Papers

Unified image generation and editing models suffer from severe task interference in dense diffusion transformers architectures, where a shared parameter space must compromise between conflicting objectives (e.g., local editing v.s. subject-driven generation). While the sparse Mixture-of-Experts (MoE) paradigm is a promising solution, its gating networks remain task-agnostic, operating based on local features, unaware of global task intent. This task-agnostic nature prevents meaningful specialization and fails to resolve the underlying task interference. In this paper, we propose a novel framework to inject semantic intent into MoE routing. We introduce a Hierarchical Task Semantic Annotation scheme to create structured task descriptors (e.g., scope, type, preservation). We then design Predictive Alignment Regularization to align internal routing decisions with the task's high-level semantics. This regularization evolves the gating network from a task-agnostic executor to a dispatch center. Our model effectively mitigates task interference, outperforming dense baselines in fidelity and quality, and our analysis shows that experts naturally develop clear and semantically correlated specializations.

Semantic information in embodied AI is inherently multi-source and multi-stage, making it challenging to fully leverage for achieving stable perception-to-action loops in real-world environments. Early studies have combined manual engineering with deep neural networks, achieving notable progress in specific semantic-related embodied tasks. However, as embodied agents encounter increasingly complex environments and open-ended tasks, the demand for more generalizable and robust semantic processing capabilities has become imperative. Recent advances in foundation models (FMs) address this challenge through their cross-domain generalization abilities and rich semantic priors, reshaping the landscape of embodied AI research. In this survey, we propose the Semantic Lifecycle as a unified framework to characterize the evolution of semantic knowledge within embodied AI driven by foundation models. Departing from traditional paradigms that treat semantic processing as isolated modules or disjoint tasks, our framework offers a holistic perspective that captures the continuous flow and maintenance of semantic knowledge. Guided by this embodied semantic lifecycle, we further analyze and compare recent advances across three key stages: acquisition, representation, and storage. Finally, we summarize existing challenges and outline promising directions for future research.

Image registration under domain shift remains a fundamental challenge in computer vision and medical imaging: when source and target images exhibit systematic intensity differences, the brightness constancy assumption underlying conventional registration methods is violated, rendering correspondence estimation ill-posed. We propose SAR-Net, a unified framework that addresses this challenge through principled scene-appearance disentanglement. Our key insight is that observed images can be decomposed into domain-invariant scene representations and domain-specific appearance codes, enabling registration via re-rendering rather than direct intensity matching. We establish theoretical conditions under which this decomposition enables consistent cross-domain alignment (Proposition 1) and prove that our scene consistency loss provides a sufficient condition for geometric correspondence in the shared latent space (Proposition 2). Empirically, we validate SAR-Net on the ANHIR (Automatic Non-rigid Histological Image Registration) challenge benchmark, where multi-stain histopathology images exhibit coupled domain shift from different staining protocols and geometric distortion from tissue preparation. Our method achieves a median relative Target Registration Error (rTRE) of 0.25%, outperforming the state-of-the-art MEVIS method (0.27% rTRE) by 7.4%, with robustness of 99.1%. Code is available at https://github.com/D-ST-Sword/SAR-NET .

The proliferation of AI-generated imagery and sophisticated editing tools has rendered traditional forensic methods ineffective for cross-domain forgery detection. We present ForensicFormer, a hierarchical multi-scale framework that unifies low-level artifact detection, mid-level boundary analysis, and high-level semantic reasoning via cross-attention transformers. Unlike prior single-paradigm approaches, which achieve <75% accuracy on out-of-distribution datasets, our method maintains 86.8% average accuracy across seven diverse test sets, spanning traditional manipulations, GAN-generated images, and diffusion model outputs - a significant improvement over state-of-the-art universal detectors. We demonstrate superior robustness to JPEG compression (83% accuracy at Q=70 vs. 66% for baselines) and provide pixel-level forgery localization with a 0.76 F1-score. Extensive ablation studies validate that each hierarchical component contributes 4-10% accuracy improvement, and qualitative analysis reveals interpretable forensic features aligned with human expert reasoning. Our work bridges classical image forensics and modern deep learning, offering a practical solution for real-world deployment where manipulation techniques are unknown a priori.

Infrared and visible image fusion generates all-weather perception-capable images by combining complementary modalities, enhancing environmental awareness for intelligent unmanned systems. Existing methods either focus on pixel-level fusion while overlooking downstream task adaptability or implicitly learn rigid semantics through cascaded detection/segmentation models, unable to interactively address diverse semantic target perception needs. We propose CtrlFuse, a controllable image fusion framework that enables interactive dynamic fusion guided by mask prompts. The model integrates a multi-modal feature extractor, a reference prompt encoder (RPE), and a prompt-semantic fusion module (PSFM). The RPE dynamically encodes task-specific semantic prompts by fine-tuning pre-trained segmentation models with input mask guidance, while the PSFM explicitly injects these semantics into fusion features. Through synergistic optimization of parallel segmentation and fusion branches, our method achieves mutual enhancement between task performance and fusion quality. Experiments demonstrate state-of-the-art results in both fusion controllability and segmentation accuracy, with the adapted task branch even outperforming the original segmentation model.

Deep learning-based automatic medical image segmentation plays a critical role in clinical diagnosis and treatment planning but remains challenging in few-shot scenarios due to the scarcity of annotated training data. Recently, self-supervised foundation models such as DINOv3, which were trained on large natural image datasets, have shown strong potential for dense feature extraction that can help with the few-shot learning challenge. Yet, their direct application to medical images is hindered by domain differences. In this work, we propose DINO-AugSeg, a novel framework that leverages DINOv3 features to address the few-shot medical image segmentation challenge. Specifically, we introduce WT-Aug, a wavelet-based feature-level augmentation module that enriches the diversity of DINOv3-extracted features by perturbing frequency components, and CG-Fuse, a contextual information-guided fusion module that exploits cross-attention to integrate semantic-rich low-resolution features with spatially detailed high-resolution features. Extensive experiments on six public benchmarks spanning five imaging modalities, including MRI, CT, ultrasound, endoscopy, and dermoscopy, demonstrate that DINO-AugSeg consistently outperforms existing methods under limited-sample conditions. The results highlight the effectiveness of incorporating wavelet-domain augmentation and contextual fusion for robust feature representation, suggesting DINO-AugSeg as a promising direction for advancing few-shot medical image segmentation. Code and data will be made available on https://github.com/apple1986/DINO-AugSeg.

Data quality plays a central role in the performance and robustness of convolutional neural networks (CNNs) for image classification. While high-quality data is often preferred for training, real-world inputs are frequently affected by noise and other distortions. This paper investigates the effect of deliberately introducing controlled noise into the training data to improve model robustness. Using the CIFAR-10 dataset, we evaluate the impact of three common corruptions, namely Gaussian noise, Salt-and-Pepper noise, and Gaussian blur at varying intensities and training set pollution levels. Experiments using a Resnet-18 model reveal that incorporating just 10% noisy data during training is sufficient to significantly reduce test loss and enhance accuracy under fully corrupted test conditions, with minimal impact on clean-data performance. These findings suggest that strategic exposure to noise can act as a simple yet effective regularizer, offering a practical trade-off between traditional data cleanliness and real-world resilience.

Scientific image manipulation in biomedical publications poses a growing threat to research integrity and reproducibility. Unlike natural image forensics, biomedical forgery detection is uniquely challenging due to domain-specific artifacts, complex textures, and unstructured figure layouts. We present the first vision-language guided framework for both generating and detecting biomedical image forgeries. By combining diffusion-based synthesis with vision-language prompting, our method enables realistic and semantically controlled manipulations, including duplication, splicing, and region removal, across diverse biomedical modalities. We introduce Rescind, a large-scale benchmark featuring fine-grained annotations and modality-specific splits, and propose Integscan, a structured state space modeling framework that integrates attention-enhanced visual encoding with prompt-conditioned semantic alignment for precise forgery localization. To ensure semantic fidelity, we incorporate a vision-language model based verification loop that filters generated forgeries based on consistency with intended prompts. Extensive experiments on Rescind and existing benchmarks demonstrate that Integscan achieves state of the art performance in both detection and localization, establishing a strong foundation for automated scientific integrity analysis.

We introduce Fiducial Exoskeletons, an image-based reformulation of 3D robot state estimation that replaces cumbersome procedures and motor-centric pipelines with single-image inference. Traditional approaches - especially robot-camera extrinsic estimation - often rely on high-precision actuators and require time-consuming routines such as hand-eye calibration. In contrast, modern learning-based robot control is increasingly trained and deployed from RGB observations on lower-cost hardware. Our key insight is twofold. First, we cast robot state estimation as 6D pose estimation of each link from a single RGB image: the robot-camera base transform is obtained directly as the estimated base-link pose, and the joint state is recovered via a lightweight global optimization that enforces kinematic consistency with the observed link poses (optionally warm-started with encoder readings). Second, we make per-link 6D pose estimation robust and simple - even without learning - by introducing the fiducial exoskeleton: a lightweight 3D-printed mount with a fiducial marker on each link and known marker-link geometry. This design yields robust camera-robot extrinsics, per-link SE(3) poses, and joint-angle state from a single image, enabling robust state estimation even on unplugged robots. Demonstrated on a low-cost robot arm, fiducial exoskeletons substantially simplify setup while improving calibration, state accuracy, and downstream 3D control performance. We release code and printable hardware designs to enable further algorithm-hardware co-design.

Scientific compound figures combine multiple labeled panels into a single image, and downstream pretraining and retrieval require panel-aligned visual-text pairs. However, in a PubMed Central (PMC)-scale crawl of 346,567 compound figures, 16.3% have no caption and are discarded by existing caption-decomposition pipelines. We propose FigEx2, a visual-conditioned framework that takes only a compound figure as input and jointly produces labeled panel boxes and panel-wise captions. FigEx2 introduces an Entity-Attention Kullback-Leibler (KL) regularizer that aligns the detector's cross-attention with scientific entities annotated for each panel, providing a stable conditioning signal that also improves localization, and applies Group Relative Policy Optimization (GRPO) with a panel-level Entity-F1 reward to optimize scientific faithfulness. We curate BioSci-Fig-Cap for in-domain supervision and contribute physics and chemistry test suites for cross-disciplinary evaluation. FigEx2 achieves 0.751 mAP@0.5:0.95 on BioSci-Fig-Cap, and outperforms Qwen3-VL-8B by 6.80 Entity-F1 on MedICaT for captioning. It also transfers zero-shot to out-of-distribution domains. The source code is available at https://github.com/Huang-AI4Medicine-Lab/FigEx2.

Maintaining or improving the performance of Deep Neural Networks (DNNs) through fine-tuning requires labeling newly collected inputs, a process that is often costly and time-consuming. To alleviate this problem, input selection approaches have been developed in recent years to identify small, yet highly informative subsets for labeling. Diversity-based selection is one of the most effective approaches for this purpose. However, they are often computationally intensive and lack scalability for large input sets, limiting their practical applicability. To address this challenge, we introduce Concept-Based Diversity (CBD), a novel and highly efficient diversity metric for image inputs that leverages Vision-Language Models (VLMs). Our results show that CBD exhibits a strong correlation with Geometric Diversity (GD), an established diversity metric, while requiring only a fraction of its computation time. Building on this finding, we propose a hybrid input selection approach that combines CBD with Margin, a simple uncertainty metric. We conduct a comprehensive evaluation across a diverse set of DNN models, input sets, selection budgets, and six most effective state-of-the-art selection baselines. The results demonstrate that the CBD-based selection consistently outperforms all baselines at guiding input selection to improve the DNN model. Furthermore, the CBD-based selection approach remains highly efficient, requiring selection times close to those of simple uncertainty-based methods such as Margin, even on larger input sets like ImageNet. These results confirm not only the effectiveness and computational advantage of the CBD-based approach, particularly compared to hybrid baselines, but also its scalability in repetitive and extensive input selection scenarios.

Zero-Shot image Anomaly Detection (ZSAD) aims to detect and localise anomalies without access to any normal training samples of the target data. While recent ZSAD approaches leverage additional modalities such as language to generate fine-grained prompts for localisation, vision-only methods remain limited to image-level classification, lacking spatial precision. In this work, we introduce a simple yet effective training-free vision-only ZSAD framework that circumvents the need for fine-grained prompts by leveraging the inversion of a pretrained Denoising Diffusion Implicit Model (DDIM). Specifically, given an input image and a generic text description (e.g., "an image of an [object class]"), we invert the image to obtain latent representations and initiate the denoising process from a fixed intermediate timestep to reconstruct the image. Since the underlying diffusion model is trained solely on normal data, this process yields a normal-looking reconstruction. The discrepancy between the input image and the reconstructed one highlights potential anomalies. Our method achieves state-of-the-art performance on VISA dataset, demonstrating strong localisation capabilities without auxiliary modalities and facilitating a shift away from prompt dependence for zero-shot anomaly detection research. Code is available at https://github.com/giddyyupp/DIVAD.

We show that for a variety of concepts in adapter-based vision-language models, the representations of their images and their text descriptions are meaningfully aligned from the very first layer. This contradicts the established view that such image-text alignment only appears in late layers. We show this using a new synthesis-based method inspired by DeepDream: given a textual concept such as "Jupiter", we extract its concept vector at a given layer, and then use optimisation to synthesise an image whose representation aligns with that vector. We apply our approach to hundreds of concepts across seven layers in Gemma 3, and find that the synthesised images often depict salient visual features of the targeted textual concepts: for example, already at layer 1, more than 50 % of images depict recognisable features of animals, activities, or seasons. Our method thus provides direct, constructive evidence of image-text alignment on a concept-by-concept and layer-by-layer basis. Unlike previous methods for measuring multimodal alignment, our approach is simple, fast, and does not require auxiliary models or datasets. It also offers a new path towards model interpretability, by providing a way to visualise a model's representation space by backtracing through its image processing components.

Current text-conditioned diffusion editors handle single object replacement well but struggle when a new object and a new style must be introduced simultaneously. We present Twin-Prompt Attention Blend (TP-Blend), a lightweight training-free framework that receives two separate textual prompts, one specifying a blend object and the other defining a target style, and injects both into a single denoising trajectory. TP-Blend is driven by two complementary attention processors. Cross-Attention Object Fusion (CAOF) first averages head-wise attention to locate spatial tokens that respond strongly to either prompt, then solves an entropy-regularised optimal transport problem that reassigns complete multi-head feature vectors to those positions. CAOF updates feature vectors at the full combined dimensionality of all heads (e.g., 640 dimensions in SD-XL), preserving rich cross-head correlations while keeping memory low. Self-Attention Style Fusion (SASF) injects style at every self-attention layer through Detail-Sensitive Instance Normalization. A lightweight one-dimensional Gaussian filter separates low- and high-frequency components; only the high-frequency residual is blended back, imprinting brush-stroke-level texture without disrupting global geometry. SASF further swaps the Key and Value matrices with those derived from the style prompt, enforcing context-aware texture modulation that remains independent of object fusion. Extensive experiments show that TP-Blend produces high-resolution, photo-realistic edits with precise control over both content and appearance, surpassing recent baselines in quantitative fidelity, perceptual quality, and inference speed.

Vision-language models achieve strong performance across a wide range of multimodal understanding and reasoning tasks, yet their multi-step reasoning remains unstable. Repeated sampling over the same input often produces divergent reasoning trajectories and inconsistent final predictions. To address this, we introduce two complementary approaches inspired by test-time scaling: (1) CASHEW, an inference-time framework that stabilizes reasoning by iteratively aggregating multiple candidate trajectories into higher-quality reasoning traces, with visual verification of object presence providing grounding cues during aggregation, and (2) CASHEW-RL, a learned variant that internalizes this aggregation behavior within a single model. CASHEW-RL is trained using Group Sequence Policy Optimization (GSPO) with a composite reward that encourages correct answers grounded in minimal yet sufficient visual evidence, while adaptively allocating reasoning effort based on task difficulty. This training objective enables robust self-aggregation at inference. Extensive experiments on 13 image understanding, video understanding, and video reasoning benchmarks show significant performance improvements, with CASHEW/CASHEW-RL achieving average gains of +4.1/+5.3 percentage points on Qwen3-VL-4B and +2.6/+3.0 percentage points on Qwen3-VL-8B, respectively.

Tear film (TF) breakup is a key driver of understanding dry eye disease, yet estimating TF thickness and osmolarity from fluorescence (FL) imaging typically requires solving computationally expensive inverse problems. We propose an operator learning framework that replaces traditional inverse solvers with neural operators trained on simulated TF dynamics. This approach offers a scalable path toward rapid, data-driven analysis of tear film dynamics.

Understanding human visual search behavior is a fundamental problem in vision science and computer vision, with direct implications for modeling how observers allocate attention in location-unknown search tasks. In this study, we investigate the relationship between Gabor-based features and gray-level co-occurrence matrix (GLCM) based texture features in modeling early-stage visual search behavior. Two feature-combination pipelines are proposed to integrate Gabor and GLCM features for narrowing the region of possible human fixations. The pipelines are evaluated using simulated digital breast tomosynthesis images. Results show qualitative agreement among fixation candidates predicted by the proposed pipelines and a threshold-based model observer. A strong correlation is observed between GLCM mean and Gabor feature responses, indicating that these features encode related image information despite their different formulations. Eye-tracking data from human observers further suggest consistency between predicted fixation regions and early-stage gaze behavior. These findings highlight the value of combining structural and texture-based features for modeling visual search and support the development of perceptually informed observer models.

We introduce VULCA-Bench, a multicultural art-critique benchmark for evaluating Vision-Language Models' (VLMs) cultural understanding beyond surface-level visual perception. Existing VLM benchmarks predominantly measure L1-L2 capabilities (object recognition, scene description, and factual question answering) while under-evaluate higher-order cultural interpretation. VULCA-Bench contains 7,410 matched image-critique pairs spanning eight cultural traditions, with Chinese-English bilingual coverage. We operationalise cultural understanding using a five-layer framework (L1-L5, from Visual Perception to Philosophical Aesthetics), instantiated as 225 culture-specific dimensions and supported by expert-written bilingual critiques. Our pilot results indicate that higher-layer reasoning (L3-L5) is consistently more challenging than visual and technical analysis (L1-L2). The dataset, evaluation scripts, and annotation tools are available under CC BY 4.0 at https://github.com/yha9806/VULCA-Bench.

We develop a new statistical ideal observer model that performs holistic visual search (or gist) processing in part by placing thresholds on minimum extractable image features. In this model, the ideal observer reduces the number of free parameters thereby shrinking down the system. The applications of this novel framework is in medical image perception (for optimizing imaging systems and algorithms), computer vision, benchmarking performance and enabling feature selection/evaluations. Other applications are in target detection and recognition in defense/security as well as evaluating sensors and detectors.

This study advances task-based image quality assessment by developing an anthropomorphic thresholded visual-search model observer. The model is an ideal observer for thresholded data inspired by the human visual system, allowing selective processing of high-salience features to improve discrimination performance. By filtering out irrelevant variability, the model enhances diagnostic accuracy and computational efficiency. The observer employs a two-stage framework: candidate selection and decision-making. Using thresholded data during candidate selection refines regions of interest, while stage-specific feature processing optimizes performance. Simulations were conducted to evaluate the effects of thresholding on feature maps, candidate localization, and multi-feature scenarios. Results demonstrate that thresholding improves observer performance by excluding low-salience features, particularly in noisy environments. Intermediate thresholds often outperform no thresholding, indicating that retaining only relevant features is more effective than keeping all features. Additionally, the model demonstrates effective training with fewer images while maintaining alignment with human performance. These findings suggest that the proposed novel framework can predict human visual search performance in clinically realistic tasks and provide solutions for model observer training with limited resources. Our novel approach has applications in other areas where human visual search and detection tasks are modeled such as in computer vision, machine learning, defense and security image analysis.

High-resolution UAV photogrammetry has become a key technology for precision agriculture, enabling centimeter-level crop monitoring and point-level plant localization. However, point-level maize localization in UAV imagery remains challenging due to (1) extremely small object-to-pixel ratios, typically less than 0.1%, (2) prohibitive computational costs of quadratic attention on ultra-high-resolution images larger than 3000 x 4000 pixels, and (3) agricultural scene-specific complexities such as sparse object distribution and environmental variability that are poorly handled by general-purpose vision models. To address these challenges, we propose the Additive Kolmogorov-Arnold Transformer (AKT), which replaces conventional multilayer perceptrons with Pade Kolmogorov-Arnold Network (PKAN) modules to enhance functional expressivity for small-object feature extraction, and introduces PKAN Additive Attention (PAA) to model multiscale spatial dependencies with reduced computational complexity. In addition, we present the Point-based Maize Localization (PML) dataset, consisting of 1,928 high-resolution UAV images with approximately 501,000 point annotations collected under real field conditions. Extensive experiments show that AKT achieves an average F1-score of 62.8%, outperforming state-of-the-art methods by 4.2%, while reducing FLOPs by 12.6% and improving inference throughput by 20.7%. For downstream tasks, AKT attains a mean absolute error of 7.1 in stand counting and a root mean square error of 1.95-1.97 cm in interplant spacing estimation. These results demonstrate that integrating Kolmogorov-Arnold representation theory with efficient attention mechanisms offers an effective framework for high-resolution agricultural remote sensing.

This paper presents the Sesame Plant Segmentation Dataset, an open source annotated image dataset designed to support the development of artificial intelligence models for agricultural applications, with a specific focus on sesame plants. The dataset comprises 206 training images, 43 validation images, and 43 test images in YOLO compatible segmentation format, capturing sesame plants at early growth stages under varying environmental conditions. Data were collected using a high resolution mobile camera from farms in Jirdede, Daura Local Government Area, Katsina State, Nigeria, and annotated using the Segment Anything Model version 2 with farmer supervision. Unlike conventional bounding box datasets, this dataset employs pixel level segmentation to enable more precise detection and analysis of sesame plants in real world farm settings. Model evaluation using the Ultralytics YOLOv8 framework demonstrated strong performance for both detection and segmentation tasks. For bounding box detection, the model achieved a recall of 79 percent, precision of 79 percent, mean average precision at IoU 0.50 of 84 percent, and mean average precision from 0.50 to 0.95 of 58 percent. For segmentation, it achieved a recall of 82 percent, precision of 77 percent, mean average precision at IoU 0.50 of 84 percent, and mean average precision from 0.50 to 0.95 of 52 percent. The dataset represents a novel contribution to sesame focused agricultural vision datasets in Nigeria and supports applications such as plant monitoring, yield estimation, and agricultural research.

The transformative potential of 3D content creation has been progressively unlocked through advancements in generative models. Recently, intuitive drag editing with geometric changes has attracted significant attention in 2D editing yet remains challenging for 3D scenes. In this paper, we introduce 3DGS-Drag – a point-based 3D editing framework that provides efficient, intuitive drag manipulation of real 3D scenes. Our approach bridges the gap between deformation-based and 2D-editing-based 3D editing methods, addressing their limitations to geometry-related content editing. We leverage two key innovations: deformation guidance utilizing 3D Gaussian Splatting for consistent geometric modifications and diffusion guidance for content correction and visual quality enhancement. A progressive editing strategy further supports aggressive 3D drag edits. Our method enables a wide range of edits, including motion change, shape adjustment, inpainting, and content extension. Experimental results demonstrate the effectiveness of 3DGS-Drag in various scenes, achieving state-of-the-art performance in geometry-related 3D content editing. Notably, the editing is efficient, taking 10 to 20 minutes on a single RTX 4090 GPU.

Maize disease classification plays a vital role in mitigating yield losses and ensuring food security. However, the deployment of traditional disease detection models in resource-constrained environments, such as those using smartphones and drones, faces challenges due to high computational costs. To address these challenges, we propose LWMSCNN-SE, a lightweight convolutional neural network (CNN) that integrates multi-scale feature extraction, depthwise separable convolutions, and squeeze-and-Excitation (SE) attention mechanisms. This novel combination enables the model to achieve 96.63% classification accuracy with only 241,348 parameters and 0.666 GFLOPs, making it suitable for real-time deployment in field applications. Our approach addresses the accuracy–efficiency trade-off by delivering high accuracy while maintaining low computational costs, demonstrating its potential for efficient maize disease diagnosis on edge devices in precision farming systems.

The dataset spans diverse artistic styles, including regionally grounded aesthetics from the Middle East, Northern Europe, East Asia, and South Asia, alongside general categories such as sketch and oil painting. All images are generated using the Moonworks Lunara model and intentionally crafted to embody distinct, high-quality aesthetic styles, yielding a first-of-its-kind dataset with substantially higher aesthetic scores, exceeding even aesthetics-focused datasets, and general-purpose datasets by a larger margin. Each image is accompanied by a human-refined prompt and structured annotations that jointly describe salient objects, attributes, relationships, and stylistic cues. Unlike large-scale web-derived datasets that emphasize breadth over precision, the Lunara Aesthetic Dataset prioritizes aesthetic quality, stylistic diversity, and licensing transparency, and is released under the Apache 2.0 license to support research and unrestricted academic and commercial use.

Large Language Models have emerged as transformative tools for Security Operations Centers, enabling automated log analysis, phishing triage, and malware explanation; however, deployment in adversarial cybersecurity environments exposes critical vulnerabilities to prompt injection attacks where malicious instructions embedded in security artifacts manipulate model behavior. This paper introduces SecureCAI, a novel defense framework extending Constitutional AI principles with security-aware guardrails, adaptive constitution evolution, and Direct Preference Optimization for unlearning unsafe response patterns, addressing the unique challenges of high-stakes security contexts where traditional safety mechanisms prove insufficient against sophisticated adversarial manipulation. Experimental evaluation demonstrates that SecureCAI reduces attack success rates by 94.7% compared to baseline models while maintaining 95.1% accuracy on benign security analysis tasks, with the framework incorporating continuous red-teaming feedback loops enabling dynamic adaptation to emerging attack strategies and achieving constitution adherence scores exceeding 0.92 under sustained adversarial pressure, thereby establishing a foundation for trustworthy integration of language model capabilities into operational cybersecurity workflows and addressing a critical gap in current approaches to AI safety within adversarial domains.

We present RefVFX, a new framework that transfers complex temporal effects from a reference video onto a target video or image in a feed-forward manner. While existing methods excel at prompt-based or keyframe-conditioned editing, they struggle with dynamic temporal effects such as dynamic lighting changes or character transformations, which are difficult to describe via text or static conditions. Transferring a video effect is challenging, as the model must integrate the new temporal dynamics with the input video's existing motion and appearance. % To address this, we introduce a large-scale dataset of triplets, where each triplet consists of a reference effect video, an input image or video, and a corresponding output video depicting the transferred effect. Creating this data is non-trivial, especially the video-to-video effect triplets, which do not exist naturally. To generate these, we propose a scalable automated pipeline that creates high-quality paired videos designed to preserve the input's motion and structure while transforming it based on some fixed, repeatable effect. We then augment this data with image-to-video effects derived from LoRA adapters and code-based temporal effects generated through programmatic composition. Building on our new dataset, we train our reference-conditioned model using recent text-to-video backbones. Experimental results demonstrate that RefVFX produces visually consistent and temporally coherent edits, generalizes across unseen effect categories, and outperforms prompt-only baselines in both quantitative metrics and human preference. See our website at https://snap-research.github.io/RefVFX/

While the Transformer architecture dominates many fields, its quadratic self-attention complexity hinders its use in large-scale applications. Linear attention offers an efficient alternative, but its direct application often degrades performance, with existing fixes typically re-introducing computational overhead through extra modules (e.g., depthwise separable convolution) that defeat the original purpose. In this work, we identify a key failure mode in these methods: global context collapse, where the model loses representational diversity. To address this, we propose Multi-Head Linear Attention (MHLA), which preserves this diversity by computing attention within divided heads along the token dimension. We prove that MHLA maintains linear complexity while recovering much of the expressive power of softmax attention, and verify its effectiveness across multiple domains, achieving a 3.6% improvement on ImageNet classification, a 6.3% gain on NLP, a 12.6% improvement on image generation, and a 41% enhancement on video generation under the same time complexity.

Video world models have emerged as promising candidates for high-fidelity world models, offering the potential to synthesize high-quality videos capturing fine-grained interactions between agents and their environments conditioned on multi-modal user inputs. Their impressive capabilities address many of the long-standing challenges faced by physics-based simulators, driving broad adoption in many problem domains, e.g., robotics. For example, video models can generate photorealistic, physically consistent deformable-body simulation without making prohibitive simplifying assumptions, which is a major bottleneck in physics-based simulation. Moreover, video models can serve as foundation world models that capture the dynamics of the world in a fine-grained and expressive way. They thus overcome the limited expressiveness of language-only abstractions in describing intricate physical interactions. However, despite their potential, video models still generate physics-violating future predictions, often manifesting as hallucinations. In this survey, we provide a review of video models and their applications as embodied world models in robotics, including efficient data generation and policy learning, dynamics and rewards modeling in reinforcement learning, policy evaluation, and visual planning. Further, we highlight important challenges hindering the trustworthy integration of video models, such as poor instruction following, hallucinations like violations of physics, unsafe content generation, in addition to significant data and compute overhead. We present potential future directions to address these open research challenges to motivate research and ultimately facilitate broader applications, especially in safety-critical settings. We provide a curated bibliography at https://github.com/irom-princeton/awesome-robotics-video-world-model-papers .

Large Vision Language Models (LVLMs) have demonstrated remarkable capabilities, yet their proficiency in understanding and reasoning over multiple images remains largely unexplored. While existing benchmarks have initiated the evaluation of multi-image models, a comprehensive analysis of their core weaknesses and their causes is still lacking. In this work, we introduce MIMIC (Multi-Image Model Insights and Challenges), a new benchmark designed to rigorously evaluate the multi-image capabilities of LVLMs. Using MIMIC, we conduct a series of diagnostic experiments that reveal pervasive issues: LVLMs often fail to aggregate information across images and struggle to track or attend to multiple concepts simultaneously. To address these failures, we propose two novel complementary remedies. On the data side, we present a procedural data-generation strategy that composes single-image annotations into rich, targeted multi-image training examples. On the optimization side, we analyze layer-wise attention patterns and derive an attention-masking scheme tailored for multi-image inputs. Experiments substantially improved cross-image aggregation, while also enhancing performance on existing multi-image benchmarks, outperforming prior state of the art across tasks. Data and code will be made available at https://github.com/anurag-198/MIMIC.