Papers

Composed Video Retrieval (CoVR) seeks the target video that results from applying a free-form textual modification to a reference video. We address the Reason-Aware CoVR (CoVR-R) challenge at the CVPR~2026 VidLLMs workshop, where retrieval is strictly zero-shot. We present R3-CoVR (Reason, Retrieve, Re-rank), a training-free pipeline built entirely from frozen foundation models. A multimodal large language model (Qwen3-VL-8B) reasons about the after-effects an edit implies – state transitions, action phases, scene, camera and tempo – and verbalises a concise post-edit description; a contrastive video–text encoder (SigLIP-2) embeds this description and the gallery for first-stage retrieval; finally a constraint-aware re-ranking stage uses the same multimodal model as a judge that scores each shortlisted candidate against the intended edited result. On the challenge test set, R3-CoVR attains 91.9% R@1 and 98.2% R@10. Two findings drive these results: (i)~matching the description length to the contrastive encoder's text window lifts \Rk{1} from $67.5$ to $72.7$; and (ii)~the constraint-aware re-ranker, which reorders only the shortlist, lifts \Rk{1} from $72.7$ to $91.9$ – the single largest gain. We analyse the re-ranker's behaviour, the retrieve/re-rank blend, and the shortlist depth, and we release a clean three-layer implementation.

Compact Vision Transformers are attractive for medical imaging in low-data and resource-constrained settings, but most existing variants assume that Euclidean latent geometry is sufficient for organizing image representations. We introduce hZACH-ViT, a family of curved-geometry extensions of ZACH-ViT, a compact zero-token Vision Transformer that removes positional embeddings and the class token and relies on global average pooling over patch representations. To isolate the role of geometry, we preserve the verified ZACH-ViT backbone and modify only the final representation space and prototype-based classifier head, enabling a controlled comparison between Euclidean, hyperbolic, and spherical latent geometries. We evaluate Poincaré, Klein, and spherical hZACH-ViT heads on seven MedMNIST datasets under an identical few-shot protocol with 50 samples per class and five random seeds. The completed benchmark contains 770 training runs spanning seven datasets, three non-Euclidean geometries, seven curvature magnitudes, and a Euclidean baseline. Across all seven datasets, the best non-Euclidean hZACH-ViT configuration improves over Euclidean ZACH-ViT, with an average gain of +0.021 in the dataset-specific primary metric and the largest improvement on OCTMNIST (+0.055 MacroF1). Fixed low-curvature configurations retain positive gains on the majority of datasets, and low curvature values (c = 0.1 or 0.2) account for six of the seven dataset-level winners. Rather than identifying a universally optimal manifold, our results establish geometry and curvature as dataset-dependent model-selection variables, with fixed low-curvature analyses confirming that gains persist beyond exhaustive per-dataset tuning.

Multi-modal Domain Generalization (MMDG) seeks to leverage complementary modalities to enhance model robustness on unseen domains. Despite extensive progress in Multi-modal Learning (MML) and Domain Generalization (DG) as individual fields, their systematic integration remains under-explored. Current MMDG research is largely confined to action recognition and lacks standardized evaluation protocols. To address this, we introduce MMDG-Bench, a comprehensive benchmark featuring two foundational frameworks: DG then MML (D2M) and MML then DG (M2D). We provide unified experimental protocols across diverse tasks, including video-audio-flow action recognition and RGB-Depth-IR face anti-spoofing. By instantiating ten MMDG baselines through pairing a unified MML configuration with five DG techniques under both D2M and M2D orderings, we demonstrate that these structured combinations frequently outperform existing state-of-the-art methods, underscoring the necessity of a unified benchmarking effort. Our analysis yields three key insights: (1) Integrating DG techniques provides consistent generalization gains across various backbones, whereas non-DG methods are highly sensitive to backbone shifts; (2) The optimal framework choice depends on inter-modal stability: D2M excels when modal relations are stable across domains, while M2D is more robust to cross-domain relational variance; (3) Stronger backbones yield amplified performance dividends when integrated into our structured frameworks. MMDG-Bench provides a principled foundation and actionable design guidelines for future research in multi-modal robustness. Code is released at https://github.com/qszhan/MMDG-Bench.

Ultrasound is the most widely used real-time imaging modality in clinical practice, yet per-frame video annotation remains a major bottleneck: expert labels are scarce and costly, and image appearance varies with speckle, shadowing, attenuation, and operator-dependent probe pose. This is especially limiting because clinically relevant information is often dynamic, from left-ventricular motion in echocardiography to muscle and bone kinematics in musculoskeletal imaging. Population atlases can amortize annotation cost by registering observations to a shared canonical coordinate system, but existing neural atlas methods mainly target single videos, small test-time image sets, or object-centric image collections. We introduce a cohort-scale neural atlas for ultrasound video: a single canonical chart with per-video Generative Latent Optimization embeddings, trained jointly over thousands of frames in DINOv3 feature space. Across five cardiac and musculoskeletal datasets with point landmarks and segmentation masks, our method learns coherent canonical templates and enables accurate atlas-space annotation transfer. On EchoNet-Dynamic and MSK-Bone, it supports single- and few-shot transfer with accuracy competitive with strong dense-correspondence baselines, while training in minutes on a single consumer GPU. The learned embeddings are interpretable: linear projections reveal structured cohort variation, image-decoder interpolation produces anatomically plausible intermediate frames, and test-time latent inversion reconstructs held-out frames through the atlas. These results suggest that cohort-scale neural atlases offer a practical, interpretable representation for reducing expert annotation burden in ultrasound video analysis.

Accurate segmentation is crucial for autonomous spacecraft, as it directly affects downstream tasks related to 3D situational awareness. The harsh illumination conditions of space, however, produce images with high variability in appearance, hindering the generalization of segmentation approaches across different spacecraft and environments. In this work, we propose GABI, a lightweight boundary-aware multi-task segmentation architecture that augments a convolutional backbone with an auxiliary distance-field prediction head. The distance field provides dense geometric supervision around object boundaries, encouraging the network to learn spatially consistent representations of spacecraft structures while maintaining low model complexity suitable for onboard perception systems. We evaluated GABI against both an established convolutional baseline and a heavier transformer-based architecture. On the SPARK benchmark, distance-field supervision improves the baseline by up to $5\%$ in Average Precision while achieving performance comparable to the transformer models. In generalization experiments, GABI improves Average Precision by more than $50\%$ over the baseline. In cross-domain evaluation, the lightweight GABI variant performs within $5\%$ in IoU and F1-score of the heavier transformer model while being approximately ten times smaller. At the same time, the heavier GABI variant surpasses the transformer architectures while remaining nearly three times lighter.

AI-generated image detection is a moving-target problem: detectors trained on one generator often fail when a new generator appears, and only a few labeled examples are available. We study a simple image-to-table formulation for this regime, where each image is encoded by a frozen DINOv3 backbone, its CLS feature is reduced to a 500-dimensional structured row with PCA, and TabPFN performs real/fake classification by in-context tabular inference rather than task-specific classifier training. This turns fake-image detection into low-data structured prediction over learned visual features, making detector adaptation depend on the labeled context set instead of gradient-based fine-tuning. On GenImage, LATTE, a recent state-of-the-art detector, remains stronger when many labeled samples from all generators are available, by 7.4% in the largest pooled setting, but DINOv3-PCA-TabPFN is stronger in the practically important low-data regime, outperforming LATTE by up to 8.2%, and in transfer settings where the detector must generalize from one generator to another. These results position tabular foundation models as a strong complementary adaptation mechanism for image forensics, shifting adaptation from detector retraining to lightweight in-context updates with a small labeled set of examples. Code URL: https://github.com/jpwalter30/Towards-Generalizable-Detection-of-AI-Generated-Images

Vision-language models (VLMs) are increasingly used to generate structured descriptions of street-level imagery for tasks such as streetscape auditing, mapping, and public consultation. These uses combine observable attributes with appraisal categories, and the human targets are often distributions of judgments with disagreement and explicit non-response. This paper argues that benchmarking VLMs for urban perception should treat disagreement and abstention as measurement outcomes, report inter-annotator reliability alongside model alignment, and treat the label space and scoring policy as negotiable artifacts when outputs are intended to inform urban governance. We ground the argument in a benchmark of 100 Montreal street scenes annotated along 30 dimensions by 12 participants from seven community organizations, and in a deterministic zero-shot evaluation of seven VLMs. Across dimensions, model agreement with human consensus co-varies with dimension-level human reliability, and for the appraisal dimension Overall Impression models and annotators exhibit distributional mismatch including different rates of Not applicable. We close with actions for benchmark creators, model developers, and institutions to make uncertainty and benchmark assumptions visible in evaluation reports.

Printed circuit board (PCB) defect detection is challenging because many defects are small and difficult to distinguish from complex background patterns. Most deep learning-based PCB inspection methods rely only on the inspected PCB image for defect detection, ignoring the defect-free reference image that encodes the expected layout of traces, pads, and other PCB structures. In this work, we propose RefDiffNet, a lightweight plug-and-play input enhancement block placed before the detector backbone to enhance the image before defect detection. RefDiffNet brings one proven idea from classical inspection into the deep learning era, using a defect-free reference image to reveal defects. RefDiffNet compares the defective image with the aligned reference, captures structural changes relative to the reference, and uses a lightweight encoder to output the original image with defective regions highlighted, thereby making the downstream detector's task easier. Results on HRIPCB and DeepPCB show that RefDiffNet consistently improves performance across detector families, including one-stage detectors from YOLOv8 to YOLOv26, the transformer-based RT-DETR, and the two-stage Faster R-CNN. It achieves up to 18% relative mAP50:95 gain with negligible overhead, introducing only 0.004 - 0.005M additional parameters and 0.7 - 0.8 GFLOPs, amounting to at most 0.25% of the parameter count of any evaluated detector. Results establish RefDiffNet as a lightweight, plug-and-play, detector-agnostic input enhancement module that substantially improves PCB defect detection with minimal computational cost.

Bounding-box regression is a fundamental component of object detection, playing a critical role in precise object localization. Existing Intersection-over-Union (IoU)-based loss functions extend the IoU objective by incorporating geometric penalties, such as center-distance and aspect-ratio mismatch, to improve bounding-box regression. However, these penalties typically remain fixed throughout training and do not account for the optimization dynamics in which predicted boxes initially exhibit large center-distance and shape errors, with later stages focusing on improving overlap with the ground truth. To address this limitation, we introduce MoEIoU, a mixture-of-experts based regression loss that jointly models overlap, center alignment, and aspect-ratio mismatch. MoEIoU aggregates these components using a log-sum-exp function, which emphasizes the dominant localization error while maintaining smooth contributions from other terms. Additionally, a curriculum-based weighting schedule is employed to prioritize correcting box position and shape in early training stages and improving overlap in later stages. We evaluated proposed MoEIoU on PASCAL VOC, HRIPCB, and MS COCO using multiple YOLO architectures, along with large-scale simulation experiments. It consistently outperforms standard and recent state-of-the-art losses, demonstrating faster convergence and improved localization accuracy. We further show that this adaptive aggregation improves existing IoU-based losses, yielding consistent gains and providing more effective optimization guidance for bounding-box regression in object detection frameworks.

EgoCross evaluates multimodal large language models on egocentric video question answering under substantial domain shift, where test videos come from surgery, industrial assembly, extreme sports, and animal-mounted cameras rather than ordinary daily-life scenes. In the source-limited track, the base model is fixed to Qwen3-VL-4B, while the official task-specific support set contains only 20 training samples. This setting makes the challenge less about model scaling and more about exposing the right visual, temporal, and answer-selection cues to a constrained model. Our key observation is that the frozen baseline model is not simply incapable of these rare scenarios; rather, it often fails to transfer its existing visual-language knowledge to the new task format without an appropriate interface. We therefore use a domain-wise inference strategy that treats the four target domains separately and designs different input, prompting, and answer-mapping procedures according to each domain's task characteristics. These strategies make the rare egocentric scenes more interpretable to the VLM by emphasizing the cues that matter for each domain. The resulting system is nearly training-free: surgery, and animal questions are answered with the base Qwen3-VL-4B model, while XSports and industry use only the official SFT checkpoint trained for two epochs on the provided 20 training samples. On the final evaluation, this simple strategy reaches 66.98% overall accuracy, suggesting that careful domain-aware inference can compensate for limited base-model strength and recover much of the ability already present in the baseline model.

Vision-Language Models (VLMs) have shown strong visual understanding and are increasingly deployed in embodied AI systems, where reliable perception under real conditions is essential. However, existing benchmarks assess VLMs using clean images or isolated perturbations rather than stresses caused by physical scene formation. This design has two limitations: it covers only a narrow subset of everyday visual stresses, and some perturbations rarely appear in realistic embodied scenes. This gap raises a fundamental question: how can we define visual stress in a principled way that captures the diverse factors encountered in physical environments? To address this question, we formulate visual perception from an inverse graphics perspective and introduce RoboStressBench, a benchmark for evaluating VLM robustness to physical visual stress in embodied scenes. Inspired by the physical rendering equation, RoboStressBench decomposes visual stress into four physically grounded dimensions: Material (M), Viewpoint (V), Lighting (L), and Geometry (G). This design enables RoboStressBench to cover a broad range of visual stresses in real-world environments, while allowing controlled analysis of their effects on VLM capabilities such as visual recognition, reasoning, and planning. Through comprehensive evaluations of state-of-the-art VLMs, we identify stress-specific failure modes and reveal that different physical factors degrade different embodied capabilities, which are often obscured by aggregate accuracy. We further introduce a stress-aware agentic solver that detects visual stressors and invokes visual-editing skills before reasoning, improving robustness in high-stress scenarios. Overall, RoboStressBench provides a principled evaluation framework for diagnosing and improving VLM perception under real-world physical stress, supporting the development of more reliable embodied AI systems.

AI glasses present a compelling platform for AI agents to serve as personalized memory assistants. To be genuinely useful, such systems must move beyond short-term video comprehension and address memory gaps that humans experience for practical, personal, or social purposes over longitudinal egocentric video streams. However, existing egocentric datasets predominantly focus on action recognition or generic QAs from short clips, measuring perceptual capabilities rather than realistic human memory needs. We introduce SuperMemory-VQA, an egocentric visual question answering (VQA) dataset for evaluating AI assistants on practical, long-horizon memory tasks. It contains 52.9 hours of everyday activities recorded with AI glasses, including synchronized RGB video, audio transcription, eye gaze, IMU, and SLAM trajectories. Through a human-verified annotation pipeline, we construct grounded 4,853 question-answer pairs that span object and location memory, intent recall, visual scene recall, timeline reconstruction, conversational memory, and in-context retrieval. Each question is posed as multiple-choice with an explicit "unanswerable" option to test hallucination robustness. Benchmarking leading agentic frameworks and LLM backbones reveals that existing systems remain far from reliable on real-world memory tasks, highlighting the need for new architectures for grounded AI memory that can answer only when evidence is sufficient. A participant survey further supports that our questions are realistic, useful, and aligned with everyday memory needs.

3D Gaussian Splatting (3DGS) renders new views of a scene in real time. Like every rasterizer, it answers only primary rays, the rays from the camera through the image. It cannot trace the secondary rays that shadows, ambient occlusion, and global illumination need. We turn a trained 3DGS scene into a ray oracle by distilling a Directed Distance Function (DDF). The DDF is a small neural field. It takes a ray, given by an origin and a direction, and returns the distance to the first surface and whether the ray hits anything. Each query is one forward pass. The field is 52~MB, and its size does not depend on the number of Gaussians, so its cost and memory stay flat as the scene grows. We make three points. First, we study what supervision a DDF needs. Depth rendered from the Gaussians is too blurry to teach thin parts, while clean distance supervision recovers them. Second, we measure speed. The DDF is 26 to 72 times faster than sphere tracing an equivalent signed distance field, and unlike a bounding volume hierarchy built over the Gaussians, even on dedicated RT-core hardware, its query time and memory do not grow with the scene. Third, we show a pipeline that needs no mesh: images give a 3DGS scene, a neural surface gives clean distances, and the DDF learns from them. We use the DDF as a secondary-ray oracle for global illumination. It reproduces reference ray-traced shadows at 30.3~dB and ambient occlusion at 21.3~dB across 142 objects, and on real captured scenes. Our codes are available at https://github.com/smlab-niser/ddf-gs.

Reconstructing the three-dimensional distribution of dark matter from weak-lensing observations is a central but highly ill-posed inverse problem in cosmology. Unlike standard 3D reconstruction with multiple viewpoints, we observe the universe from a single line of sight, through noisy shape distortions of galaxies with uncertain distances, so meaningful recovery of the 3D matter field requires strong prior assumptions. Existing methods either produce point estimates with handcrafted priors or use neural ensembles for approximate Bayesian uncertainty, and struggle to capture the non-Gaussian, filamentary structure of the cosmic web. With the advent of new high-resolution cosmological simulations, we now have an alternative source of prior knowledge that captures the nonlinear statistics of structure formation with far greater fidelity than analytic prescriptions. We leverage these simulations to build a new dataset $\texttt{Conicus3D}$, which enables us to learn a data-driven diffusion-model prior capturing the full 3D distribution of dark matter structure across cosmic time. Building on recent plug-and-play approaches, we modify a diffusion-based posterior sampling scheme to the 3D weak-lensing setting, combining the learned prior with a differentiable physical forward model. On realistic simulations targeting a modern weak lensing survey, our approach yields substantially improved 2D and 3D reconstruction accuracy over baseline methods. Moreover, it produces posterior samples whose statistics closely track the underlying simulations, while remaining robust to moderate shifts in cosmology.

Parameter compression of class-conditional diffusion models exposes a structural limitation in output-level distillation: supervising only the guided output leaves the two score branches non-identifiable, so the classifier-free guidance gap is not determined in the student. The zero-loss set admits degenerate solutions in which both branches collapse toward identical predictions, and guidance loses its effect at inference despite low training loss. For each fixed input the objective is exactly flat in a branch-output direction at every residual value, so the ambiguity is a property of the output-level objective rather than a poor local minimum. This paper introduces DASH, which supervises the conditional and unconditional branches independently. An anchor term regularises the conditional prediction toward ground-truth noise, and the teacher's final learned per-timestep curriculum transfers into the student as a frozen prior. Across CIFAR-10, CIFAR-100, and class-conditional ImageNet-64, a more than 5x compressed student stays within four FID points of its teacher and recovers at least 89% of the teacher's guidance-gap magnitude, where no other two-branch baseline with defined calibration exceeds 82%. Ablation isolates unconditional supervision as the term that separates this formulation from output-level distillation. The account is quantitative: the null space fixes the ratio between a composite student's two branch errors, and the measured ratios approximately match that prediction on all three datasets.

Recent advancements in video-based world models have demonstrated an unprecedented ability to synthesize high-fidelity visual sequences. However, a fundamental gap persists between visually plausible video generation and the functional requirements of a world model, particularly in maintaining a stable and reasonable internal state over extended temporal horizons. While existing benchmarks primarily emphasize visual quality, motion coherence, and text-video alignment, they largely overlook memory, the core capability of a world model to preserve consistency across long-term horizons and complex interactions. To address this gap, we present MBench, a comprehensive benchmark dedicated to quantifying and evaluating the memory capability of video world models. We systematically decompose the memory capability of video world models into three hierarchical and complementary core dimensions: entity consistency, environment consistency, and causal consistency, which are further refined into 12 quantifiable sub-dimensions for comprehensive characterization of long-term memory. Our benchmark is built upon rigorously curated real-captured long videos, and evaluated by rule-based quantitative matrices and VLM to enable objective and comprehensive consistency assessment. Extensive evaluations of mainstream state-of-the-art video world models reveal critical systemic limitations of existing methods in long-term state retention, providing a standardized benchmark and clear research direction to advance the field.

Cross-view geo-localization (CVGL) is critical for Unmanned Aerial Vehicle (UAV) self-positioning and target localization in GNSS-denied environments. However, acquiring robust semantics while preserving finegrained spatial details remains challenging. To address this, we propose DINO-GFSA, a framework leveraging a LoRA (Low-Rank Adaptation) adapted DINOv3 (ViTL) backbone for parameter-efficient, high-capacity representation. Crucially, we introduce a Semantic Gated Residual Fusion module, which utilizes high-level semantics to selectively calibrate and integrate low-level spatial cues, effectively bridging the semantic gap. Furthermore, a Mamba-based Sequential Aggregation Head is designed to capture long-range spatial dependencies with linear complexity. Experiments demonstrate state-of-the-art performance on University-1652 and DenseUAV benchmarks, notably surpassing the previous best on DenseUAV by 3.48% on Recall@1. These results validate DINO-GFSA as a generalized, robust solution for UAV CVGL.

Open-vocabulary object detection (OVD) has achieved remarkable progress through large-scale vision-language pre-training. Existing methods, however, typically formulate OVD as a discriminative prediction problem, where decoder queries are either static or initialized from encoder features, thus limiting their diversity and flexibility. In this paper, we introduce a generative perspective by modeling decoder query generation as a continuous transport process in latent space. We propose FlowOVD, a text-conditioned query generation framework based on rectified flow that progressively transforms text-agnostic queries into text-guided queries. By introducing continuous latent query dynamics into a vision-language model (VLM) based detector, our method avoids heuristic discrete query construction and enables more expressive semantic alignment for open-vocabulary detection. Without requiring additional training data, FlowOVD achieves 49.5 AP on COCO and 31.5 AP on LVIS, outperforming GroundingDINO by +1.2 AP (+2.5 %) and +4.1 AP (+15.0 %), respectively. The larger gain on the challenging long-tailed LVIS benchmark further highlights the effectiveness of continuous query generation for open-vocabulary generalization.

Video Moment Retrieval (VMR) task requires accurately localizing temporal boundaries aligned with natural language queries, but many models suffer from a misalignment between continuous surrogate losses and non-differentiable metrics, leading to optimization stagnation during the late stages of training and trapping boundary predictions in suboptimal solutions. Although Reinforcement Learning (RL) post-training successfully optimizes localization results for large models, applying it directly to lightweight networks easily disrupts the fragile feature representations established during the supervised phase. To overcome this optimization bottleneck, we propose Gradient-Isolated Reinforcement Learning for DETR (GIRL-DETR), introducing RL post-training into a lightweight temporal localization framework for the first time. The input video and text features first establish early alignment through Cross-Modal Interaction (CMI) before entering the transformer encoder. Subsequently, a Text-Guided Gating (TGG) mechanism dynamically injects semantic priors into the queries before the transformer decoder generates candidate proposals, providing high signal-to-noise ratio inputs for temporal prediction. After the supervised training reaches convergence, the backbone network is frozen to protect the feature manifold, while the detection head directly optimizes the non-differentiable evaluation metric tIoU to enhance localization accuracy through a Three-stage Progressive Reinforcement Learning (TPRL) strategy. This approach achieves an orthogonal decoupling of state representation and metric optimization. Experiments on Charades-STA, QVHighlights, and TACoS demonstrate that GIRL-DETR effectively resolves surrogate loss degradation and achieves substantial accuracy improvements with minimal parameter updates, providing a robust new pathway for RL applications in lightweight VMR models.

Visual Speech Recognition (VSR) aims to recognize speech from visual cues such as lip movements. Still, its performance is fundamentally limited by viseme ambiguity and pose-induced variations that introduce geometric distortions and occlusions. Existing approaches mainly rely on linguistic context or implicit invariance, leaving visual representations insufficiently robust under non-frontal views. In this work, we propose a pose-aware phoneme-level framework, termed HP-VSR-ResFiLM, that explicitly incorporates head-pose information into visual feature extraction. The proposed framework adopts a two-stage pipeline consisting of a pose-conditioned visual encoder in Stage~1 and a pretrained NLLB language model in Stage~2 for phoneme-to-text reconstruction. Specifically, Stage~1 incorporates a pose-conditioned residual Feature-wise Linear Modulation (FiLM) block after the 2D CNN frontend to refine visual representations adaptively using head-pose information. Experiments on LRS2 and LRS3 demonstrate that HP-VSR-ResFiLM achieves competitive performance under comparable training conditions, attaining word error rates (WER) of 24.7% and 30.3%, respectively, without relying on additional training data. Comprehensive ablation studies reveal that different FiLM modulation strategies exhibit complementary strengths: HP-VSR-FiLMFuse (L4) achieves the best overall performance on LRS2, whereas the proposed HP-VSR-ResFiLM provides the greatest improvements under large head-pose variations on LRS3 and consistently outperforms previous pose-aware methods for high-yaw samples. These results demonstrate the effectiveness of explicit pose-conditioned feature modulation for robust visual speech recognition in unconstrained settings.

For low-altitude Unmanned Aerial Vehicle (UAV) autonomy, 3D spatial understanding is not merely a perception objective, but the safety interface between human instructions and physical flight. In human-scale urban airspace below 20 meters, thin geometry, occlusions, vegetation, and urban clutter define whether an aerial agent can safely enter the space ahead. However, existing UAV datasets mainly provide 2D annotations or 3D boxes, while driving-oriented occupancy benchmarks assume stable ground-level sensor rigs. Both miss the defining regime of low-altitude flight: a front-facing monocular camera observing occupied and free space from a moving aerial body with frame-wise changing 6-DoF pose and camera extrinsics. To bridge this gap, we introduce SkyShield, to the best of our knowledge the first front-view monocular semantic occupancy benchmark for urban UAV flight below 20 meters. Built on CARLA, SkyShield contains 36K front-view UAV samples across diverse urban scenes and weather conditions, pairing each image with frame-wise 6-DoF UAV pose, frame-wise dynamic camera geometry, UAV states, and front-frustum semantic occupancy labels. We further propose KAR-mIoU, a UAV-centric and dynamics-aware metric that re-weights voxel-level evaluation by kinematic reachability and time-to-collision, revealing safety-critical risks hidden by conventional mIoU. To tackle this challenging new setting, we provide SkyOcc, a geometry-first monocular baseline that integrates frame-wise UAV attitude into projection, fuses temporal occupancy features, and applies safety-prior optimization to preserve sparse collision-critical structures. Together, SkyShield, KAR-mIoU, and SkyOcc establish occupancy as a safety interface for low-altitude aerial autonomy. Code and dataset will be released publicly.

Vision foundation models are bottlenecked by the quadratic cost of self-attention, which limits usable resolution and increases the cost of large-scale pretraining. Subquadratic alternatives such as linear attention and state-space models reduce this cost, but often serialize images into 1D token streams and weaken the 2D spatial structure important for vision. Generalized Spatial Propagation Networks (GSPN) instead propagate context directly on the 2D grid through line-scan recurrences, achieving near-linear complexity without positional embeddings, but have seen little use as foundation-scale encoders. We present C-GSPN, a foundation-scale vision encoder based on 2D spatial propagation. C-GSPN makes the operator practical through three improvements: (1) a fast GSPN CUDA kernel that fuses per-step launches into a single warp-specialized implementation with shared-memory tiling, coalesced access, and a compact multi-channel propagation, reaching over 90% of peak memory bandwidth and running up to 40–52x faster than the original GSPN implementation; (2) a compressed latent-space propagation block with fused normalization, which turns kernel-level speed into block- and model-level efficiency; and (3) a two-stage cross-operator distillation recipe that trains the new architecture from an attention teacher without the cost of from-scratch foundation-scale training. Distilled with 600M image-text pairs, C-GSPN matches an isomorphic ViT baseline with 15% fewer parameters, improves ADE20K segmentation by +2.1%, transfers to high resolution with a fraction of the data needed from scratch, and delivers a 4x end-to-end block speedup at 2K with single-pass, tiling-free inference.

Adversarial Training (AT) is a leading defense against adversarial examples but often suffers from Catastrophic Overfitting (CO) in efficient single-step variants, where robustness to multi-step attacks collapses despite high single-step performance. We address this failure mode with two contributions. First, we formalize Epsilon Overfitting (EO), a perspective in which fixed perturbation magnitudes and directions exacerbate CO, and show that introducing perturbation variability significantly improves robust generalization across different architectures and datasets. Second, we propose PertAlign (Perturbation Alignment), a theoretically grounded, computationally negligible metric that predicts CO onset by measuring gradient alignment across attack stages. Leveraging these insights, we introduce SORA, an adaptive step-size AT method that dynamically adjusts perturbations based on loss surface geometry. SORA consistently prevents CO, achieves state-of-the-art robustness and clean accuracy, and generalizes across datasets and architectures using a single fixed set of hyperparameters, which is essential for applicability in fast AT. Extensive experiments on diverse datasets and architectures show that SORA matches or surpasses the robustness of prior methods while delivering higher clean accuracy and superior efficiency. Code is available at https://github.com/SecondOrderAT/SORA.

The CASTLE Challenge @ EgoVis 2026 evaluates long-form egocentric video question answering over 600+ hours of multi-perspective recordings. Each four-choice question requires evidence from videos, transcripts, auxiliary photos, people, days, rooms, and temporal context. We propose an evidence-aware multimodal reasoning pipeline based on Qwen. Our system parses question hints, retrieves ASR chunks, attaches auxiliary images, samples candidate video frames, and routes questions into static visual, speech/text, temporal, and mixed types with specialized prompts. Multiple inference passes are aggregated by confidence-weighted voting and converted into the official Codabench format. In ablation, LoRA improves the score from 0.21 to 0.50, and more sampled frames further raise it to 0.58. Our final system ranks first in the CASTLE Challenge @ EgoVis 2026.

Cross-modal remote sensing image retrieval aims to retrieve semantically related scenes across heterogeneous sensing modalities. This remains challenging because paired observations may differ substantially in imaging physics, spatial resolution, spectral configuration, and visual appearance. Moreover, a single retrieval projection trained with one objective may be insufficient to jointly support cross-modal semantic alignment and same-modal neighbourhood preservation. We propose CR-JEPA, a Cross-modal Retrieval Joint-Embedding Predictive Architecture for dual-modality remote sensing retrieval. The model uses modality-specific stems, a shared transformer trunk, and JEPA-style predictive objectives to estimate masked latent target features within and across modalities. Inspired by LeJEPA, we apply Sketched Isotropic Gaussian Regularization to raw retrieval projections to stabilize embeddings and mitigate collapse. CR-JEPA further employs a decoupled-head design with a unified retrieval head for same-modal retrieval and a cross-modal retrieval head for cross-modal search. We evaluate CR-JEPA on BEN-14K, CBRSIR_VS, and DSRSID. On BEN-14K, CR-JEPA improves S1 to S2 retrieval from 61.23% to 75.82% and S2 to S1 retrieval from 63.73% to 75.40% over X-JEPA, while also achieving competitive same-modal retrieval with fewer parameters.

Real-world image super-resolution (Real-ISR) requires balancing structural fidelity to degraded observations with realistic detail synthesis. However, existing generative Real-ISR methods often rely on entangled conditioning mechanisms, leading to structural drift or semantically inconsistent details. To address this issue, we propose Visual In-Context Restoration (VICR), a Diffusion Transformer (DiT)-based framework that formulates Real-ISR as image completion. Specifically, we introduce a decoupled visual prior injection mechanism that derives local and global cues from the low-quality (LQ) image: local cues help recover image structures and support high-frequency detail synthesis, while global cues guide overall generation and promote semantic consistency. For ambiguous regions under severe degradation, VICR employs an inference-time agent to refine semantic prompts using visual evidence from the LQ input while keeping model parameters fixed. Experiments show that VICR achieves state-of-the-art performance across multiple Real-ISR benchmarks with only 127M trainable parameters.

Short-term anticipation in egocentric video requires more than recognizing the current scene: a system must infer which object the camera wearer will contact, which action will follow, and how soon the contact will happen. This report describes FROST-STA, our submission to the Ego4D Short-Term Object Interaction Anticipation (STA) Challenge at EgoVis 2026. For each query time, the model produces a ranked set of structured hypotheses containing an active-object box, noun label, verb label, time-to-contact (TTC), and confidence. FROST-STA builds on the V-JEPA 2.1 STA evaluation protocol, but adapts it to the challenge by using object-centric decoding, multi-head prediction, and a submission-oriented training and ensembling recipe. We keep the V-JEPA 2.1 ViT-G backbone fixed and extract two dense token streams: video tokens from a short clip resized to 384 pixels before the query, and image tokens from the last observed high-resolution frame. A compact alignment module, consisting of an attentive probe and frame-guided temporal pooling, maps the clip representation onto the spatial reference of the final frame before fusing it with image features. The fused maps are decoded by Faster R-CNN-style STA heads that estimate box offsets, nouns, verbs, TTC values, and interaction quality. For the final leaderboard entry, we train for 25 epochs with the official training split plus additional permitted validation annotations, and combine predictions across eight heads and checkpoints from epochs 15-25. FROST-STA obtains 5.13 Overall Top-5 mAP on the official test server, ranking second in the challenge and showing that frozen dense image-video features can serve as a strong basis for object-level interaction forecasting.

Multimodal MRI provides complementary information for neuroimaging analysis, where different imaging modalities capture distinct anatomical, tissue, and pathological features that support the development and evaluation of downstream AI applications. Although large-scale structural MRI resources are increasingly available, their modality coverage is often uneven across public and pooled neuroimaging datasets. This uneven modality coverage is further complicated by heterogeneity across sites, scanners, and acquisition protocols, as well as demographic and clinical variables that are often sparse, inconsistently recorded, or unavailable across studies. Synthetic MRI generation can help address this imbalance by synthesizing target-modality volumes for dataset augmentation and controlled synthetic cohort creation. However, many existing MRI synthesis approaches are trained on narrow modality sets or relatively homogeneous cohorts, limiting their applicability to large pooled neuroimaging resources where modality availability, acquisition protocols, and metadata coverage vary substantially across datasets. Diffusion models have become an attractive approach for MRI synthesis because of their strong sample fidelity and diversity, but sampling directly in 3D voxel space is computationally expensive and slow at inference. Latent diffusion improves practicality by synthesizing MRI in a learned, 3D latent space, although generation quality depends on the autoencoder's reconstruction fidelity and the resulting latent distribution. Our approach combines a Wavelet-Fusion variational autoencoder (WF-VAE) latent compressor with a conditional 3D U-Net diffusion model trained in the learned latent space using explicit modality and metadata conditioning. Our proposed Wavelet-Fusion Diffusion Model (WFDM) achieved the strongest distributional alignment among the evaluated synthetic MRI generators.

Single-stage fully sparse 3D object detectors rely on point clouds data to detect objects in autonomous driving scenarios. However, the sparsity and incompleteness of point clouds significantly limit the performance of 3D object detection. To address this issue, this paper proposes a point clouds completion method specifically designed for single-stage fully sparse detectors. The entire shape-prior-based completion process consists of two consecutive steps. In the first step, we design a novel Instance Selection module, which is capable of identifying point clouds corresponding to foreground objects even when the baseline model does not generate proposals, while effectively ignoring the point clouds of background regions. In the second step, we introduce a novel Alignment-Based Point Completion module, which aligns the point clouds of foreground objects with prototypes in terms of both their centers and orientations. Subsequently, points are selected from the prototype to fill in the missing parts of the foreground object. We evaluated our method on two single-stage fully sparse detectors using the KITTI dataset. The experimental results demonstrate that the proposed method significantly improves the detection performance, confirming its effectiveness and generalizability.

This work presents a configurable pipeline for generating semantic-segmentation-ready agricultural datasets from Sentinel-2 imagery and EuroCrops parcel-level annotations. The workflow transforms heterogeneous vector crop annotations into aligned multispectral image–mask pairs through label harmonization, Sentinel-2 product selection, spatial alignment, rasterization, patch extraction, quality filtering, and class-aware sample selection. The generated dataset contains 67,337 patches from five European countries and uses a reduced taxonomy of ten crop classes plus background. A four-level U-Net with Group Normalization was trained using 10 Sentinel-2 spectral bands and a composite loss combining class-weighted cross-entropy and Dice loss. On the internal EuroCrops-based test split, the model achieved a mean Intersection over Union (mIoU) of 0.7665, a pixel accuracy of 0.8693, and a mean class accuracy of 0.9072. Compared with spectral and spatial-context Random Forest baselines, the U-Net showed the importance of learned multi-scale spatial representations for crop segmentation. External evaluation was performed on unseen Belgian EuroCrops subsets, DACIA5, and PASTIS. The results show a clear performance gap under external and cross-dataset evaluation, especially for benchmarks with different taxonomies, annotation protocols, spatial coverage, or temporal organization. The model transfers more reliably to dominant and taxonomically aligned classes such as maize and wheat, while performance remains limited for several minority classes and for the adapted single-date PASTIS setting. These findings highlight both the potential and the limitations of using EuroCrops-derived supervision for Sentinel-2 crop segmentation under realistic domain shifts.