Visual Representation Learning

Robot manipulation critically depends on perception that preserves the action-relevant aspects of a scene. Yet most robot learning pipelines are built upon visual encoders pre-trained for static recognition or vision-language alignment, leaving motion understanding to downstream policies. We introduce DynaFLIP, a dynamics-aware multimodal pre-training framework that pushes motion understanding upstream into perception. We construct image-language-3D flow triplets from heterogeneous human and robot videos, and use these triplets as training-time supervision to shape an image-only encoder. Our key idea is to encourage the three modalities to span a small simplex volume in the shared hyperspherical space -- a smaller simplex volume indicating stronger alignment. To avoid the geometric ambiguity and trivial collapse of naive volume minimization, we combine simplex-volume minimization with a cosine regularizer and a contrastive objective. Our analyses show that DynaFLIP focuses on control-relevant regions critical for manipulation. The resulting dynamics-aware representations serve as reusable visual backbones and consistently outperform baselines across diverse downstream policies, including VLAs. We validate this across diverse simulation and real-world setups, with gains reaching +22.5% under out-of-distribution scenarios. Our results suggest that robot generalization improves when visual representations are trained to encode not just what is present, but how the world changes under action.

Psychovisual models suggest human vision decouples low-level feature extraction from higher cognition by first forming intermediate abstractions. In contrast, deep learning-based vision models routinely extract and aggregate features using homogeneous stacks of spatial layers, rendering their decision-making processes opaque. In this paper, we propose Deep Visual Coding, a learned frequency-domain representation inspired by 1990s image codes that quantised perceptually salient frequencies, which together with complex-valued image representations produces psychovisual-style abstractions. This approach enables the first psychovisual-based deep learning framework, utilizing data-driven spectral filters that learn to encode task-relevant semantic structures within distinct frequency sub-bands. Salience analyses reveal that our psychovisual models extract highly interpretable object parts compared to the amorphous regions produced by regular Convolutional Neural Networks (CNNs). Furthermore, we find that our models are less depth dependent than CNNs for model scaling, since our complex-valued representations and learned abstractions subsume the role of the deep spatial layers. Together, these findings demonstrate that psychovisual coding provides a promising path toward more efficient and transparent vision models.

Backpropagation is the core learning mechanism underlying deep learning. However, whether and how this algorithm is implemented in the brain remains highly debated. In particular, while forward activations of pretrained models reliably map onto the cortical hierarchy of visual processing, it is unknown whether backpropagated gradients exhibit a similar correspondence. Here, we address this question using functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG) recordings of human brain responses to natural images. For this, we extend standard encoding analyses of forward activations to map backpropagated gradients onto neural data. Focusing on a recent self-supervised vision model (DINOv3) and reproducing results on eight vision models, we find that backpropagated gradients can reliably predict both fMRI and MEG signals, specifically in higher-level visual cortex and for later latencies. However, the spatial and temporal organization of these backpropagated gradients in the brain diverges from the patterns expected under a biologically plausible backpropagation mechanism: specifically, both the order in which gradients are computed and their spatial organization diverge from the temporal and spatial hierarchies of the human brain. Together, these results suggest that, although deep networks and the brain may share similar representational content, they likely rely on fundamentally different mechanisms to learn those representations.

Discrete visual tokenizers map images to ordered sequences of tokens, providing a natural representation for structural scene descriptions. Most use a fixed sequence length, while adaptive methods often require post-hoc search or choose among a small set of rates that control the length. We propose STROP, a discrete tokenizer that learns both a visual program and its image-dependent active length. A length head is trained with a four-phase curriculum using local rate-distortion probes against frozen DINOv3 features, then predicts the active prefix in a single forward pass. At a matched rate of about $250$ nominal bits per crop, the adaptive model improves segmentation over a separately trained fixed-length baseline on four benchmarks (by $1.6$-$3.1$ mIoU), and it also beats a fixed $K{=}32$ baseline that uses more bits. STROP programs also yield higher segmentation mIoU than FlexTok, One-D-Piece, and ALIT at similar or higher rates, under the same readout architecture and training protocol. STROP therefore learns useful per-image sequence lengths without post-hoc search or a predefined set of compression rates.

Aligning structured data is a fundamental problem in computer vision and machine learning, underlying tasks such as time series analysis, human action recognition, and visual representation learning. Existing alignment methods, including Dynamic Time Warping (DTW) and its differentiable variants, rely on deterministic similarity measures and are therefore sensitive to heterogeneous and noisy features. In this work, we introduce uncertainty-aware alignment, a probabilistic framework that models pairwise correspondences with heteroscedastic uncertainty and performs structured matching along alignment paths. Our formulation, uncertainty-DTW (uDTW), assigns each correspondence a Normal distribution and parametrizes each alignment path by a Maximum Likelihood Estimate objective consisting of (i) a precision-weighted matching term that suppresses unreliable features, and (ii) a log-variance regularization that prevents degenerate solutions. This yields a probabilistic alignment mechanism that is robust to noise and interpretable, as uncertainty directly reflects the reliability of matches. We further generalize this framework from temporal sequences to tokenized visual representations, enabling structured matching over sets of visual tokens. The learned uncertainty can be interpreted as a reverse-attention: semantically relevant regions exhibit low uncertainty and dominate the alignment, while ambiguous/noisy regions have high uncertainty. This provides a connection between alignment, attention, and uncertainty modeling. We evaluate the proposed framework across diverse domains. The results demonstrate consistent improvements over state-of-the-art methods and show that learned uncertainty correlates with semantic importance. These findings establish uncertainty-aware alignment as a general, robust, and interpretable framework for learning from structured data.

Vision Transformers (ViTs) can learn strong image-level representations while their patch representations become less effective for dense prediction during prolonged training. We revisit this dense degradation phenomenon and argue that it is not fully explained by high-norm artifacts alone. Instead, we characterize \emph{semantic diffusion}: an optimization shortcut in which global semantic information spreads through patch tokens beyond what is locally justified. Our analysis shows that dense representation quality is not captured by locality alone: shallow features can remain better aligned with foreground regions yet underperform deeper features, and \texttt{[CLS]} features remain complementary for dense prediction. These observations suggest that the goal should not be to remove global context, but to make token interactions more selective. We therefore study sparse attention as a minimal intervention, replacing softmax attention with entmax-1.5 while preserving global token connectivity. On DINOv1 ViT-S/16 trained for 200 epochs on ImageNet-1K, this change preserves ImageNet linear probing accuracy and substantially improves semantic segmentation performance: VOC mIoU increases from 42.80 to 48.78, ADE20K from 19.85 to 21.97, and Cityscapes from 36.79 to 37.87. These results suggest that selective token mixing is a simple and effective bias for improving dense ViT representations.

A central question in computational vision is whether human-like visual representations are better explained by discriminative or generative learning. Existing comparisons, however, often confound the learning objective with architecture, scale, and training data, leaving open whether the objective itself drives alignment. We address this confound using Joint Energy-Based Models (JEMs), which interpolate continuously between discriminative and generative training within a fixed architecture. By varying a single mixing coefficient, we isolate the effect of the learning objective and evaluate the resulting models across six human-alignment benchmarks spanning perceptual similarity, gloss perception, human response uncertainty, robustness, shape-texture cue conflict, and diagnostic feature attribution. Across this diverse suite, human alignment is consistently maximized at intermediate points of the generative-discriminative continuum, rather than at either endpoint. Hybrid JEMs combine the categorical structure induced by discriminative learning with the sensitivity to input structure induced by generative learning, yielding more human-like behavior across multiple levels of vision. These results suggest that the generative-discriminative dichotomy is the wrong axis for understanding human-aligned vision: alignment emerges not from choosing one objective over the other, but from balancing both.

Does the relationship between learning rules and brain alignment generalize across species? We extend our prior finding that untrained CNNs match backpropagation at human V1 by testing the same five learning rules against macaque electrophysiology. The rules are backpropagation (BP), feedback alignment (FA), predictive coding (PC), spike-timing-dependent plasticity (STDP), and an untrained random-weights baseline. The macaque data come from two datasets: MajajHong2015 (V4/IT, 3,200 stimulus presentations, 88/168 neurons) and FreemanZiemba2013 (V1/V2, 135 stimuli, 102/103 neurons). Using RSA with identical model weights from our human study, we find: (1) all models achieve higher alignment with macaque early visual cortex (rho = 0.15-0.30 at V1/V2) than with human fMRI (rho = 0.01-0.08), consistent with the higher signal-to-noise ratio of electrophysiology; (2) STDP and PC produce the highest macaque V1/V2 alignment (rho ~ 0.30 and 0.28), consistent with their leading position among trained rules in human V1; (3) at IT, learning rule rankings show no detectable correlation across species (Kendall's tau = 0.00, p = 1.00), though this null result is expected given that n = 5 provides power only at tau = +/-1.0, and is further confounded by stimulus set differences; (4) a pretrained ResNet-50 (ImageNet) achieves rho = 0.25 at macaque IT, substantially above all custom CNN conditions (rho = 0.07-0.14), suggesting IT alignment is limited by model capacity and training data rather than by the learning rule. Noise ceilings, multi-seed variability (5 seeds), and a stimulus-control analysis are reported. These results demonstrate that early visual alignment is robust across species, while higher-area alignment is modulated by model capacity and stimulus domain.

The platonic representation hypothesis suggests that sufficiently large models converge to a shared representation geometry, even across modalities. Motivated by this, we ask: Can the semantic knowledge of a language model efficiently improve a vision model? As an answer, we introduce TextTeacher, a simple auxiliary objective that injects text embeddings as additional information into image classification training. TextTeacher uses readily available image captions, a pre-trained and frozen text encoder, and a lightweight projection to produce semantic anchors that efficiently guide representations during training while leaving the inference-time model unchanged. On ImageNet with standard ViT backbones, TextTeacher improves accuracy by up to +2.7 percentage points (p.p.) and yields consistent transfer gains (on average +1.0 p.p.) under the same recipe and compute. It outperforms vision knowledge distillation, yielding more accuracy at a constant compute budget or similar accuracy, but 33% faster. Our analysis indicates that TextTeacher acts as a feature-space preconditioner, shaping deeper layers in the first stages of training, and aiding generalization by supplying complementary semantic cues. TextTeacher adds negligible overhead, requires no costly multimodal training of the target model and preserves the simplicity and latency of pure vision models. Project page with code and captions: https://nauen-it.de/publications/text-teacher

Flow matching with $x$-prediction -- regressing the clean data point rather than the ambient velocity -- is known to exploit low-dimensional manifold structure effectively in pixel space \cite{li2025back}. We ask whether a pretrained representation space, while containing a low-dimensional data manifold of comparable intrinsic dimensionality, offers a distribution more favorable for flow-matching learning. Comparing pixel, SD-VAE, and DINOv2 features along four geometric axes, we find that pixel and DINOv2 share nearly identical intrinsic dimensionalities (both $\hat{d}\!\approx\!33$) yet DINOv2 exhibits $7.3\times$ higher effective rank, $35\times$ better covariance conditioning, $11.5\times$ lower excess kurtosis, and $1.7\times$ lower on-manifold interpolation error; SD-VAE latents are consistently intermediate, indicating that the advantage stems from representation-learning objectives rather than mere compression. These statistical properties render the flow-matching regression well-conditioned and remove the need for the specialized prediction heads or Riemannian transport used by prior DINOv2 diffusion methods. We propose the \emph{Representation Image Transformer} (RiT): a vanilla Diffusion Transformer trained by $x$-prediction on frozen DINOv2 features, augmented only by a dimension-aware noise schedule and joint \texttt{[CLS]}-patch modeling. On ImageNet $256{\times}256$, RiT attains FID 1.45 without guidance and 1.14 with classifier-free guidance, outperforming DiT$^\text{DH}$-XL with $19\%$ fewer parameters (676M vs.\ 839M). The resulting ODE is efficiently solvable at coarse discretizations: with classifier-free guidance, $5$ Heun steps already reach FID 2.0 and $10$ steps reach 1.25, without distillation or consistency training. Code at https://github.com/lezhang7/RiT.

Mixture-of-Experts (MoE) models are often interpreted by analysing which categories are routed to which experts. However, routing alone does not reveal what each expert actually encodes. We train sparsely-gated convolutional MoE models with a contrastive objective on natural images and characterise expert specialisation using tools from visual neuroscience. Extending from gating-level to expert-level analyses, we measure per-expert category separability, and per-expert tuning using the most exciting inputs. Extending from category-level to feature-level explanations, we interpret tuning via semantic dimensions derived from a dataset of human behavioural judgements (THINGS). Finally, we use tuning and representational similarity analysis to assess the stability of expertise-allocation across independent initialisations. We find that an animate-inanimate distinction dominates expert partitioning, apparent from gating through to expert readout, and is stable across independently trained models. Although routing statistics suggest relatively sparse, categorical preferences, expert analyses reveal broader tuning to continuous visual and semantic dimensions that extend beyond category boundaries. Experts exhibit similar category-separability to one another, despite distinct feature tuning, demonstrating the explanatory benefits of moving beyond category-level analyses. Together, these results show that expert specialisation in vision MoEs extends well beyond category routing and is better understood by probing fine-grained expert-level tuning and representational structure.

How interpretable are the features of leading vision models? The question is increasingly pressing as these models move from research benchmarks into high-stakes deployments, yet existing methods cannot answer it reliably. We close this gap with a framework for measuring and comparing the human interpretability of vision models, built around two complementary psychophysics protocols: (1) localizability -- can an observer predict where a feature fires on a novel image? -- and (2) nameability -- can an observer accurately describe what the feature represents? Features are recovered via sparse autoencoders, and a chance-anchored scoring function places every model on a common scale. Applying the framework to six vision transformers -- two supervised ViTs and four foundation models (DINOv2, DINOv3, CLIP, SigLIP) -- we collected more than $15{,}000$ behavioral responses, analyzing the $13{,}400$ responses from the $377$ participants who passed our pre-specified quality checks. Foundation models are consistently less interpretable than their supervised counterparts, and the gap is not a capability tradeoff: interpretability does not correlate with downstream task performance on any benchmark we examine. What does correlate is the locality of a feature's activations and coarse-grained semantic alignment with humans -- models with focal activations and representations that reflect the world's broad categorical structure produce more interpretable features, whereas fine-grained perceptual alignment does not. The two protocols yield strongly correlated rankings and share the same predictors, establishing interpretability as an independent, measurable dimension of representation quality -- and, surprisingly, one on which every foundation model we tested falls below the supervised baselines that came before. Capability alone cannot close that gap; locality and coarse-grained alignment can.

Visual Place Recognition (VPR) is critical for autonomous navigation, yet state-of-the-art methods lack well-calibrated uncertainty estimation. Standard pipelines cannot reliably signal when a query is ambiguous or a match is likely incorrect, posing risks in safety-critical robotics. We propose KappaPlace, a principled framework for learning uncertainty-aware VPR representations. Our core contribution is a Prototype-Anchored supervision strategy that leverages latent class representatives as targets for a probabilistic objective. By modeling image descriptors as von Mises-Fisher (vMF) variables, we learn a lightweight module to predict the concentration parameter as a direct proxy for aleatoric uncertainty. While existing VPR uncertainty methods are typically restricted to a query-centric view, we derive a novel match-level formulation to quantify the reliability of specific query-reference pairs. Across five diverse benchmarks, KappaPlace reduces Expected Calibration Error (ECE@K) by up to 50% compared to existing methods while maintaining or improving retrieval recall. We provide both a joint-training variant and a post-training extension for frozen backbones. Our results demonstrate that KappaPlace provides a robust, stable, and well-calibrated signal that enables reliable decision-making within the VPR pipeline. Our code is available at: https://github.com/mayayank95/UncertaintyAwareVPR

Non-contrastive self-supervised learning (SSL) is an effective framework for predictive representation learning, but popular (and in practice effective) methods such as SimSiam, BYOL, I-JEPA or DINO, which rely on a form of self-distillation to train a teacher-student network, remain poorly understood as they typically do not minimize a well-defined objective. We analyze the dynamics of a variant of the Joint Embedding Predictive Architecture (JEPA) using a regularized linear regressor to predict the learned representations of two views of the data from one another, and fully characterize its stability: non-collapsed stable equilibria align with leading nonlinear canonical correlation subspaces, while collapsed equilibria may also be stable attractors. Motivated by this result, we introduce PEIRA, a non-contrastive SSL method with an explicit objective defined through the trace of the optimal linear regressor. We show that its only stable equilibria are nontrivial global minimizers and recover the same canonical correlation subspaces, with regularization selecting the effective dimension. Experiments on ImageNet-1K and CIFAR-10 show PEIRA is competitive with VICReg and LeJEPA baselines, and qualitative empirical results support the theory.

Children acquire object category representations from their everyday experiences in the first few years of life. What do the inputs to this learning process look like? We analyzed first-person videos of young children's visual experience at home from the BabyView dataset ($N$ = 31 participants, 868 hours, ages 5--36 months), using a supervised object detection model to extract common object categories from more than 3 million frames. We found that children's object category exposure was highly skewed: a few categories (e.g., cups, chairs) dominated children's visual experiences while most categories appeared rarely, replicating previous findings from a more restricted set of contexts. Category exemplars were highly variable: children encountered objects from unusual angles, in highly cluttered scenes, and partially occluded views; many categories (especially animals) were most frequently viewed as depictions. Surprisingly, despite this variability, detected categories (e.g., giraffes, apples) showed stronger groupings within superordinate categories (e.g., animals, food) relative to groupings derived from canonical photographs of these categories. We found this same pattern when using high-dimensional embeddings from both self-supervised visual and multimodal models; this effect was also recapitulated in densely sampled data from individual children. Understanding the robustness and efficiency of visual category learning will require the development of models that can exploit strong superordinate structure and learn from non-canonical, sparse, and variable exemplars.

The field of deep visual recognition is undergoing a paradigm shift toward universal representations. The Platonic Representation Hypothesis suggests that diverse architectures trained on massive datasets are converging toward a shared, "ideal" latent space. This again raises a critical question: is a "Good Embedding All You Need?" In this paper, we leverage this convergence to demonstrate that off-the-shelf embeddings are inherently "good enough" for complex tasks, rendering intensive task-specific fine-tuning unnecessary. We explore this hypothesis within the few-shot learning framework, proposing a straightforward, non-parametric pipeline that entirely bypasses backpropagation. By utilizing a k-Nearest Neighbor classifier on frozen DINOv2-L features, we conduct a layer-wise characterization to identify an optimal feature extraction. We further demonstrate that manifold refinement via PCA and ICA provides a beneficial regularizing effect. Our results across four major benchmarks demonstrate that our approach consistently surpasses sophisticated meta-learning algorithms, achieving state-of-the-art performance.

Deep neural networks trained with different architectures, objectives, and datasets have been reported to converge on similar visual representations. However, what remains unknown is which visual properties models actually converge on and which factors may underlie this convergence. To address this, we decompose the object similarity structure of 162 diverse vision models into a small set of non-negative dimensions. To determine universal versus model-specific dimensions, we then estimate how often each dimension reappears across models. In contrast to model-specific dimensions, universal dimensions are more interpretable and more strongly driven by conceptual image properties, indicating the relevance of interpretability and semantic content as implicit factors driving universality across models. Differences in architecture, objective function, training data, model size, and model performance do not explain the emergence of universal dimensions. However, models with more universal dimensions also better predict macaque IT activity and human similarity judgments, suggesting that universality reflects representations relevant to biological vision. These findings have important implications for understanding the emergent representations underlying deep neural network models and their alignment with biological vision.

Vision Transformers (ViTs) learn rich visual-semantic representations through all-to-all self-attention among patch tokens. However, this design implicitly assumes that direct pairwise patch interactions are necessary for effective representation learning. In this work, we challenge this assumption and show that representations supporting both global recognition and dense prediction can be learned without direct patch-to-patch interaction. We propose VECA (Visual Elastic-Core Attention), a vision transformer with core-periphery structured attention mediated by a small set of learned cores. Patch tokens exchange global information exclusively through these cores, while the full set of dense patches are preserved and iteratively updated across layers. This reduces attention complexity from $O(N^2)$ to $O(N)$, linear in the number of patches $N$ for a fixed core budget $C$. Unlike prior latent-token cross-attention architectures, VECA facilitates sparse global communication without compressing the spatial representation itself. Nested training along the core axis further enables a single model to elastically trade off computation and accuracy at inference time without retraining. Across image classification and dense prediction tasks, VECA remains competitive with full-attention backbones and outperforms the evaluated linear-complexity alternatives on most benchmarks. Moreover, without explicit supervision, these cores develop semantically organized structures that support object-label transfer across video frames. These results show that effective visual representations can be learned without direct all-to-all patch interaction.

Learning latent representations that capture both semantic and spatial information is central to efficient spatio-semantic reasoning. However, many existing approaches rely on implicit latent structures combined with dense feature maps or task-specific heads, limiting computational efficiency and flexibility. We propose WorldComp2D, a novel lightweight representation learning framework that explicitly structures latent space geometry according to object identity and spatial proximity using multiscale local receptive fields. This framework consists of (i) a proximity-dependent encoder that maps a given observation into a spatio-semantic latent space and (ii) a localizer that infers the coordinates of objects in the input from the resulting spatio-semantic representation. Using facial landmark localization as a proof-of-concept, we show that, compared to SoTA lightweight models, WorldComp2D reduces the numbers of parameters and FLOPs by up to 4.0X and 2.2X, respectively, while maintaining real-time performance on CPU. These results demonstrate that explicitly structured latent spaces provide an efficient and general foundation for spatio-semantic reasoning. This framework is open-sourced at https://github.com/JinSeongmin/WorldComp2D.

Cross-view geo-localization (CVGL), which matches an oblique drone view to a geo-referenced satellite tile, has emerged as a key alternative for autonomous drone navigation when GNSS signals are jammed, spoofed, or unavailable. Despite strong recent progress, three limitations persist: (1) global-descriptor designs compress the patch grid into a single vector without separating layout from texture across the view gap; (2) altitude-related scale variation is retained in the learned embedding rather than marginalized; and (3) multi-objective training relies on hand-tuned scalars over losses on incompatible gradient scales. We propose SkyPart, a lightweight swappable head for patch-based vision transformers (ViTs) that institutes explicit part grouping over the patch grid. SkyPart has four theory-grounded components: (i) learnable prototypes competing for patch tokens via single-pass cosine assignment; (ii) altitude-conditioned linear modulation applied only during training, making the retrieval embedding altitude-free at inference; (iii) a graph-attention readout over active prototypes; and (iv) a Kendall uncertainty-weighted multi-objective loss whose stationary points are Pareto-stationary. At 26.95M parameters and 22.14 GFLOPs, SkyPart is the smallest among top-performing methods and sets a new state of the art on SUES-200, University-1652, and DenseUAV under a single-pass, no-re-ranking, no-TTA protocol. Its advantage over the strongest baseline widens under the ten-condition WeatherPrompt corruption benchmark.

Intermediate feature representations represent the backbone for the expressivity and adaptability of deep neural networks. However, their geometric structure remains poorly understood. In this submission, we provide indirect insights into this matter by applying a broad selection of manipulations in input space, ranging from geometric and photometric transformations to local masking and semantic manipulations using generative image editing models, and assess the feasibility of learning a mapping in the feature space, mapping from the original to the manipulated feature map. To this end, we devise different types of mappings, from linear to non-linear and local to global mappings and assess both the reconstruction quality of the mapping as well as the semantic content of the mapped representations. We demonstrate the feasibility of learning such mappings for all considered transformations. While global (transformer) models that operate on the full feature map often achieve best results, we show that the same can be achieved with a shared linear model operating on a single feature vector typically with very little degradation in reconstruction quality, even for highly non-trivial semantic manipulations. We analyze the corresponding mappings across different feature layers and characterize them according to dominance of weight vs. bias and the effective rank of the linear transformations. These results provide hints for the hypothesis that the feature space is to a first degree of approximation organized in linear structures. From a broader perspective, the study demonstrates that generative image editing models might open the door to a deeper understanding of the feature space through input manipulation.

Fine-grained image retrieval (FGIR) typically relies on supervision from seen categories to learn discriminative embeddings for retrieving unseen categories. However, such supervision often biases retrieval models toward the semantics of seen categories rather than the underlying appearance characteristics that generalize across categories, thereby limiting retrieval performance on unseen categories. To tackle this, we propose GAPan, a Generative Appearance Prior alignment network that reformulates the learning objective from category prediction toward appearance modeling. Technically, GAPan treats retrieval features with an invertible density model based on normalizing flows. In the forward direction, the flow maps all instance features into a latent density space, where each seen category is modeled by a class-conditional Gaussian prior and optimized via exact likelihood estimation. This formulation preserves richer appearance details by leveraging the invertible property of the flows. In the reverse direction, samples from the high-density regions of these learned priors are mapped back to the feature space to produce appearance-aware anchors that reflect intra-category variation. These anchors supervise a prior-driven alignment objective that aligns retrieval embeddings with category-specific appearance distributions, thereby improving generalization to unseen categories. Evaluations demonstrate that our GAPan achieves state-of-the-art performance on both widely-used fine- and coarse-grained benchmarks.

Continual learning studies how models can adapt to new tasks while retaining previously acquired knowledge. Although a broad spectrum of methods has been proposed to mitigate catastrophic forgetting, the field remains predominantly performance-driven, with limited insight into what forgetting actually corresponds to within the vision model's representation space. Prior work has primarily analyzed forgetting through task-level performance or coarse measures of representational drift, without disentangling output-level accessibility from changes in finer-grained internal structure. To this end, we propose a diagnostic framework that leverages Sparse Autoencoders (SAEs) to define a task-anchored latent feature space, enabling analysis of how task-specific information evolves at a finer granularity, where individual SAE latents are treated as concept proxies for recurring and relatively disentangled visual patterns in the model's internal computations. Within this framework, we decompose forgetting into apparent concept deletion, recoverability, and decodability. We show that a large portion of seemingly lost concept-level information can often be recovered under linearity assumption, with concept decodability degrading as more tasks are introduced. Overall, our findings suggest that a significant part of concept-level forgetting can be attributed to changes in the representational accessibility rather than complete information erasure.

Latent representations learned by neural networks often exhibit semantic structure, where concept similarity is reflected by geometric proximity in embedding space. However, comparing such spaces across models remains difficult: changes in architecture, pretraining data, objective, or random seed can yield embeddings with similar content but incompatible geometry. This latent space alignment problem is central to interpretability, transfer and multimodal learning, federated systems, and semantic communication; however, progress remains limited by the lack of large-scale, model-diverse, and metadata-rich benchmarks. To address this gap, we introduce SEMASIA, a large-scale collection of latent representations extracted from approximately 1,700 pretrained vision models across eight standard image-classification benchmarks. SEMASIA pairs embeddings with structured metadata describing architectures, training regimes, pretraining sources, and model scale. We demonstrate three applications of the resource. First, we analyze the conceptual organization of individual latent spaces, showing consistent prototype-like clustering and hierarchical semantic neighborhoods across models and datasets. Second, we benchmark supervised alignment mappings between latent spaces using reconstruction error and downstream task performance. Third, we perform a large-scale regression analysis of how pretraining-data complexity, specialization, transfer learning, augmentation, and model scale relate to geometric and probing properties of embeddings. By coupling representational scale with standardized metadata, SEMASIA provides a reproducible foundation for studying latent geometry, evaluating alignment methods, and developing next-generation heterogeneous and interoperable AI systems.

Self-supervised pretraining has become the mainstream approach for learning MRI representations from unlabeled scans. However, most existing objectives still treat each scan primarily as static aggregations of slices, patches or volumes. We ask whether there exists an intrinsic form of self-supervision signal that is different from reconstructing the masked patches, through transforming the 3D volumes into controllable 2D rendered sequences: by rendering slices at continuous positions, orientations, and scales, a 3D volume can be converted into dense video-action sequences whose controls are the action trajectories. We study this formulation with an action-conditioned pretraining objective, where a tokenizer encodes slice observations and a latent dynamics model predicts the evolution of latent features. Across representative anatomical and spatial downstream tasks, the proposed pretraining is evaluated against standard static-volume baselines, tokenizer-only pretraining, and dynamics variants without aligned actions. These results suggest that controllable MRI slice navigation provides a useful complementary pretraining interface for learning anatomical and spatial representations from large unlabeled MRI collections.

Video semantic search in densely crowded scenes remains a challenging task due to visual encoders tendency to prioritize salient foreground regions while neglecting contextually important, background areas. We propose an Inverse Attention Embedding mechanism that explicitly captures and highlights these overlooked regions. By combining inverse attention embeddings with traditional visual embeddings, our method significantly enhances semantic retrieval performance without additional training. Initial experiments and ablation studies demonstrate promising improvements over existing approaches in recall for video semantic search in crowded environments.

Domain generalization requires identifying stable representations that support reliable classification across domains. Domains may differ in low-level attributes, such as color, texture, or visual style, while preserving the same structural relationships among their underlying components. Existing methods primarily address these differences by improving the training process or aligning features across domains. However, since they leave this shared compositional structure implicit, they may overlook a more reliable source of invariance and consequently generalize less effectively to unseen domains. We propose Compositional Relational Invariance from Spatial Primitives (CRISP), an image classification framework that factors visual recognition into visual primitives and their relational composition. We represent these compositions using soft unary, binary, and ternary predicates over primitive locations and appearance, yielding differentiable measures of spatial and visual alignment that can be learned end-to-end. To learn primitives and relational structure jointly, we design an end-to-end architecture with three components: (1) a visual backbone that extracts generalized features, (2) a concept bottleneck layer that maps these features to primitive heatmaps with differentiable spatial coordinates, and (3) a structural scoring layer that evaluates candidate spatial relations among the detected primitives. Finally, we compute class probability from the joint evidence of its class-specific relational compositions and localized primitive appearance. We evaluate \method{} on five real-world image-classification datasets from the widely used DomainBed suite, covering shifts in depiction style, dataset provenance, and camera-trap location and achieving the new state-of-the-art on both benchmarks.

Unified visual tokenization faces a fundamental trade-off between high-fidelity pixel reconstruction (spatial equivariance) and semantic abstraction (conceptual invariance). We attribute this conflict to Manifold Misalignment: naive joint optimization induces opposing gradients, creating a zero-sum game between reconstruction and perception. To address this, we propose MUSE, a framework based on Topological Orthogonality. By treating Structure as an orthogonal bridge, MUSE decouples optimization within Transformers: structural gradients refine attention topology, while semantic gradients update feature values. This turns destructive interference into Mutual Reinforcement. Experiments show that MUSE breaks the trade-off, achieving state-of-the-art generation quality (gFID 3.08) and surpassing its teacher InternViT-300M in linear probing (85.2% vs. 82.5%), demonstrating that structurally aligned reconstruction can enhance semantic perception. Code is available at https://github.com/PanqiYang1/MUSE.

Artificial neural networks trained on visual tasks develop internal representations resembling those of the primate visual system, a discovery that has guided a decade of computational neuroscience. Research on building brain-aligned models has progressively embraced finer-grained learning ob- jectives, from object classification to contrastive self-supervised objectives that maximize distinc- tions among individual images. Yet the effect of learning-signal granularity on brain alignment remains largely unexamined. Here we systematically investigate how the granularity of a learning signal shapes representational alignment with human vision. We parametrically vary the number of training classes using a data-driven approach that partitions a set of training images into differ- ent numbers of categories via PCA-based splits of pretrained embeddings. We train hundreds of neural networks across convolutional and transformer architectures on these coarse classification tasks and compare their representations with human fMRI responses, macaque electrophysiology recordings, and human behavior. We find that networks trained to distinguish as few as eight broad categories learn representations that match or exceed the neural alignment of models distinguishing 1,000 classes. Even more strikingly, these coarsely trained networks align more closely with hu- man perceptual similarity judgments than all other models evaluated, including networks trained with fine-grained supervision or self-supervision as well as leading large-scale vision models. These results demonstrate that human-like visual representations can emerge from surprisingly simple learning objectives, reframing what learning signals vision may require and opening a path toward building AI systems that are more aligned with human perception.

In this paper, we explore deep learning techniques for individual identification of animals based on their skin patterns. Individual identification is crucial in biodiversity monitoring, since it enables analysis of decline or growth of populations, or intra-species interactions within populations. Models trained for the task of individual identification often do not focus on the skin pattern of animals, but on background details or body shape details. These characteristics are not individually specific, or can change drastically through time. We focus on techniques that will make machine learning models more responsive to skin pattern structure when extracting individual visual embeddings from images. For this, we explore image inpainting of task-specific masks as an auxiliary task to enhance ML-based individual identification from animal skin patterns. We propose a comparative analysis among four models as an encoder backbone for the individual identification task. We focus on the case study of zebrafish, which is a widely recognized biological model organism, and which exhibits individually identifying skin patterns. To evaluate encoder backbone performance, we present standard metrics for classification accuracy, embedding clustering metrics, and GradCAM visualizations.