Self-Supervised Learning (SSL) has led to considerable progress in Speaker Verification (SV). The standard framework uses same-utterance positive sampling and data-augmentation to generate anchor-positive pairs of the same speaker. This is a major limitation, as this strategy primarily encodes channel information from the recording condition, shared by the anchor and positive. We propose a new positive sampling technique to address this bottleneck: Self-Supervised Positive Sampling (SSPS). For a given anchor, SSPS aims to find an appropriate positive, i.e., of the same speaker identity but a different recording condition, in the latent space using clustering assignments and a memory queue of positive embeddings. SSPS improves SV performance for both SimCLR and DINO, reaching 2.57% and 2.53% EER, outperforming SOTA SSL methods on VoxCeleb1-O. In particular, SimCLR-SSPS achieves a 58% EER reduction by lowering intra-speaker variance, providing comparable performance to DINO-SSPS.
Self-supervised learning (SSL) models offer powerful representations for sound event detection (SED), yet their synergistic potential remains underexplored. This study systematically evaluates state-of-the-art SSL models to guide optimal model selection and integration for SED. We propose a framework that combines heterogeneous SSL representations (e.g., BEATs, HuBERT, WavLM) through three fusion strategies: individual SSL embedding integration, dual-modal fusion, and full aggregation. Experiments on the DCASE 2023 Task 4 Challenge reveal that dual-modal fusion (e.g., CRNN+BEATs+WavLM) achieves complementary performance gains, while CRNN+BEATs alone delivers the best results among individual SSL models. We further introduce normalized sound event bounding boxes (nSEBBs), an adaptive post-processing method that dynamically adjusts event boundary predictions, improving PSDS1 by up to 4% for standalone SSL models. These findings highlight the compatibility and complementarity of SSL architectures, providing guidance for task-specific fusion and robust SED system design.
As AI systems evolve from static tools to dynamic agents, traditional categorical governance frameworks -- based on fixed risk tiers, levels of autonomy, or human oversight models -- are increasingly insufficient on their own. Systems built on foundation models, self-supervised learning, and multi-agent architectures increasingly blur the boundaries that categories were designed to police. In this Perspective, we make the case for dimensional governance: a framework that tracks how decision authority, process autonomy, and accountability (the 3As) distribute dynamically across human-AI relationships. A critical advantage of this approach is its ability to explicitly monitor system movement toward and across key governance thresholds, enabling preemptive adjustments before risks materialize. This dimensional approach provides the necessary foundation for more adaptive categorization, enabling thresholds and classifications that can evolve with emerging capabilities. While categories remain essential for decision-making, building them upon dimensional foundations allows for context-specific adaptability and stakeholder-responsive governance that static approaches cannot achieve. We outline key dimensions, critical trust thresholds, and practical examples illustrating where rigid categorical frameworks fail -- and where a dimensional mindset could offer a more resilient and future-proof path forward for both governance and innovation at the frontier of artificial intelligence.
Existing point cloud representation learning tend to learning the geometric distribution of objects through data-driven approaches, emphasizing structural features while overlooking the relationship between the local information and the whole structure. Local features reflect the fine-grained variations of an object, while the whole structure is determined by the interaction and combination of these local features, collectively defining the object's shape. In real-world, objects undergo elastic deformation under external forces, and this deformation gradually affects the whole structure through the propagation of forces from local regions, thereby altering the object's geometric properties. Inspired by this, we propose a physics-driven self-supervised learning method for point cloud representation, which captures the relationship between parts and the whole by constructing a local-whole force propagation mechanism. Specifically, we employ a dual-task encoder-decoder framework, integrating the geometric modeling capability of implicit fields with physics-driven elastic deformation. The encoder extracts features from the point cloud and its tetrahedral mesh representation, capturing both geometric and physical properties. These features are then fed into two decoders: one learns the whole geometric shape of the point cloud through an implicit field, while the other predicts local deformations using two specifically designed physics information loss functions, modeling the deformation relationship between local and whole shapes. Experimental results show that our method outperforms existing approaches in object classification, few-shot learning, and segmentation, demonstrating its effectiveness.
Few-shot classification of hyperspectral images (HSI) faces the challenge of scarce labeled samples. Self-Supervised learning (SSL) and Few-Shot Learning (FSL) offer promising avenues to address this issue. However, existing methods often struggle to adapt to the spatial geometric diversity of HSIs and lack sufficient spectral prior knowledge. To tackle these challenges, we propose a method, Spectral-Spatial Self-Supervised Learning for Few-Shot Hyperspectral Image Classification (S4L-FSC), aimed at improving the performance of few-shot HSI classification. Specifically, we first leverage heterogeneous datasets to pretrain a spatial feature extractor using a designed Rotation-Mirror Self-Supervised Learning (RM-SSL) method, combined with FSL. This approach enables the model to learn the spatial geometric diversity of HSIs using rotation and mirroring labels as supervisory signals, while acquiring transferable spatial meta-knowledge through few-shot learning. Subsequently, homogeneous datasets are utilized to pretrain a spectral feature extractor via a combination of FSL and Masked Reconstruction Self-Supervised Learning (MR-SSL). The model learns to reconstruct original spectral information from randomly masked spectral vectors, inferring spectral dependencies. In parallel, FSL guides the model to extract pixel-level discriminative features, thereby embedding rich spectral priors into the model. This spectral-spatial pretraining method, along with the integration of knowledge from heterogeneous and homogeneous sources, significantly enhances model performance. Extensive experiments on four HSI datasets demonstrate the effectiveness and superiority of the proposed S4L-FSC approach for few-shot HSI classification.
Retinal foundation models have significantly advanced retinal image analysis by leveraging self-supervised learning to reduce dependence on labeled data while achieving strong generalization. Many recent approaches enhance retinal image understanding using report supervision, but obtaining clinical reports is often costly and challenging. In contrast, metadata (e.g., age, gender) is widely available and serves as a valuable resource for analyzing disease progression. To effectively incorporate patient-specific information, we propose PRETI, a retinal foundation model that integrates metadata-aware learning with robust self-supervised representation learning. We introduce Learnable Metadata Embedding (LME), which dynamically refines metadata representations. Additionally, we construct patient-level data pairs, associating images from the same individual to improve robustness against non-clinical variations. To further optimize retinal image representation, we propose Retina-Aware Adaptive Masking (RAAM), a strategy that selectively applies masking within the retinal region and dynamically adjusts the masking ratio during training. PRETI captures both global structures and fine-grained pathological details, resulting in superior diagnostic performance. Extensive experiments demonstrate that PRETI achieves state-of-the-art results across diverse diseases and biomarker predictions using in-house and public data, indicating the importance of metadata-guided foundation models in retinal disease analysis. Our code and pretrained model are available at https://github.com/MICV-yonsei/PRETI
Open set recognition (OSR) is devised to address the problem of detecting novel classes during model inference. Even in recent vision models, this remains an open issue which is receiving increasing attention. Thereby, a crucial challenge is to learn features that are relevant for unseen categories from given data, for which these features might not be discriminative. To facilitate this process and "optimize to learn" more diverse features, we propose GradMix, a data augmentation method that dynamically leverages gradient-based attribution maps of the model during training to mask out already learned concepts. Thus GradMix encourages the model to learn a more complete set of representative features from the same data source. Extensive experiments on open set recognition, close set classification, and out-of-distribution detection reveal that our method can often outperform the state-of-the-art. GradMix can further increase model robustness to corruptions as well as downstream classification performance for self-supervised learning, indicating its benefit for model generalization.
Reconstruction and joint embedding have emerged as two leading paradigms in Self Supervised Learning (SSL). Reconstruction methods focus on recovering the original sample from a different view in input space. On the other hand, joint embedding methods align the representations of different views in latent space. Both approaches offer compelling advantages, yet practitioners lack clear guidelines for choosing between them. In this work, we unveil the core mechanisms that distinguish each paradigm. By leveraging closed form solutions for both approaches, we precisely characterize how the view generation process, e.g. data augmentation, impacts the learned representations. We then demonstrate that, unlike supervised learning, both SSL paradigms require a minimal alignment between augmentations and irrelevant features to achieve asymptotic optimality with increasing sample size. Our findings indicate that in scenarios where these irrelevant features have a large magnitude, joint embedding methods are preferable because they impose a strictly weaker alignment condition compared to reconstruction based methods. These results not only clarify the trade offs between the two paradigms but also substantiate the empirical success of joint embedding approaches on real world challenging datasets.
Self-Supervised Learning (SSL) has become a powerful solution to extract rich representations from unlabeled data. Yet, SSL research is mostly focused on clean, curated and high-quality datasets. As a result, applying SSL on noisy data remains a challenge, despite being crucial to applications such as astrophysics, medical imaging, geophysics or finance. In this work, we present a fully self-supervised framework that enables noise-robust representation learning without requiring a denoiser at inference or downstream fine-tuning. Our method first trains an SSL denoiser on noisy data, then uses it to construct a denoised-to-noisy data curriculum (i.e., training first on denoised, then noisy samples) for pretraining a SSL backbone (e.g., DINOv2), combined with a teacher-guided regularization that anchors noisy embeddings to their denoised counterparts. This process encourages the model to internalize noise robustness. Notably, the denoiser can be discarded after pretraining, simplifying deployment. On ImageNet-1k with ViT-B under extreme Gaussian noise ($\sigma=255$, SNR = 0.72 dB), our method improves linear probing accuracy by 4.8% over DINOv2, demonstrating that denoiser-free robustness can emerge from noise-aware pretraining. The code is available at https://github.com/wenquanlu/noisy_dinov2.
A key factor in effective Self-Supervised learning (SSL) is preventing dimensional collapse, which is where higher-dimensional representation spaces span a lower-dimensional subspace. Therefore, SSL optimization strategies involve guiding a model to produce representations ($R$) with a higher dimensionality. Dimensionality is either optimized through a dimension-contrastive approach that encourages feature decorrelation or through a sample-contrastive method that promotes a uniform spread of sample representations. Both families of SSL algorithms also utilize a projection head that maps $R$ into a lower-dimensional embedding space $Z$. Recent work has characterized the projection head as a filter of irrelevant features from the SSL objective by reducing mutual information, $I(R;Z)$. Therefore, the current literature's view is that a good SSL representation space should have a high $H(R)$ and a low $I(R;Z)$. However, this view of the problem is lacking in terms of an understanding of the underlying training dynamics that influences both terms, as well as how the values of $H(R)$ and $I(R;Z)$ arrived at the end of training reflect the downstream performance of an SSL model. We address both gaps in the literature by demonstrating that increases in $H(R)$ due to feature decorrelation at the start of training lead to a higher $I(R;Z)$, while increases in $H(R)$ due to samples distributing uniformly in a high-dimensional space at the end of training cause $I(R;Z)$ to plateau or decrease. Furthermore, our analysis shows that the best performing SSL models do not have the highest $H(R)$ nor the lowest $I(R;Z)$, but arrive at an optimal intermediate point for both. We develop a method called AdaDim to exploit these observed training dynamics by adaptively weighting between losses based on feature decorrelation and uniform sample spread.
We propose a novel benchmarking methodology for graph neural networks (GNNs) based on the graph alignment problem, a combinatorial optimization task that generalizes graph isomorphism by aligning two unlabeled graphs to maximize overlapping edges. We frame this problem as a self-supervised learning task and present several methods to generate graph alignment datasets using synthetic random graphs and real-world graph datasets from multiple domains. For a given graph dataset, we generate a family of graph alignment datasets with increasing difficulty, allowing us to rank the performance of various architectures. Our experiments indicate that anisotropic graph neural networks outperform standard convolutional architectures. To further demonstrate the utility of the graph alignment task, we show its effectiveness for unsupervised GNN pre-training, where the learned node embeddings outperform other positional encodings on three molecular regression tasks and achieve state-of-the-art results on the PCQM4Mv2 dataset with significantly fewer parameters. To support reproducibility and further research, we provide an open-source Python package to generate graph alignment datasets and benchmark new GNN architectures.
While Graph Contrastive Learning (GCL) has attracted considerable attention in the field of graph self-supervised learning, its performance heavily relies on data augmentations that are expected to generate semantically consistent positive pairs. Existing strategies typically resort to random perturbations or local structure preservation, yet lack explicit control over global structural consistency between augmented views. To address this limitation, we propose Fractal Graph Contrastive Learning (FractalGCL), a theory-driven framework that leverages fractal self-similarity to enforce global topological coherence. FractalGCL introduces two key innovations: a renormalisation-based augmentation that generates structurally aligned positive views via box coverings; and a fractal-dimension-aware contrastive loss that aligns graph embeddings according to their fractal dimensions. While combining the two innovations markedly boosts graph-representation quality, it also adds non-trivial computational overhead. To mitigate the computational overhead of fractal dimension estimation, we derive a one-shot estimator by proving that the dimension discrepancy between original and renormalised graphs converges weakly to a centred Gaussian distribution. This theoretical insight enables a reduction in dimension computation cost by an order of magnitude, cutting overall training time by approximately 61%. The experiments show that FractalGCL not only delivers state-of-the-art results on standard benchmarks but also outperforms traditional baselines on traffic networks by an average margin of about remarkably 7%. Codes are available at (https://anonymous.4open.science/r/FractalGCL-0511).
Auto-regressive partial differential equation (PDE) foundation models have shown great potential in handling time-dependent data. However, these models suffer from the shortcut problem deeply rooted in auto-regressive prediction, causing error accumulation. The challenge becomes particularly evident for out-of-distribution data, as the pretraining performance may approach random model initialization for downstream tasks with long-term dynamics. To deal with this problem, we propose physics-informed temporal alignment (PITA), a self-supervised learning framework inspired by inverse problem solving. Specifically, PITA aligns the physical dynamics discovered at different time steps on each given PDE trajectory by integrating physics-informed constraints into the self-supervision signal. The alignment is derived from observation data without relying on known physics priors, indicating strong generalization ability to the out-of-distribution data. Extensive experiments show that PITA significantly enhances the accuracy and robustness of existing foundation models on diverse time-dependent PDE data. The code is available at https://github.com/SCAILab-USTC/PITA.
Event cameras deliver visual data with high temporal resolution, low latency, and minimal redundancy, yet their asynchronous, sparse sequential nature challenges standard tensor-based machine learning (ML). While the recent asynchronous-to-synchronous (A2S) paradigm aims to bridge this gap by asynchronously encoding events into learned representations for ML pipelines, existing A2S approaches often sacrifice representation expressivity and generalizability compared to dense, synchronous methods. This paper introduces EVA (EVent Asynchronous representation learning), a novel A2S framework to generate highly expressive and generalizable event-by-event representations. Inspired by the analogy between events and language, EVA uniquely adapts advances from language modeling in linear attention and self-supervised learning for its construction. In demonstration, EVA outperforms prior A2S methods on recognition tasks (DVS128-Gesture and N-Cars), and represents the first A2S framework to successfully master demanding detection tasks, achieving a remarkable 47.7 mAP on the Gen1 dataset. These results underscore EVA's transformative potential for advancing real-time event-based vision applications.
Recent advances in self-supervised learning (SSL) have revolutionized computer vision through innovative architectures and learning objectives, yet they have not fully leveraged insights from biological visual processing systems. Recently, a brain-inspired SSL model named PhiNet was proposed; it is based on a ResNet backbone and operates on static image inputs with strong augmentation. In this paper, we introduce PhiNet v2, a novel Transformer-based architecture that processes temporal visual input (that is, sequences of images) without relying on strong augmentation. Our model leverages variational inference to learn robust visual representations from continuous input streams, similar to human visual processing. Through extensive experimentation, we demonstrate that PhiNet v2 achieves competitive performance compared to state-of-the-art vision foundation models, while maintaining the ability to learn from sequential input without strong data augmentation. This work represents a significant step toward more biologically plausible computer vision systems that process visual information in a manner more closely aligned with human cognitive processes.
Self-supervised learning shows promise in harnessing extensive unlabeled data, but it also confronts significant privacy concerns, especially in vision. In this paper, we perform membership inference on visual self-supervised models in a more realistic setting: self-supervised training method and details are unknown for an adversary when attacking as he usually faces a black-box system in practice. In this setting, considering that self-supervised model could be trained by completely different self-supervised paradigms, e.g., masked image modeling and contrastive learning, with complex training details, we propose a unified membership inference method called PartCrop. It is motivated by the shared part-aware capability among models and stronger part response on the training data. Specifically, PartCrop crops parts of objects in an image to query responses within the image in representation space. We conduct extensive attacks on self-supervised models with different training protocols and structures using three widely used image datasets. The results verify the effectiveness and generalization of PartCrop. Moreover, to defend against PartCrop, we evaluate two common approaches, i.e., early stop and differential privacy, and propose a tailored method called shrinking crop scale range. The defense experiments indicate that all of them are effective. Finally, besides prototype testing on toy visual encoders and small-scale image datasets, we quantitatively study the impacts of scaling from both data and model aspects in a realistic scenario and propose a scalable PartCrop-v2 by introducing two structural improvements to PartCrop. Our code is at https://github.com/JiePKU/PartCrop.
Accurate and robust localization is a critical enabler for emerging 5G and 6G applications, including autonomous driving, extended reality (XR), and smart manufacturing. While data-driven approaches have shown promise, most existing models require large amounts of labeled data and struggle to generalize across deployment scenarios and wireless configurations. To address these limitations, we propose a foundation-model-based solution tailored for wireless localization. We first analyze how different self-supervised learning (SSL) tasks acquire general-purpose and task-specific semantic features based on information bottleneck (IB) theory. Building on this foundation, we design a pretraining methodology for the proposed Large Wireless Localization Model (LWLM). Specifically, we propose an SSL framework that jointly optimizes three complementary objectives: (i) spatial-frequency masked channel modeling (SF-MCM), (ii) domain-transformation invariance (DTI), and (iii) position-invariant contrastive learning (PICL). These objectives jointly capture the underlying semantics of wireless channel from multiple perspectives. We further design lightweight decoders for key downstream tasks, including time-of-arrival (ToA) estimation, angle-of-arrival (AoA) estimation, single base station (BS) localization, and multiple BS localization. Comprehensive experimental results confirm that LWLM consistently surpasses both model-based and supervised learning baselines across all localization tasks. In particular, LWLM achieves 26.0%--87.5% improvement over transformer models without pretraining, and exhibits strong generalization under label-limited fine-tuning and unseen BS configurations, confirming its potential as a foundation model for wireless localization.
Masked image modeling is one of the most poplular objectives of training. Recently, the SparK model has been proposed with superior performance among self-supervised learning models. This paper proposes a new mask pattern for this SparK model, proposing it as the Mesh Mask-ed SparK model. We report the effect of the mask pattern used for image masking in pre-training on performance.
The success of deep learning in computer vision over the past decade has hinged on large labeled datasets and strong pretrained models. In data-scarce settings, the quality of these pretrained models becomes crucial for effective transfer learning. Image classification and self-supervised learning have traditionally been the primary methods for pretraining CNNs and transformer-based architectures. Recently, the rise of text-to-image generative models, particularly those using denoising diffusion in a latent space, has introduced a new class of foundational models trained on massive, captioned image datasets. These models' ability to generate realistic images of unseen content suggests they possess a deep understanding of the visual world. In this work, we present Marigold, a family of conditional generative models and a fine-tuning protocol that extracts the knowledge from pretrained latent diffusion models like Stable Diffusion and adapts them for dense image analysis tasks, including monocular depth estimation, surface normals prediction, and intrinsic decomposition. Marigold requires minimal modification of the pre-trained latent diffusion model's architecture, trains with small synthetic datasets on a single GPU over a few days, and demonstrates state-of-the-art zero-shot generalization. Project page: https://marigoldcomputervision.github.io
Scaling up model and data size have demonstrated impressive performance improvement over a wide range of tasks. Despite extensive studies on scaling behaviors for general-purpose tasks, medical images exhibit substantial differences from natural data. It remains unclear the key factors in developing medical vision foundation models at scale due to the absence of an extensive understanding of scaling behavior in the medical domain. In this paper, we explored the scaling behavior across model sizes, training algorithms, data sizes, and imaging modalities in developing scalable medical vision foundation models by self-supervised learning. To support scalable pretraining, we introduce BioVFM-21M, a large-scale biomedical image dataset encompassing a wide range of biomedical image modalities and anatomies. We observed that scaling up does provide benefits but varies across tasks. Additional analysis reveals several factors correlated with scaling benefits. Finally, we propose BioVFM, a large-scale medical vision foundation model pretrained on 21 million biomedical images, which outperforms the previous state-of-the-art foundation models across 12 medical benchmarks. Our results highlight that while scaling up is beneficial for pursuing better performance, task characteristics, data diversity, pretraining methods, and computational efficiency remain critical considerations for developing scalable medical foundation models.