-object scenarios. Many rely on optical flow, synthetic motion cues, or task-specific pretraining, limiting scalability and generalisation. We propose a self-supervised framework, Cross-Temporal Consistency and Clustering, that learns mid-level, part-aware representations by combining attention-guided token selection with lightweight temporal clustering. Instead of operating at the pixel or whole-object level, the method aligns soft part assignments across time using a saliency-weighted symmetric consistency objective. The framework leverages a frozen transformer backbone with lightweight modules for adaptive token selection and multi-offset temporal alignment, enabling efficient scaling across resolutions and motion patterns.

The accuracy of existing leaf-wood segmentation methods for tree point clouds varies across forest types and sites. Self-supervised learning (SSL) on point clouds has improved the generalization of deep learning models for forestry point cloud tasks, including biomass regression and individual tree segmentation, but its applicability to leaf-wood segmentation remains untested. In this study, we pretrained Point-M2AE, a widely used SSL architecture for point clouds, on ShapeNet-55 augmented with 2,400 individual tree point clouds. For fine-tuning and inference, we used recursive voxel subdivision to handle the wide variation in point density across inputs, allowing the same model to operate at both individual-tree and plot scales without architecture change. Compared to the model without pretraining, the pretrained model improved wood IoU from 60.5% to 70.0% for needleleaf and from 69.7% to 76.3% for broadleaf trees. On a benchmark spanning four countries across three climatic zones, the pretrained model achieved the smallest cross-site variation and highest overall performance among compared methods (LeWos, CWLS, and PointTransformer). Plot-level segmentation maintained accuracy comparable to individual-tree performance, with mIoU of 84.7% for broadleaf and 77.7% for needleleaf plots, showing that the model generalizes across scales without additional finetuning. As a downstream test in tropical forests, where dense canopies make segmentation challenging, we applied our model and a quantitative structure model to estimate wood volume for 28 trees from Guyana, Indonesia, and Peru to assess whether the segmentation improvements from SSL pretraining translate into improved downstream performance. The resulting volume estimates achieved the lowest error among all methods tested (MAE = 2.40 m$^3$), less than half that of algorithmic baselines (LeWos: 5.94 m$^3$; CWLS: 5.27 m$^3$).