Vision Transformer

Recent findings indicate that Vision Transformers settle into locally similar computational phases, implying a level of depthwise computational redundancy. However, existing methods to exploit this redundancy either fail to reduce inference compute or severely degrade model expressivity. In this work, we formalise a unified view of block redundancy that decouples the geometry from specific surrogate interventions. We then introduce Transformer-Within-Transformer (TWT), a post-hoc method that fuses contiguous groups of redundant layers into a single learned surrogate layer. TWT reduces parameter count and inference compute while remaining competitive with original models using half the depth on natural images, and in several downstream histopathology settings, TWT matches or even improves on the original baseline.

Scaling deep learning faces critical bottlenecks: data exhaustion, exponential training costs, and resource concentration. Model merging combines pre-trained checkpoints without gradient descent, offering orders-of-magnitude savings versus retraining. Combining independently trained vision models is difficult when their architectures and parameter shapes differ. Existing weight-space merging methods generally assume aligned, shape-compatible checkpoints, whereas a Vision Transformer (ViT) and a state-space model (SSM) implement token mixing with different operators. We study a hybrid Heterogeneous merging setting that retains both architectures while aligning parameter groups by semantic role. Our proposed Riemannian--Lorentz Parameter Fusion (RLPF) method projects aligned groups to common coordinates, lifts selected coordinates to the Lorentz hyperboloid model of hyperbolic space, computes a regularized geodesic barycenter, and decodes the result into the two branches. A learned gate then combines branch logits for each input. Component groups use fixed curvature values, with normalization parameters treated as Euclidean. In the results available in this manuscript, the fine-tuned system obtains 82.37% on CIFAR-10, 75.04% on Oxford-IIIT Pet, and 78.58% top-1 accuracy on ImageNet-1K; the corresponding best-parent accuracies are 76.54%, 71.42%, and 76.42%. On ImageNet-1K, the reported pre-fine-tuning initialization reaches 77.80%. These results support further study of geometry-aware heterogeneous fusion, but not a training-free single-checkpoint merge: RLPF is a two-branch hybrid whose gate and reported final models are trained.

Vision Transformers (ViTs) are central to most modern vision models, yet obtaining input attributions that are fine-grained, faithful, and stable remains challenging. Layer-wise Relevance Propagation (LRP) has been adapted to transformer attention, but in ViTs it often produces noisy, unfaithful explanations. We show that the missing ingredient is the treatment of residual connections: cancellation effects in residual pathways lead to attribution explosion. Moreover, we find that these cancellations are substantially stronger in ViTs than in language transformers. To address this issue, we introduce Residual-aware Layer-wise Relevance Propagation (ResLRP), a simple extension of LRP whose propagation rules explicitly account for cancellations in residual branches, are exactly conservative, and provably bound relevance explosion. Causal channel-wise interventions confirm that residual cancellation, not a generic regularization effect, drives the instability. ResLRP substantially improves attribution quality across faithfulness and localization, evaluated on ViT architectures spanning supervised, self-supervised, contrastive, hierarchical, and multimodal families, as well as on the ground-truth-controlled FunnyBirds benchmark. The largest gains arise in modern Vision Language Models (VLMs), with +27-29% localization and up to 3.4x faithfulness scores. Beyond benchmarks, ResLRP localizes Sparse Autoencoder (SAE) features in input space, and our residual amplification measure serves as an architecture-level diagnostic predicting where attribution degrades.

Wheat streak mosaic virus (WSMV) is a destructive pathogen of sweet corn and other cereal crops, causing yield losses and complicating early detection because symptoms are spatially variable and subtle. In sweet corn seed production, WSMV also has regulatory importance, as phytosanitary regulations from countries such as New Zealand and Chile require seed lots to be certified virus-free. Visual scouting is unreliable because symptoms can resemble abiotic stress, while enzyme-linked immunosorbent assay (ELISA) is accurate but expensive, labor-intensive, and difficult to scale. We present an automated pipeline for plant-level WSMV detection using unmanned aircraft systems (UAS) multispectral imagery. The framework integrates orthomosaic reconstruction, geospatial alignment, plant extraction, and classification using a Vision Transformer with seven-channel inputs (five spectral bands, NDVI, and NDRE). Using treatment-based labels, the model achieved 89% accuracy on over 6,500 test patches across multiple growth stages. However, ELISA-based ground truth revealed substantial label noise: only a small fraction of sampled plants in inoculated plots were infected. Treatment labels therefore did not reliably represent infection status, and the high accuracy was largely driven by label bias rather than disease detection. Performance decreased markedly against row-level symptom severity and plant-level ELISA labels. Under these higher-fidelity but smaller-sample conditions, both deep learning and classical machine learning showed limited generalization and weak separability between ELISA-confirmed mock-inoculated and infected plants. These results show that UAS-based disease detection is constrained by label fidelity and data availability, emphasizing biologically grounded labels and models aligned with real-world conditions.

State Space Models (SSMs), particularly VMamba, have emerged as efficient alternatives for modeling long-range dependencies in medical image analysis. However, distinguishing subtle pathological features from visually similar anatomical backgrounds remains a significant challenge. Existing SSM architectures often learn entangled representations, lacking explicit mechanisms to separate disease-specific signals from normal anatomy. To address this limitation, we propose Spatial-Contextual Differential Mamba (SCDM), an asymmetric dual-branch architecture designed for selective representational disentanglement. SCDM introduces a Positive Branch for extracting discriminative features and a Negative Branch that actively models and suppresses normal anatomical context. This separation is achieved through a similarity-driven repulsion gate and a differential inference rule, which promote competitive feature learning without requiring additional branch labels or increasing model capacity. Evaluated on the RSNA Pneumonia dataset, SCDM achieves competitive classification performance (AUC of 0.858) while requiring significantly fewer parameters (29.4M) and FLOPs (1.44G) compared to standard VMamba and vision transformer baselines. Furthermore, activation analyses demonstrate that our differential mechanism yields highly precise localization, effectively isolating lesions by inhibiting irrelevant anatomical distractors.

Texture recognition remains challenging for modern vision models because discriminative evidence is often carried by higher-order spatial statistics rather than by object shape alone. While Vision Transformers provide strong long-range modeling capacity, their standard object-centric representations do not explicitly expose such statistical structure, which limits texture sensitivity in fine-grained recognition settings. We present HiPerViT, a compact vision-only architecture that injects an explicit second-order statistical prior into a transformer-based recognition pipeline. The method combines global and local image views with a compact bilinear descriptor encoded as a statistical token, and integrates this token with first-order spatial representations through Perceiver-style latent distillation. This design enables direct interaction between spatial tokens and second-order feature co-occurrence statistics, providing the model with explicit access to texture-relevant information without requiring multimodal pretraining or ensemble construction. Across six texture recognition benchmarks, HiPerViT achieves consistent improvements over strong vision-only baselines under the reported evaluation protocols, including gains of +3.05 percentage points on DTD, +10.48 on GTOS-Mobile, and +10.10 on 1200Tex. Beyond benchmark performance, our analyses show that these gains are largely invariant to the backbone depth used to extract second-order statistics and to the ordering of interaction and distillation stages. This pattern suggests that the primary source of improvement is not a specific fusion topology, but the explicit availability of second-order statistical information as a first-class representational signal. These results support explicit statistical tokenization as an effective and robust design principle for texture-centric visual recognition.

Accurate plant growth estimation is essential for greenhouse monitoring, yet obtaining labeled data remains costly and time-consuming. To address this, we propose a few-shot regression framework that combines Vision Transformer (ViT) feature embeddings, clustering-based task construction, and gradient-based meta-learning, and show that task construction in embedding space is a primary driver of performance. The approach leverages an unlabeled image pool to organize data into structured tasks using fuzzy c-means clustering, enabling efficient learning from a small number of labeled samples. We systematically evaluate meta-learning methods and show that second-order methods (e.g., Model-Agnostic Meta-Learning variants such as MAML++) outperform classical baselines in the few-shot regime. Furthermore, intra-cluster support selection has a limited and dataset-dependent impact. Experiments on two plant datasets show that structured task design combined with meta-learning enables reliable plant growth estimation under severe label scarcity.

Vision transformers (ViTs) have achieved remarkable generalization across visual domains, yet little is known about how they internally represent the structure of the world. To address this gap, we use Cross-Layer Transcoders (CLTs) to read concept circuits from ViTs: directed graphs whose nodes correspond to sparse, interpretable concepts and edges capture concept interactions across layers. Our method yields two complementary views of model behavior. The global concept circuit is input-invariant and can be recovered directly from learned cross-layer weights, exposing the reusable "world knowledge" encoded in the model. The instance concept circuit is input-dependent and identifies the concepts and pathways actually used for a specific prediction, enabling faithful example-level explanations. We demonstrate the utility of concept circuits in three ways: (1) Automatic spurious correlation discovery: leveraging the statistics of our global concept circuits to identify shortcut dependencies within the model. (2) Spurious correlation removal: intervening on the instance concept circuit to steer the model towards correct predictions. Empirical results show that our method outperforms existing counterparts by 11.0% on the Waterbird dataset. (3) Model comparison: contrasting the global concept circuits of different foundation models (e.g., CLIP vs. DINO) to reveal how supervision paradigms shape representational structure. Our code is available at https://github.com/deep-real/VisionCLT

Model compression is key to mitigate deployment challenges of ever growing machine learning models. In this area of research, singular value decomposition (SVD)-based compression offers a compelling trade-off between computational efficiency and model accuracy. Fisher-weighted SVD in particular provides principled, loss-aware compression. However, we find that improving the fidelity of Fisher approximation used in the compression is poorly predictive of post-compression accuracy for Vision Transformers (ViTs). Motivated by this observation, we propose FACTS, a structured Fisher Approximation tailored to Compressing ViTs with Fisher-weighted SVD, which enforces token-local aggregation while preserving within-token activation-gradient dependence. Additionally, we introduce a fast Constrained Rank Search (CoRS), that optimizes layer-wise rank allocation while adhering to a fixed floating point operation (FLOP) constraint. Extensive experiments across ViTs and hybrid architectures demonstrate that FACTS consistently improves accuracy-efficiency trade-offs without requiring finetuning. Notably, it outperforms the strongest SVD baseline by up to +5.8 percentage points (p.p.) Top-1 on Swin-B, with further gains driven by our search method. Code is available at https://github.com/MoritzTho/FACTS.

Vision Transformers (ViTs) typically process every image using a fixed input resolution and model width, even though many images can be classified with substantially less computation. We introduce ProgResViT, an input-adaptive ViT that performs inference progressively across multiple rounds. The first round processes a low-resolution image with a narrow subnetwork. Inference terminates when the prediction is sufficiently confident; otherwise, the model reuses the representations produced in the current round and proceeds with a higher-resolution input and a wider subnetwork to refine its prediction. As all rounds share a single backbone, we propose Progress-Conditioned Soft Gating (PSG), which conditions token fusion and layer outputs on the current round, block, and input resolution. On image classification, applying ProgResViT to DeiT yields better accuracy-compute trade-offs than adaptive-width, adaptive-depth, and dynamic-token baselines. With knowledge distillation, a DeiT-based ProgResViT achieves 84.9% top-1 accuracy, slightly exceeding the reported DeiT-III-S accuracy under a comparable evaluation setting. We show that the same design also provides favorable accuracy-compute trade-offs for self-supervised DINO representations and downstream semantic segmentation. Code is available at https://github.com/ds-kiel/ProgResViT.

Vision Transformers (ViT) excel in semantic understanding but fail to discriminate between object instances (e.g., identical embeddings for two dogs), limiting their use in instance-level tasks such as object detection and instance segmentation. We propose Contrastive Vision Transformer (CoViT), a self-supervised learning framework that injects instance-awareness into ViT through geometry-guided contrastive learning. CoViT uniquely coordinates ViT's attention maps and embeddings by constructing triplets: (1) Attention-guided masking: Refine multi-head attention via adaptive thresholding and morphological operations to generate instance masks, identifying foreground anchors; (2) Hardest contrastive mining: For each anchor, computing pairwise embedding similarities to select the intra-instance hardest positive (least similar patch within its mask) and inter-instance hardest negative (most similar patch from other instances), with intra-instance regions masked during negative search. These triplets drive a contrastive loss that simultaneously compresses intra-instance variance and expands inter-instance margins, forcing ViT to discern subtle geometric and appearance differences between instances. CoViT consistently achieves stable performance gains of over 2 AP points across multiple instance-level perception tasks by using ViT as backbone architecture. Notably, CoViT requires no extra decoders or labels, demonstrating that a pure ViT can learn instance-aware representations via inherent attention priors and targeted contrastive constraints. Code and models will be released.

Vision Transformer (ViT) design has become increasingly diverse, with backbones combining convolutional stems, windowed, linear, or multi-axis attention, patch merging, and spatial reduction in various configurations. This diversity poses challenges for existing attribution methods, whose assumptions often do not hold across ViT variants: Grad-CAM requires a terminal spatial feature map, attention rollout assumes global softmax attention, and layer-wise relevance propagation (LRP) requires module-specific rules. To the best of our knowledge, no existing method provides a unified attribution framework across this architectural space. We show that this architectural diversity can be captured by a simpler underlying structure. The attention and resolution-reduction operators in current ViTs can be decomposed into four operation types: linear maps, bilinear mixing, normalization or gating, and reindexing. Each operation admits a relevance rule that satisfies conservation. Based on these rules, HiLRP supports new backbones by construction rather than by architecture-specific derivation, and its attribution maps decompose the prediction rather than relying on heuristic assumptions. We prove conservation and conditional equivariance and verify both to machine precision. Across 14 attribution methods and 10 architectures, we find that no prior method remains reliable across ViT families, while Faithfulness Correlation becomes uninformative for backbones robust to spatial masking. HiLRP alone preserves conservation across windowed, spatial-reduction, multi-axis, and linear-attention models, where naive extensions can produce zero or inflated relevance. It also localizes attribution failures in class activation mapping, achieving 0.97 Pointing compared with 0.55 for competing methods on EfficientViT.

Vision Transformers (ViTs) are increasingly used in split-inference systems, where edge devices transmit intermediate token representations to a remote cloud. In this setting, token reduction lowers computation and communication costs, while token shuffling disrupts the spatial organization of the transmitted tokens, potentially limiting information leakage. However, their privacy benefits remain unclear against feature inversion attacks, which attempt to reconstruct the input from the transmitted embeddings. In this work, we show that, despite disrupting the spatial structure required by conventional reconstruction attacks, transmitted token embeddings retain substantial positional information. Based on this observation, we introduce the Spatially Aligned Reconstruction Attack (SARA), a unified pipeline that predicts token positions, restores their spatial layout, reconstructs missing embeddings using a feature-space masked autoencoder, and recovers the input image. Our results demonstrate that token shuffling provides only apparent privacy, as SARA largely reconstructs the original token organization. Token reduction offers stronger protection, but significant leakage persists when the retained tokens preserve sufficient semantic and positional information. Finally, we introduce a lightweight edge-side defense that removes positional embeddings and progressively adapts the edge-side transformer blocks through knowledge distillation. It substantially reduces attack performance against SARA, while preserving downstream task accuracy and requiring no changes to the cloud-side model.

Vision Transformers provide strong visual representations but typically rely on slowly updated parameters, limiting their ability to organize newly acquired information across different memory timescales. This work proposes \textit{Hierarchical Hebbian Memory}, a three-level memory architecture composed of rapid Working Memory, persistent Routed Episodic Memory, and slower Semantic Memory. A learned controller regulates memory contribution, read and write routing, plasticity, retention, and consolidation. A causal read-before-write lifecycle ensures that the current outcome cannot influence the prediction it supervises. The architecture is evaluated on Omniglot 5-way 1-shot recognition and CORe50 continual object recognition. With Swin-Tiny, the hierarchical model reaches 97.39% accuracy on Omniglot and 95.37% final accuracy on CORe50 when combined with experience replay. Learned multi-bank retrieval reaches 47.50% delayed-association accuracy, compared with 24.17% for a single persistent bank and 25.00% without memory. After intervening distractors, Episodic Memory retains approximately 0.96 cosine similarity with stored associations, while Working Memory falls to approximately 0.05. These results show that Hebbian association and learned memory routing can jointly organize online visual experience across rapid, persistent, and consolidated memory timescales within Vision Transformers.

Clinical AI often optimizes predictive performance without engaging how clinicians decide where to look and what to write. We present Co-Annotator, which distills expert gaze and dictation into two guidance components: a gaze-aligned Vision Transformer producing fixation-aligned areas of interest (AOIs), and an ontology-bounded vision-language model (VLM) that pre-fills editable biomarker summaries for retinal optical coherence tomography (OCT). We first collect expert gaze and dictations (US1) to train the models, significantly improving diagnostic accuracy and biomarker generation. We then deploy the system with ophthalmology residents: a controlled resident study (US2) confirmed each modality is safe and independently beneficial, with AOI guidance producing lasting perceptual efficiency gains through post-guidance carryover and VLM guidance more than doubling biomarker documentation breadth. In a combined deployment across two academic institutions (US3), providing both modalities simultaneously produced efficiency gains that substantially exceeded either modality alone: correct diagnoses per minute increased by 40% and comment editing time fell by 67%, without compromising diagnostic accuracy. Notably, neither modality improved efficiency during guidance in US2, which makes the in-guidance efficiency gain under combined guidance in US3 the more striking result. Expert-distilled multimodal guidance can remove two distinct clinical workflow bottlenecks at once (visual search overhead and documentation burden) without compromising the diagnostic accuracy clinicians already achieve.

Joint-Embedding Predictive Architectures (JEPAs) for world modeling typically employ fixed-size Vision Transformer encoders that are over-provisioned for simple tasks and under-provisioned for complex ones, with significant redundancy across attention heads. We propose Successive Capacity Growth (SCG), a method that starts from a minimal encoder (1 head, 2 layers, 283K parameters) and grows incrementally in width (adding attention heads for low-level semantic capacity) or depth (adding transformer blocks for higher-order semantic abstraction), driven by a task-agnostic test-and-verify mechanism that exploits function-preserving expansion to safely trial architectural changes and roll back if they do not improve prediction loss. The Sketched Isotropic Gaussian Regularizer (SIGReg) ensures that all learned semantic dimensions remain statistically independent and aligned with the predictive objective, preventing collapse even as the architecture grows. On a 60-dimensional multi-object dynamics task, SCG naturally triggers depth expansion, improving prediction loss by 20.3% over the fixed small baseline with 56 times greater parameter efficiency than scaling to the fixed large model; on a 2D navigation task, a single width expansion yields even an 23% improvement over the fixed large model. Across all three tested environments of increasing complexity, the adaptive encoder matches or exceeds the fixed small baseline, with zero false-positive expansions and bit-exact function preservation (ratio = 1.0, absolute difference = 0.0). The take-away is that JEPA world model encoders need not be pre-allocated at maximum capacity - they can grow successively as the task demands, achieving significant compute and data efficiency while maintaining representation quality.

Vision Transformers (ViTs) demonstrate exceptional performance in computer vision but suffer from large parameter counts and quadratic computational complexity, severely limiting their deployment on resource-constrained edge hardware. While recursive weight-sharing reduces parameter counts and token merging mitigates computational and memory bottlenecks, integrating these two paradigms without costly retraining is non-trivial, leaving this intersection largely unexplored. We propose MergeOver, a post-training approach that integrates Token Merging (ToMe) into the recursively weight-shared Sliced Recursive Transformer (SReT). Through an Unmerge tracking stack, constraint-safe merge-rate adjustment, and synchronised token-mass tracking across spatial permutations, MergeOver resolves the spatial and merging constraints of this integration. We further employ a stage-wise single-shot schedule that performs token reduction at the first block of each stage and maintains a fixed sequence length throughout its subsequent recursive iterations. Benchmarked on ImageNet-1K, our selected configuration reduces top-1 accuracy by 1.47 percentage points. On the GPU, it reduces peak activation memory by 37.3% and 38.4% at batch sizes 1 and 16, while throughput decreases by 21.7% at batch size 1 but increases by 21.7% at batch size 16. On a Raspberry Pi 5 (ARM CPU), it reduces latency by 2.4% and 17.6% at batch sizes 1 and 16. These results show that MergeOver can recover a meaningful part of the throughput and memory cost that recursive weight-sharing introduces, without retraining, and provides a baseline for combining token merging with hierarchical recursive transformers.

This paper addresses the challenge of multi-label defect classification in electroluminescence (EL) images of photovoltaic (PV) cells. Training models on images where multiple defects co-occur creates learning ambiguity, making it difficult to disentangle visual features for specific defect types, a problem compounded by the scarcity of examples for individual classes. To tackle this, we introduce Generative Defect Isolation (GDI), utilizing the LaMa inpainting model with Fast Fourier Convolutions to remove selected defects and generate realistic, single-defect training samples. Extensive experiments on Vision Transformer (ViT-S, ViT-L) and EfficientNetV2-L architectures demonstrate that GDI significantly outperforms baselines. The performance gains are most pronounced in low-data scenarios; class-wise analysis shows substantial improvements, boosting the F1-Score for rare defect classes by up to 63.6%. Furthermore, GDI effectively resolves learning ambiguity from co-occurring defects, yielding a 26% reduction in such co-occurring classification errors. Our work establishes GDI as an effective method for maximizing the value of existing segmentation datasets and sets a new performance benchmark for multi-label classification in this domain.

Hyperspectral imaging (HSI) offers nondestructive assessment of fish freshness by detecting biochemical alterations across spectral bands. However, conventional deep learning approaches do not fully address the particular characteristics of HSI data, such as spectral dominance over spatial textures, ordinal label structure, and a small number of training samples. We propose SGNet (Spectral-Grouped Network), a lightweight architecture that separates spectral and spatial feature extraction using grouped convolutions and a depthwise spatial pathway. A dual attention mechanism that couples channel-wise squeeze-and-excitation with spatial gating adaptively highlights informative features. SGNet achieves 97.8% classification accuracy and 0.64 days mean absolute error (MAE) with just 4.75M parameters when tested on our newly developed 16-day refrigerator-stored salmon fillet dataset. Ablation studies validate the contribution of each component, while comparisons demonstrate a five- to eighteen-fold parameter reduction relative to ResNet-50 and Vision Transformers. Our findings indicate that domain-aware design supports precise, real-time freshness prediction for industrial implementation.

Identifying dengue virus-infected mosquitoes from control mosquitoes is a major challenge in analyzing mosquito locomotion behavior due to the small size and complexity of the video background. Conventional AI methods are often unable to extract accurate features from video frames and produce erroneous features. In this study, a three-step framework is introduced: first, mosquitoes are identified and the background is removed using the YOLO 11M model, then visual features are extracted using the Vision Transformer (ViT), and finally the videos are classified with a convolutional GRU (ConvGRU) classifier. A comparative analysis of different models, including Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and their convolutional versions showed that the ConvGRU model achieved the best performance; it achieved 88.88% accuracy, 84.45% precision, 82.82% recall, and 82.81% F1 score. These results demonstrate that combining convolutional models with sequence-based networks, especially in the ConvGRU model, allows the simultaneous extraction of precise spatial features and long-term temporal dependencies from mosquito movements. Finally, the proposed framework provides a reliable solution for analyzing mosquito behavior in complex environments.

Token-pruning policies are usually designed for a single recognition pipeline, but pretrained Vision Transformers are reused across tasks with different spatial demands. We ask which parts of a pruning policy transfer across image classification, semantic segmentation, and object detection. For each pipeline, controlled probes freeze the no-pruning checkpoint and apply a series of parameter-free reduction criteria at one eligible layer at a time without retraining. The probes reveal three differences: segmentation and detection rank the criteria differently, classification is especially sensitive to attention-based pruning in the earliest layers, and the dense tasks prefer opposite recovery endpoints. These findings motivate Task-Adaptive Pruning (TAP). Existing register tokens serve as task-agnostic storage for feature artifacts. TAP instead introduces one task register per task and activates only the current one. Its evolving state ranks tokens, distributes an exact removal budget over depth, and sets the recovery scale for dense features. At a final keep rate of $ρ=0.5$, our jointly adapted model, TAP-J, reaches $47.0$ mIoU at $1.30\times$ encoder throughput on ADE20K and $53.7$ box AP at $1.32\times$ encoder throughput on COCO while remaining competitive on ImageNet-1K.

Vision Transformers demonstrate remarkable global modeling capacity but often underperform in data-scarce regimes. Distilling convolutional inductive biases from a CNN teacher provides an effective remedy while leaving the deployed model unchanged. However, general-purpose feature distillation transfers little in this setting. In CNN-to-CNN distillation, pooling, flattening, and logit-space projections remove the spatial grid that encodes locality and translation equivariance. Unlike a convolutional student, a ViT cannot readily reconstruct this structure on its own. In this paper, we propose iBKD, a distillation framework that preserves the spatial grid throughout the entire transfer process. Its core module, the Inductive Bias Attention Module, aggregates features from all student layers onto the teacher's grid using learned weights. It then enhances structural cues through channel and deformable spatial attention and injects them via convolutional cross-attention operating directly between spatial grids rather than token sets. The module is used only during training, leaving the deployed model as an unmodified ViT with no inference overhead. Across seven Transformer backbones and six data-scarce benchmarks, iBKD consistently outperforms both locality-guidance methods and general knowledge distillation baselines, with its advantage increasing as the amount of training data decreases.

We present DINO-A, an adaptation of self-distillation from vision to general audio representation learning. While DINO has become a canonical method in self-supervised vision and prior audio work has explored latent prediction (BYOL-A) and masked modeling (Audio-MAE, BEATs), no prior work has brought canonical DINO to general audio classification in the way BYOL-A brought BYOL. DINO-A retains DINO's multi-crop, EMA teacher, and high-dimensional projection, replacing only the input modality and augmentations with log-mel spectrograms and the BYOL-A v2 augmentation block. We pretrain three backbones, two Vision Transformers with 8x8 and 16x16 patches and a convolutional encoder, on FSD50K and evaluate them with linear probing on ESC-50, Speech Commands v2, UrbanSound8K, and GTZAN. Three findings characterize the resulting representations. Patch resolution within the Vision Transformer family has consistent effect on representation quality, with smaller patches winning across all four tasks. The choice between Vision Transformer and convolutional backbone interacts with task type: convolutional networks lead on speech while Vision Transformers lead on environmental sounds and music. Under identical pretraining and evaluation conditions, DINO-A and BYOL-A v2 differ by 11.96 percentage points on average, and we trace this difference to two mechanisms: the interaction between DINO's high-dimensional projection space and FSD50K's limited scale, and the additional cost of multi-crop augmentation, which DINO uses but BYOL-A v2 does not. The high-dimensional projection space, central to DINO's success in vision, becomes a liability at FSD50K scale.

Vision Transformers (ViTs) are widely believed to require more labeled data than CNNs for industrial dense prediction. Through controlled experiments on four industrial datasets, we show that the data-efficiency gap stems from pretraining incoherence, which refers to the statistical mismatch between ImageNet-pretrained ViT backbones and COCO-pretrained CNN necks, rather than from inherent self-attention deficits. We characterize the cross-architecture feature gap and propose a lightweight AlignBlock family for pyramid-level feature recalibration. Our core finding empirically identifies a data-efficiency frontier: for domain-proximal scenes with >= 200 samples, Swin-Graft surpasses YOLOv11x (terminal 703-shot: 0.973 vs 0.956 mAP@50); for domain-distant scenes, CNNs retain advantage (hook 141-shot: 0.900 vs 0.600 mAP@50). Grafted neck weights yield up to 2.5x the mAP of a randomly initialized neck.

AI training's rising resource intensity is straining electricity supplies and carbon budgets, motivating systematic study of memory-efficient training on constrained hardware. We benchmark five gradient optimizers (SGD, Adam, Adagrad, Adadelta, and Conjugate Gradient Descent) under three memory strategies (standard training, gradient checkpointing, and gradient accumulation) across four transformer architectures (ViT, ModernBERT, Llama 3.1 1B, and NanoVLM), measuring training loss, GPU utilization, training time, and memory usage. Gradient accumulation emerges as the most reliable strategy, cutting training loss by roughly an order of magnitude on the vision-language model and about four-fold on the language model without additional GPU memory. Contrary to common practice, Adam is not universally superior: Adadelta and SGD outperform it on the encoder and autoregressive architectures. Gradient checkpointing's effect is strongly architecture-dependent, improving vision transformer loss while severely degrading the encoder model, and it increases training time by up to 60% on memory-bound models. GPU utilization is governed primarily by architecture, ranging from 8-15% for the memory-bound language model to 96-99% for compute-bound vision models. These findings provide practical guidelines for optimizer and gradient-strategy selection in resource-efficient model training and deployment.

Distributed deployment of large vision foundation models often partitions a ViT backbone and exchanges intermediate token features between computing nodes, making efficient feature compression critical under bandwidth and computation constraints. Existing ViT feature codecs typically flatten heterogeneous global and patch tokens into an L x C pseudo image, causing entropy models to mainly capture sequence-axis dependencies while overlooking the native two-dimensional patch-grid structure. In this paper, we show that ViT patch tokens retain strong local spatial correlations on the original grid. To exploit this structural prior, we propose the Visual Token Codec (VTC), a dual-path learned codec that separates global and patch tokens into dedicated coding paths. Global tokens are compressed with a lightweight factorized prior, whereas patch tokens are encoded on the patch-token grid using a spatial-channel context entropy model. To support intermediate-layer compression and practical rate adaptation, VTC further incorporates feature-matching supervision after subsequent ViT blocks and variable-rate modules within a single codec. Experiments on DINOv2 and SAM3 show that VTC consistently outperforms representative ViT feature coding baselines on classification, segmentation, and detection tasks. At 90% of uncompressed-feature performance, VTC reduces bitrate by 15.7x-37.4x across these tasks. We further provide intermediate-layer rate-utility analyses for practical transmission- and storage-oriented deployment scenarios.

Parameter-Efficient Fine-Tuning (PEFT) has become the de facto standard for adapting Vision Transformers (ViTs) to downstream tasks. While parameter count has been the dominant efficiency metric in PEFT, it does not imply \textit{compute efficiency}: parameter-sparse methods can still incur full-model training cost per step, and typically need long schedules to reach peak accuracy. We introduce Circuit Fine-Tuning (CFT), a compute-efficient framework that uses circuit discovery---conventionally used to explain trained models---to select modules for fine-tuning before training. Whereas attribution is conventionally formulated against a trained task head, we formulate it against a near-zero-initialized probe head, which isolates the response of the backbone to the target distribution rather than the preferences of a particular classifier. CFT then fine-tunes only the recovered subgraph. CFT needs no learning-rate warmup and reaches peak accuracy in ${\sim}20$ epochs on average---versus $44$--$96$ for strong PEFT baselines---yielding $2.3$--$6.6\times$ fewer training FLOPs and up to $16\times$ less wall-clock time, while adding zero parameters and no inference operations. Experiments across a standard visual transfer benchmark (VTAB-1k), hierarchical backbones (Swin), domain-shifted medical imaging (CBIS-DDSM), and a vision-language model (Gemma-3 on CUB-200) demonstrate the effectiveness of CFT. Code is available at https://github.com/UriKialy/CFT

Vision transformers face significant computational overheads in high-resolution dense prediction due to the quadratic complexity of self-attention. Linear attention offers efficiency but sacrifices local context modeling. We propose \textbf{HSMLA (Hierarchical Softmax Multi-scale Linear Attention)}, which combines ReLU-based linear attention for global context, selective softmax refinement for critical local features, and multi-scale token representations via depthwise convolutions. HSMLA achieves superior accuracy-efficiency trade-offs: up to $4.2\times$ inference-time speedup across dense prediction tasks, $87.3$ Dice with $3.2\times$ speedup on CT organ segmentation, and $94.2$ AUC with $4.1\times$ speedup on pathology WSI.

Vision transformers (ViTs) have become the de facto standard for image encoding across many perception tasks. Despite their empirical success, it remains mechanistically unclear how they encode low-level features, given their lack of inductive biases: ViTs process information globally rather than relying on local structure. Biological visual systems, in contrast, build low-level features, such as orientation selectivity in the primary visual cortex, by combining information from small, localized regions of the visual field. These features are general-purpose representations, shared and required across multiple specialized neural pathways, unlike higher-level, task-specific semantic features. This raises the question if such biologically-grounded features arise in ViTs. In this work, we systematically study how orientation selectivity emerges in ViTs by introducing a suite of neuroscience-inspired metrics: representational similarity score (RSS), orientation recruitment score (ORS), and orientation tuning bandwidth to quantify how orientation is encoded in representational geometry and as a function of model depth. Through extensive analysis, we find that: (1) the training paradigm is the strongest determinant of orientation selectivity, with models sharing an objective, peaking at comparable relative depths regardless of scale (2) many units are orientation-selective early in training, with early-to-middle layers recruiting more such units over time, while deeper layers lose selectivity and broaden their tuning toward semantic encoding and (3) our metrics offer a mechanistic heuristic for how many layers to unfreeze for best downstream generalization. Our framework presents a way to track biologically-grounded features during ViT training, probes how desired properties are encoded in transformer representations, and builds a systematic understanding of how ViTs generalize across tasks.

Gradient-weighted Class Activation Mapping (Grad-CAM) is widely used to visualize model decisions, but it was originally formulated for convolutional neural networks, where spatial feature maps and channel dimensions have clear architectural meanings. Vision Transformers (ViTs) do not provide the same structure, instead representing images through tokens, attention, residual streams, and multimodal interactions. This paper presents a systematic taxonomy and literature audit of how Grad-CAM and related methods are adapted, justified, and reported for ViT-based architectures. From an initial search of more than 550 papers, we identify 175 papers that apply Grad-CAM or Grad-CAM-adjacent methods to ViTs. We find that most papers do not provide a full mathematical or implementation-level account of how Grad-CAM is adapted to transformer representations. To characterize this gap, we introduce a descriptive taxonomy of ViT Grad-CAM adaptations that makes explicit the feature locations, gradient targets, spatial reconstruction steps, and aggregation choices that are often left implicit. This taxonomy is not intended to prescribe a single correct adaptation, but to clarify the range of methodological choices being made. The study shows that Grad-CAM on ViTs is often treated as a trivial extension of CNN-based Grad-CAM, despite requiring nontrivial choices that affect rigor, reproducibility, and interpretation.

Vision Transformers (ViTs) achieve strong image-recognition performance, but their parameter count grows linearly with depth when each block is independently parameterized. Single-block recurrent ViTs (bViT) remove this growth by repeatedly applying one shared block. Rather than proposing a new architecture, we fix a bViT and provide a controlled empirical characterization of three training and inference regimes under a common CIFAR-100 protocol, asking: (i)~when does recurrence beat independently parameterized depth---at matched FLOPs or at matched parameter memory? (ii)~when a residual recurrent block is trained through an ODE solver, does solver order act as numerical refinement or as an architectural bias? and (iii)~what does robustness beyond the training horizon cost in nominal accuracy? We find that standard ViTs remain preferable when FLOPs are the primary constraint, whereas recurrent ViTs offer a better accuracy--parameter trade-off under memory constraints. Consistent with the standard view of residual networks as Euler discretizations of ODEs, the continuous-time analogue of a residual recurrent block is the state-subtracted vector field $\dot{z}=F_θ(z)-z$; although known in principle, this distinction is easy to violate when the block is wrapped as a black-box vector field, and we qualify the cost at few accuracy points. Because the vector field is learned jointly with the solver, higher-order solvers act as a solver-induced architectural bias rather than a numerical-accuracy improvement, and their gains are not uniform. Finally, stage-wise deep supervision traces an accuracy--robustness frontier: it does not improve nominal accuracy, but degrades gracefully far beyond the training horizon, where naive recurrence collapses to near-random performance.

To adopt the Vision Transformers (ViTs) in resource-constrained environment, token pruning is widely used to reduce computational cost without impacting accuracy. However, adversaries have developed targeted attacks against said token pruning techniques to undermine such attempts to make ViTs efficient. In this paper, we propose MOAT, a model-agnostic pre-processing defense pipeline that applies a combination of input transformations to protect efficient ViT implementations against adversarial efficiency attacks. MOAT operates directly on the input without requiring modifications to the model architecture or token pruning mechanism. Experimental results demonstrate that, across all evaluated ViT models, MOAT limits GFLOPs degradation under adversarial attacks to within 3.4% of the original unattacked model.

Vision Transformers (ViTs) increasingly rely on input-adaptive inference, such as token pruning and early halting, to meet energy and latency budgets. This survey examines a recent class of adversarial efficiency degradation attacks that target these mechanisms to increase computation without necessarily degrading accuracy. We unify and compare two representative attacks, SlowFormer (a universal adversarial patch) and DeSparsify (per-image perturbations), across three popular token-pruning frameworks: A-ViT, ATS, and AdaViT. We standardize reporting using GFLOPs, accuracy loss, and an Attack Success (AS) metric that measures how much of the model's compute savings the attack takes away. Understanding these attacks is crucial for designing countermeasures that not only mitigate risk but also remain lightweight, since deployment often occurs in low-power settings such as mobile or embedded devices. To organize our analysis, we focus on three questions: how input-adaptive optimizations (e.g., token pruning and early halting) create attack surfaces for efficiency degradation; how such attacks operate in practice and which optimizations are most vulnerable; and which defenses exist today and whether they meaningfully restore efficiency under attack.

Transfer-based adversarial attacks often transfer poorly across heterogeneous architectures because CNNs favor local textures while Vision Transformers (ViTs) rely on global shapes. We propose Season, a spectrum-aware orthogonal gradient refinement framework for L-infinity transfer attacks against black-box target models on ImageNet, using a white-box surrogate. Season decomposes each update into a low-frequency branch capturing structural cues and a high-frequency branch capturing textures. A low-saliency guidance scheme reallocates high-frequency energy to background regions, preserving foreground structures that ViTs depend on. An orthogonal projection then forces the textural update to lie in the orthogonal complement of the structural direction, mitigating feature interference. As a training-free plug-and-play wrapper, Season enhances eight gradient-stabilization and input-enhancement attacks without modifying their cores. Across eight CNN, ViT, and MLP targets, Season improves transfer success rate by 6.6 percentage points on average and up to 16.0 points over strong baselines under a unified protocol.

Pose-agnostic Anomaly Detection (PAD) remains challenging as anomalies can appear under arbitrary viewpoints, requiring methods to handle significant pose variations. Existing approaches rely on complex 3D reconstruction, which are computationally expensive and require extensive multi-view data. We propose PADFormer, a novel image-space approach that leverages Vision Transformer (ViT) to directly reconstruct anomaly-free versions of query images while preserving pose information. Our key insight is to adapt cross-view masked reconstruction for anomaly detection through training exclusively on normal data, combined with dynamic patch selection and spatial alignment mechanisms that enable effective learning from sparse reference views under significant pose variations. During inference, we perform multiple forward passes with different masking patterns to generate an ensemble of anomaly-free reconstructions, ensuring comprehensive coverage of the query image. Anomalies are detected by comparing these reconstructions with the query image. PADFormer achieves state-of-the-art results on the PAD benchmark while maintaining comparable performance on classic few-shot anomaly detection (FSAD) tasks, demonstrating superior efficiency and generalization without requiring 3D reconstruction.

Existing scaling strategies for Multimodal Large Language Models (MLLMs) typically expand either model parameters or sequential inference computation, incurring substantial memory or latency overhead. More importantly, most existing methods fail to alter the rigid, fixed computation allocation between the Vision Transformer and the Large Language Model components, limiting task-specific optimization. To address this, we introduce the Parallel Vision-Language (ParVL) scaling framework for MLLMs, which scales parallel computation by reusing the existing ViT and LLM backbone parameters across multiple vision and language branches. This framework raises a central question: given a fixed backbone parameter budget, how should additional shared-backbone computation be allocated between the vision and language modalities? We instantiate each parallel computational stream with branch-specific prefix parameters over a shared backbone, and train the entire model end-to-end via full-parameter supervised fine-tuning on roughly 13B tokens. We systematically study the computation-allocation trade-off between the ViT encoder and LLM decoder. ParVL improves overall multimodal performance over same-recipe single-branch baselines, and the best evaluated vision--language allocation varies across tasks. Code is available at https://github.com/YangYangGirl/ParVL.

Most evidence on the effectiveness of explainable artificial intelligence (XAI) attribution methods has been established on convolutional neural networks (CNNs), with limited investigation into whether these conclusions generalize to the diverse Vision Transformer (ViT) architectures that now dominate computer vision. This paper presents a controlled benchmark that evaluates attribution quality across five dimensions: faithfulness, localization, robustness, complexity, and computational cost. A standardized framework assesses 13 attribution methods from four algorithmic families on eight representative backbones spanning CNNs, isotropic ViTs, hierarchical transformers, hybrid architectures, and linear-attention transformers. The results show that attribution performance is strongly architecture-dependent and that rankings established on CNNs do not reliably transfer to transformer-based models. CAM-based methods achieve the highest scores under the conventional bounding-box localization metric on CNNs and most ViTs but perform poorly on linear-attention architectures. Pixel-level dense-mask evaluation further reveals that these gains largely reflect metric saturation rather than accurate localization. CAM-based methods also exhibit limited robustness on global-attention transformers, whereas attention rollout provides consistently stable explanations with poor localization. Furthermore, faithfulness correlation offers limited discrimination between attribution methods, highlighting the limitations of single-metric evaluation. These findings challenge prevailing conclusions on attribution performance and demonstrate the need for architecture-aware, multi-dimensional evaluation. The open-source code for the evaluation framework and benchmark results is available at https://github.com/Nishan-Charlie/VIT_XAI_Bench.

With the widespread adoption of Vision Transformers in modern AI, the need to analyze their inherent representational behavior has become increasingly important. While most existing studies emphasize token geometries and training dynamics, the evolution of representational covariance structures and class-level geometric organization remains comparatively underexplored. In this work, we investigate semantic geometry and class separability as representations evolve across the layers of ViT-Small/16 through TGO-III: Semantic Geometry Observatory. It is a framework designed to analyze the emergence of semantic organization, feature evolution, and class-wise representation geometry throughout training. The framework employs multiple complementary observatories, including Linear Probe Accuracy, Fisher Ratio, Class Centroid Distances, Local Intrinsic Dimension, and Local PCA Rank, to quantify the progressive evolution of discriminative representations. Our analysis reveals that class representations become progressively more linearly separable, Fisher discriminability increases, class centroids move farther apart, and local representation manifolds exhibit structured class-dependent geometric complexity. These observations provide empirical evidence supporting the Semantic Expansion Hypothesis, suggesting that the manifold expansion observed in previous observatories is accompanied by the progressive organization of representations into increasingly discriminative semantic structures. Collectively, TGO-III extends the Transformer Geometry Observatory framework by establishing a direct connection between manifold geometry, covariance evolution, and semantic organization during Transformer training.

The wide adoption of Vision Transformers (ViTs) in safety-critical applications raises reliability concerns related to hardware faults. Algorithm-Based Fault Tolerance (ABFT) methods have emerged as lightweight and symmetric protection mechanisms for DNNs. However, they are particularly challenging for ViTs due to their significant computational requirements. This work comprehensively evaluates the reliability of ViTs, emphasizing the need for symmetric protection in their layers. Furthermore, we present CheckOne, a novel, cost-effective method for fault detection and mitigation in ViTs that significantly reduces the computational cost compared to conventional ABFT. Through extensive experiments with multiple ViTs, CheckOne mitigates critical faults by up to $26\times$ and achieves an average 3.8x higher performance than ABFT in ViTs.

Low-Rank Adaptation (LoRA) enables parameter-efficient fine-tuning, but standard LoRA produces a single deterministic model and does not directly support predictive uncertainty estimation. We introduce EulerLoRA, a stochastic extension of LoRA that generates multiple predictive trajectories by sampling structured variations along the rank-one components of shared low-rank adapters, while preserving the deterministic LoRA transformation in expectation. We evaluate EulerLoRA with vision transformers on CIFAR-10, CIFAR-100, and HAM10000, together with out-of-distribution detection on SVHN. Across these benchmarks, EulerLoRA achieves comparable or improved performance relative to strong LoRA-Ensemble baselines. Using two rank-20 adapters, EulerLoRA requires approximately 3 million trainable adapter parameters, compared with about 10 million for a rank-8, 16-adapter LoRA-Ensemble, corresponding to roughly 69% fewer trainable parameters. These results show that useful predictive diversity can be obtained from a small number of shared adapters.

Vision foundation models, such as DINOv2, learn highly expressive representations but rely on massive, opaque architectures that demand substantial computational power and memory. To provide an interpretable-guided and efficient solution to this issue, we first propose a spectral analysis and new visualization technique for individual attention heads based on the Laplacian eigenvectors of their attention maps. Building upon recent observations regarding the block structure of Vision Transformers, we perform semantic clustering of attention heads and identify functional redundancies. Leveraging these insights, we introduce SAPER (Soft Attention PrunER), an end-to-end differentiable pruning framework based on the LapSum Soft Top-K approach. Extensive experiments on ImageNet-1K demonstrate that SAPER achieves a highly favorable accuracy-efficiency trade-off, outperforming the competitive RAPTOR baseline in FLOPs reduction while preserving strong classification performance.

Visual imitation learning enables robots to acquire visuomotor skills directly from images, yet RGB observations lack explicit geometric cues, making learned policies brittle to camera perturbations. To address this, we propose \textbf{Ray-conditioned Vision Transformer Encoder (RayViT)}, a lightweight architecture that injects camera geometry into pretrained ViT backbones. RayViT represents camera geometry as a Plücker ray map, patchifies it into ray features, and uses gated cross-attention to produce a ray-conditioned class token. These ray features are added as dense positional embeddings, while the ray class token replaces the original ViT class token to provide a geometry-aware summary representation. We combine this approach with an auxiliary cosine similarity loss to consistently improve the performance and robustness for geometry-aware tokens. Experiments on sim- and real-robot tasks demonstrate that RayViT improves robustness by approximately 13 percentage points under camera perturbations in multi-task RoboCasa benchmark and by 1.78 average completed stages in real-world multi-task success rate compared to baselines.

Post-training quantization (PTQ) has emerged as an effective solution for deploying Vision Transformers (ViTs) on resource-constrained devices. However, existing PTQ methods typically employ uniform bit-widths across transformer components, overlooking their heterogeneous sensitivity to quantization and leading to inefficient precision allocation. In this paper, we propose {MixFrag, a fragility-guided mixed-precision PTQ framework for Vision Transformers. MixFrag first estimates component-level quantization fragility by measuring the Kullback--Leibler (KL) divergence between full-precision and isolated quantized output distributions using a small calibration set. It then formulates bit allocation as a Multiple-Choice Knapsack Problem (MCKP), enabling adaptive layer-wise precision assignment under a target bit budget. Extensive experiments on ImageNet-1K across multiple Vision Transformer architectures demonstrate that MixFrag achieves competitive classification performance under practical mixed-precision settings. Furthermore, evaluations on COCO object detection and instance segmentation show that MixFrag achieves state-of-the-art performance among existing mixed-precision PTQ methods, improving the previous best method by up to 9.6 AP under the challenging MP3/MP3 setting. Additional analyses validate the proposed fragility metric and demonstrate its strong correlation with the learned bit allocation. These results establish MixFrag as an effective framework for mixed-precision post-training quantization of Vision Transformers.

Operational Earth observation increasingly calls for answering queries such as find the image pairs where a new building appeared.'' This means searching an archive of before-and-after (bi-temporal) satellite image pairs and ranking each pair by how well it matches a natural-language description of the change. The component that performs this match, the fusion module that combines the before'' and ``after'' views, must be run at query time across many candidate pairs, so its speed largely sets the cost of every search. We present a controlled comparison of how to build that module. Using one fixed image encoder (a frozen CLIP model) and one training recipe for all variants, we evaluate eight designs drawn from three families: attention, state-space models (Mamba), and learned compression (our Temporal Bottleneck Fusion, TBF). Each design is tested on two benchmarks (LEVIR-CC and Dubai-CC) with ten random seeds, so the reported differences are statistically grounded. We outline three findings: first, a training-free two-stage search (a cheap difference model that shortlists candidates, followed by attention fusion that re-ranks them) matches or exceeds full-fusion recall on LEVIR-CC while cutting query cost $10$-$15\times$, with comparable R@1/R@5 on Dubai-CC; second, the linear-time scan of Mamba, attractive on paper, gives no speed benefit at the patch counts typical of vision transformers ($L{=}196$): the scan is limited by memory bandwidth, whereas attention maps cleanly onto parallel hardware; and third, compressing the fused representation (TBF) reduces parameters by $2.3\times$ and latency by $1.6\times$ for a change-only BLEU-1 cost of $0.007$, although more aggressive compression quietly discards change-relevant detail that aggregate metrics fail to reveal.

Wind turbine blade defect detection remains highly challenging in real-world inspection scenarios due to limited on-site data and the subtle visual characteristics of defects. In practice, blade defects are often small-scale, low-contrast, and difficult to distinguish from complex backgrounds, which significantly limits the robustness of existing detectors. To address these challenges, we propose BladeYOLO, a defect detection framework for wind turbine blades. Specifically, we integrate a Vision Transformer (ViT) backbone initialized with DINOv3 self-supervised pre-trained weights into YOLOv12-L, enabling the transfer of large-scale generic visual priors to blade defect detection and improving feature representation under limited training annotations. To enhance the perception of subtle defects, we further develop a Mamba-guided Weak-Defect Enhancement module, which consists of a Detail-Enhanced Multi-scale Branch for preserving high-frequency structural cues and a Cross-Mamba module for progressively propagating high-level semantic guidance to shallow features. In addition, we introduce a lightweight Style-Injector module that captures environment-related style information via Fourier decomposition and injects it into selected ViT self-attention layers, thereby improving robustness against environment-induced appearance variations. Extensive experiments demonstrate that BladeYOLO achieves superior performance on the WTBlade-Defect dataset, with additional annotation-budget experiments showing its favorable performance under reduced training annotations. Evaluation on the public Wind Surface Defect dataset further provides supportive evidence for the cross-dataset robustness of BladeYOLO. In particular, on this public dataset, BladeYOLO outperforms the best competing method by 3.5% in mAP$_{50}$ and 2.5% in mAP$_{50-95}$.

Vision foundation models have enabled strong training-free anomaly detection (AD). However, most existing approaches rely primarily on independent local patch features, leaving the global contextual information encoded by Vision Transformers (ViTs) underexploited. In this work, we identify the dual characteristics of the ViT [CLS] token: its embedding provides anomaly-invariant global semantic representation, while its attention maps implicitly highlight spatially abnormal regions. Building on this observation, we propose a fully automated AD framework leveraging global context to remove manual tunings. Our framework introduces (1) an automatic augmentation selection strategy driven by [CLS]-level semantic consistency, and (2) an attention-guided feature reweighting mechanism that dynamically adjusts patch contributions according to [CLS] attention saliency. By integrating these components over multi-level features, our method achieves stable anomaly scoring and precise localization without training or parameter tuning. Under the one-shot setting, it achieves Image-AUC scores of 97.7%, 93.2%, and 84.5% on MVTec-AD, VisA, and Real-IAD. Using a single fixed configuration across categories, backbones, and datasets, the method establishes a new state-of-the-art for plug-and-play, training-free anomaly detection while maintaining strong robustness and practical scalability.

Pretrained dense visual features from Vision Transformers (ViTs) are powerful yet have been underutilized in robot learning. Modern robot policies either compress each observation into a single global token, or rely on visual backbones trained from scratch, sacrificing both fine-grained spatial detail and the benefits of large-scale visual pre-training. While there exist policies that do operate on dense patch features like large vision-language-action models (VLAs), they tend to be heavy and slow, inheriting the full cost of a billion-parameter vision-language model (VLM) backbone. We close this gap with Patch Policy, a minimal architectural extension that enables transformer-based policies to consume dense pre-trained patch tokens directly without the computational overhead of a full VLM. At its core is a block-causal attention mask that preserves the temporal causality of standard policies while letting the model attend over many patch tokens per observation, alongside other state information. Patch Policy is lightweight, fast, and highly effective. Across four simulated and three real-world environment suites, our method achieves a 40% relative improvement over policies using state-of-the-art global-pooled representations. Furthermore, it surpasses fine-tuned OpenVLA-OFT by 18% while using roughly 0.7% of the parameters. We believe Patch Policy provides a pipeline for the robotics community to readily leverage continuing progress in visual representation learning, without sacrificing the training efficiency or inference speed required for high-frequency, reactive control. Videos can be viewed at https://patch-policy.github.io

Remote sensing has become an increasingly valuable tool for agricultural monitoring, particularly through the use of publicly available satellite imagery. However, effectively integrating domain knowledge into machine learning methods remains challenging. This study presents a real-world example of early sugar beet harvest yield forecasting from purely optical Sentinel-2 imagery, demonstrating how a tight integration of domain knowledge and machine learning can lead to synergistic gains. We empirically find that using very small vision transformer patch sizes and all available Sentinel-2 spectral bands improves our model despite being uncommon design choices in the domain. As a practical contribution, we were able to identify a large fraction of low-yield fields in a different year early on in the growth cycle through a modified training setup and a ranking-based detection of underperforming fields.

Vision Transformers process spatially redundant tokens efficiently only when coarse token summaries preserve the evidence required by exponential attention aggregation. We identify a boundary-minority underestimation failure in which a spatially small, high-response region contributes dominant Gibbs mass while remaining nearly invisible to a block mean. We formalize the failure through the discrepancy between normalized log-mean-exp free energy and mean summarization, prove that minority Gibbs mass can remain non-vanishing as its spatial support and mean contribution vanish, and characterize the limitations of finite-order moment corrections. Building on the resulting analysis, we introduce Boundary-Minority Free-Energy Adaptive Screening (BMFA), which constructs a hierarchical piecewise-constant approximation and recursively refines blocks according to a computable lower-bound increment of local free energy. Controlled synthetic tests, COCO and LVIS diagnostic probes, closed-loop DeiT-Tiny evaluations, and ImageNet-1K experiments establish a consistent evidence chain. BMFA reduces the mean synthetic underestimate from 2.582 to 0.261 at a 5.794% leaf ratio, lowers the COCO image-edge mean gap from 2.254 to 0.526, and preserves 71.520% ImageNet Top-1 accuracy at a 55.861% leaf ratio. The current prototype evaluates selection quality after full QK computation; the reported leaf ratio therefore characterizes representation granularity rather than verified sparse-kernel speedup.

Silicon-photonic (SiPh) accelerators have emerged as a promising platform for Vision Transformer (ViT) inference by performing matrix multiplications on microring-resonator (MRR) banks with high throughput and energy efficiency. Extending these platforms to support on-chip fine-tuning remains challenging because backpropagation requires large activation storage, frequent weight write-back to MRRs, and tolerance to device-level noise. We present Opto-ViT-v2, the first framework for parameter-efficient fine-tuning (PEFT) on a near-sensor SiPh ViT accelerator. Our tensorized low-rank decomposition separates pretrained optical weights from a small set of trainable electronic factors (as few as 8K parameters for ViT-Base), greatly reducing activation storage and weight updates while enabling practical on-chip training. We further introduce a gradient-accumulated sparse classifier that freezes low-importance weights through one-shot top-k gradient masking, reducing classifier training cost by about 40 percent. We also develop the first system-level noise model for photonic on-chip training, capturing the effects of MRR crosstalk, thermal drift, and laser amplitude noise during both forward and backward propagation. Calibrated using measurements from more than 200 fabricated MRR devices, the model shows that low-rank factor updates are more robust than full fine-tuning and conventional layer-wise low-rank adaptation under identical noise conditions. Experiments on VTAB-1K (19 tasks) and FGVC few-shot benchmarks demonstrate that Opto-ViT-v2 recovers within 0.3 to 0.8 percent of clean software accuracy under measured photonic noise while achieving more than 100 KFPS/W, enabling practical on-chip domain adaptation for photonic edge vision systems.

Inverse rendering is traditionally solved via differentiable renderers and gradient descent, which requires substantial problem-specific engineering and is prone to getting stuck in local minima due to ambiguities. Derivative-free approaches alleviate engineering requirements, but often heavily depend on a good problem initialization. In this work, we propose Feature-Informed Diffusion Evolution (FIDE), a fully black-box framework that requires no gradients or specific initialization: the renderer is treated as an opaque function whose only requirement is to produce images. Our key insight is feature guiding: rather than reducing each candidate rendering to a scalar loss value, we use a Vision Transformer (ViT) to extract dense visual features from it. We subsequently use these features to train a diffusion-based candidate proposal model, allowing the network to use visual cues to predict parameters that would match the target image. The candidate solutions proposed by this diffusion model are then refined in a closed loop with a CMA evolution strategy, continuously narrowing the proposal region as optimization progresses. We validate across diverse inverse problems from path tracing, vector splines, Voronoi shaders, and robotics, and demonstrate that feature-guiding substantially improves convergence over scalar-loss baselines and reliably escapes local minima where gradient-based methods stall.

Recent visual place recognition (VPR) methods based on vision transformers, particularly foundation models, have achieved remarkable recognition performance. However, these models process all visual tokens throughout the entire network, resulting in substantial computational overhead, which hinders their deployment in real-time and resource-constrained scenarios. A natural question thus arises: are all visual tokens necessary for VPR? To answer this question, we present the first systematic benchmark of token reduction for efficient visual place recognition. Our benchmark comprehensively evaluates representative token pruning, token merging, and hybrid pruning-merging methods across multiple state-of-the-art VPR models and diverse benchmark datasets covering urban, suburban, and natural environments. We further investigate token reduction from multiple perspectives, including recognition performance under different reduction configurations, computational complexity, inference speed, qualitative visualization, and deployment efficiency on edge devices. Through extensive experiments and in-depth analysis, our benchmark reveals multiple important characteristics of token reduction in VPR and provides several practical insights into the trade-offs between accuracy and inference efficiency. For example, token reduction can reduce computational cost by up to 29% and improve throughput by up to 44%, while incurring less than 1% degradation in recognition accuracy. Overall, this work establishes a comprehensive foundation for future research on token-efficient VPR and efficient visual retrieval systems. Our codes and models will be available at https://github.com/Tong-Jin01/TokenReduction4VPR

Vision Transformer (ViT) has been widely used as a powerful framework for modeling global dependencies among image patches. However, its core component, self-attention assigns softmax-normalized relative weights to all patches, making it difficult to evaluate the relevance between patches independently. In visual recognition, images often contain many background or redundant patches, yet self-attention cannot explicitly reject such irrelevant patches, which may introduce unnecessary information into feature aggregation. To address this limitation, Screening has been proposed in the field of language modeling, where the relevance of each token is independently evaluated based on query-key similarity and low-relevance tokens are explicitly excluded through thresholding. In this work, we propose VisionScreen, a new vision model that extends Screening mechanism to visual recognition. VisionScreen treats image patches as tokens arranged on a two-dimensional grid and extends absolute relevance estimation based on query-key similarity to the two-dimensional spatial domain. This allows each patch to selectively aggregate only content-wise and spatially relevant patches without relying on competition among patches. Experiments on image classification benchmarks demonstrate that the proposed method outperforms conventional ViT. These results suggest that Screening can be effective for visual recognition, offering an alternative to relative feature aggregation based on softmax attention.

Do vision models see colors the way humans do? Existing evaluations of color representations usually compare them with geometric spaces such as CIELAB or with discrete color labels. These references capture perceptual distance or category membership, but not the graded way in which people organize colors. We evaluate color grounding against a fuzzy perceptual model with 86 graded categories fitted to human survey data. The framework can be applied to any image encoder and measures three complementary properties: category boundaries, category compactness, and graded alignment beyond what color geometry alone can explain. Across eleven Vision Transformer encoders, the category-level results are broadly similar, whereas graded alignment differs substantially. Masked Autoencoders achieve the strongest beyond-geometry alignment, with confidence intervals that do not overlap those of the other encoders. A layer-wise analysis further shows that masked reconstruction preserves this structure toward the output. On natural images, MAE represents surface color globally, while language-supervised models encode color more strongly in relation to the foreground object. These results show that human-like color grounding has several distinct aspects that should not be reduced to a single score.

Most multimodal learning methods improve how heterogeneous representations are aligned and fused, while post-fusion enhancement remains less explored. We propose Parallel Quantum Feature Augmentation (PQFA), a hybrid quantum-classical framework that applies multiple shallow variational quantum circuits to fused multimodal features. Text and image representations extracted by frozen RoBERTa and ViT encoders are processed through bidirectional cross-attention, attentive pooling, and adaptive gated fusion. The fused feature is then amplitude-encoded into parallel quantum circuits, whose measurement readouts are concatenated with the classical representation for prediction. We evaluate PQFA on MM-IMDb and N24News through controlled comparisons using the same encoders, fusion backbone, data splits, projection dimension, and augmentation output width. PQFA consistently outperforms both the fusion backbone without quantum augmentation and a width-matched MLP augmentation baseline, while using approximately 2.2K augmentation parameters compared with 24.0K for the MLP branch. Missing-modality experiments further show improved robustness when textual or visual inputs are incomplete, with particularly clear gains when the more informative textual modality is severely degraded. Controlled ablations and feature-space analyses indicate that the improvement cannot be reproduced by random feature mappings, increased classical width, or untrained quantum transformations. Quantum-state diagnostics additionally show stable predictive performance across the tested simulated noise levels and distinct branch-specific transformations of the encoded states. These results establish PQFA as an effective and parameter-efficient strategy for post-fusion augmentation in hybrid quantum-classical multimodal learning.

Scene Text Recognition (STR) remains challenging due to the diversity of text appearances, including curvature, rotation, and perspective distortion. Recent Transformer-based approaches perform well but usually rely on one-dimensional positional encodings that ignore the 2D spatial structure of text images. Axial 2D extensions of Rotary Position Embedding (RoPE) exist for vision Transformers, but they assume roughly square, isotropic image content and apply the rotation only within encoder self-attention. Scene text violates both assumptions: crops are markedly anisotropic, and STR models are encoder-decoder, so the decoder must relate its queries to the encoder's 2D layout through cross-attention. We introduce 2D-RoPE-STR, which adapts axial 2D-RoPE to this setting through (1) an anisotropic row/column dimension allocation matched to the aspect ratio of text, and (2) an extension of the rotary coupling into encoder-decoder cross-attention, letting autoregressive decoding steps attend to encoder tokens by their 2D layout, a setting not addressed by prior encoder-only formulations. Both changes are essentially parameter-free and require no architectural redesign beyond the positional-encoding module. We further introduce a diagnostic protocol (a controlled ablation pair isolating only the positional encoding, an image-level net-win disagreement analysis, and encoder attention visualization) that identifies where and why relative 2D position helps: curved, rotated, and perspective-distorted layouts where reading order departs from a straight horizontal line. On six standard benchmarks (IIIT5K, SVT, ICDAR 2013, ICDAR 2015, CUTE80, SVTP), gains concentrate on exactly these irregular layouts, with ablations isolating each design choice against 1D RoPE and 2D sinusoidal and learnable alternatives.

Vision Transformers (ViTs) have demonstrated remarkable performance in computer vision tasks. However, their self-attention mechanism often diffuses focus across background regions, relying on spurious correlations rather than object-relevant cues. Inspired by inhibitory mechanisms observed in biological vision systems, we propose the Inhibited Self-Attention (ISA), a novel self-attention that integrates inhibitory signals to enhance feature selectivity and suppress spurious responses. In contrast to conventional self-attention, which relies solely on positive attention values due to softmax normalization, our approach retains and utilizes negative attention scores to suppress irrelevant features and sharpen focus on objects of interest. Experiments across multiple datasets, including ImageNet-1k and COCO, and several robustness benchmarks demonstrate that ISA enhances object-centric selectivity, reduces shortcut reliance, and improves out-of-distribution generalization. Our analysis of relevance maps confirms that ViTs with ISA exhibit sharper, more localized focus on object-relevant regions while reducing distractions from non-relevant (background) features, enabling more reliable models. We release our code at https://github.com/prdvanderwal/inhibited-self-attention

Recently, Vision Transformer (ViT)-based models have exhibited remarkable performance in image super-resolution. However, the quadratic computational complexity of ViTs with respect to spatial resolution severely constrains their efficiency, leading to high latency and massive memory consumption. To alleviate this, various window-based attention mechanisms have been proposed; yet, they inherently compromise the long-range dependency modeling that is the primary advantage of ViTs. To overcome these limitations, we propose the Clustered Unit-level Similarity Transformer (CUST), a novel architecture that efficiently integrates global and local information. Specifically, CUST enables each patch to aggregate and attend to similar patches within a broadened regional scope outside its local window, thereby capturing extensive contextual understanding. Furthermore, it employs overlapping attention windows to capture local dependencies, while explicitly extracting high-frequency details by computing the residual difference between the original features and their downsampled-upsampled counterparts. Comprehensive experiments demonstrate that our proposed model achieves a practical balance between computational efficiency and restoration performance. It achieves a lower memory footprint and faster inference speed compared to recent global context or lightweight models under realistic constraints. Code is available at [https://github.com/jwgdmkj/CUST].

Physical AI systems, such as autonomous vehicles and intelligent machines, require transformer-based perception models that satisfy stringent edge latency and energy constraints. However, heterogeneous edge-GPU deployment remains limited by underutilized hardware engines and accelerator-incompatible operators, causing fragmented execution and lower throughput per watt. This paper presents Heterogeneous Frame Dispatch Scheduling (H-FraDS), a hardware-aware frame scheduling methodology for transformer inference on a recent NVIDIA edge GPU. H-FraDS routes frames across the GPU and dual deep learning accelerator (DLA) cores using fixed dispatch ratios to improve utilization under latency and power constraints. To enable scheduling, incompatible transformer components are adapted for DLA execution by reshaping tensors, approximating error function (ERF) with tanh, and replacing layer normalization with bounded tanh. The adapted model maintains a 92% F1 score, with only a 2% reduction from the original. Optical flow accelerator (OFA) is further used for inference-side optical-flow estimation. To the best of the authors' knowledge, prior work has not addressed these combined issues. Using Swin Transformer for autonomous-driving perception, H-FraDS Balanced Dispatch (1:2) achieves 125.93 FPS, a 2.36x speedup over standalone adapted-DLA execution, 4.0 FPS/W, and approximately 24 ms DLA latency, satisfying 30 FPS real-time operation; the GPU-DLA-OFA case achieves a 2.02x DLA throughput speedup.

Despite exposing rich intermediate representations, Vision Transformers (ViTs) are almost exclusively utilized as black-box feature extractors, where only the last layer is considered for downstream tasks. We challenge this convention by introducing the notion of recoverability: the capacity of intermediate representations to correct last-layer failures. By evaluating independent classification probes at every model depth across 16 datasets, we observe that intermediate probes correctly classify 18% to 76% of samples that the last-layer probe misclassifies. We show that these gains are not primarily driven by predictive diversity, but by a redundancy-correctness correspondence, where the internal hierarchy acts as a series of stable, redundant probes of a shared discriminative signal. While established horizontal ensemble strategies (i.e., across multiple models) can improve performance, they incur high computational cost and ignore this vertical signal within a single model. To bridge this gap, we propose VFusion, a principled vertical aggregation strategy employing a learnable mapping into a low-dimensional latent space that synthesizes features across the internal ViT hierarchy. VFusion substantially outperforms established aggregation baselines in both in-distribution and out-of-distribution settings, notably closing 45% of the accuracy gap between the best individual layer and a theoretical oracle performance. Our gains consistently generalize across model sizes and pre-training regimes, confirming that VFusion offers a robust and efficient alternative to horizontal ensemble methods. The code is available at https://github.com/francescodisalvo05/vit-vertical-fusion.