Authors: Nirmal Elamon, Rouzbeh Davoudi
Abstract: The field of object detection and understanding is rapidly evolving, driven by advances in both traditional CNN-based models and emerging multi-modal large language models (LLMs). While CNNs like ResNet and YOLO remain highly effective for image-based tasks, recent transformer-based LLMs introduce new capabilities such as dynamic context reasoning, language-guided prompts, and holistic scene understanding. However, when used out-of-the-box, the full potential of LLMs remains underexploited, often resulting in suboptimal performance on specialized visual tasks. In this work, we conduct a comprehensive comparison of fine-tuned traditional CNNs, zero-shot pre-trained multi-modal LLMs, and fine-tuned multi-modal LLMs on the challenging task of artificial text overlay detection in images. A key contribution of our study is demonstrating that LLMs can be effectively fine-tuned on very limited data (fewer than 1,000 images) to achieve up to 36% accuracy improvement, matching or surpassing CNN-based baselines that typically require orders of magnitude more data. By exploring how language-guided models can be adapted for precise visual understanding with minimal supervision, our work contributes to the broader effort of bridging vision and language, offering novel insights into efficient cross-modal learning strategies. These findings highlight the adaptability and data efficiency of LLM-based approaches for real-world object detection tasks and provide actionable guidance for applying multi-modal transformers in low-resource visual environments. To support continued progress in this area, we have made the code used to fine-tune the models available in our GitHub, enabling future improvements and reuse in related applications.
Authors: Saumya B
Abstract: Brain tumor segmentation is crucial for diagnosis and treatment planning, yet challenges such as class imbalance and limited model generalization continue to hinder progress. This work presents a reproducible evaluation of U-Net segmentation performance on brain tumor MRI using focal loss and basic data augmentation strategies. Experiments were conducted on a publicly available MRI dataset, focusing on focal loss parameter tuning and assessing the impact of three data augmentation techniques: horizontal flip, rotation, and scaling. The U-Net with focal loss achieved a precision of 90%, comparable to state-of-the-art results. By making all code and results publicly available, this study establishes a transparent, reproducible baseline to guide future research on augmentation strategies and loss function design in brain tumor segmentation.
Authors: Hyeonggeun Han, Sehwan Kim, Hyungjun Joo, Sangwoo Hong, Jungwoo Lee
Abstract: Despite their impressive generative capabilities, text-to-image diffusion models often memorize and replicate training data, prompting serious concerns over privacy and copyright. Recent work has attributed this memorization to an attraction basin-a region where applying classifier-free guidance (CFG) steers the denoising trajectory toward memorized outputs-and has proposed deferring CFG application until the denoising trajectory escapes this basin. However, such delays often result in non-memorized images that are poorly aligned with the input prompts, highlighting the need to promote earlier escape so that CFG can be applied sooner in the denoising process. In this work, we show that the initial noise sample plays a crucial role in determining when this escape occurs. We empirically observe that different initial samples lead to varying escape times. Building on this insight, we propose two mitigation strategies that adjust the initial noise-either collectively or individually-to find and utilize initial samples that encourage earlier basin escape. These approaches significantly reduce memorization while preserving image-text alignment.
Authors: Siiri Lepp\"alampi, Sonja M. Hyrynsalmi, Erno Vanhala
Abstract: Generative AI offers vast opportunities for creating visualisations, such as graphics, videos, and images. However, recent studies around AI-generated visualisations have primarily focused on the creation process and image quality, overlooking representational biases. This study addresses this gap by testing representation biases in AI-generated pictures in an occupational setting and evaluating how two AI image generator tools, DALL-E 3 and Ideogram, compare. Additionally, the study discusses topics such as ageing and emotions in AI-generated images. As AI image tools are becoming more widely used, addressing and mitigating harmful gender biases becomes essential to ensure diverse representation in media and professional settings. In this study, over 750 AI-generated images of occupations were prompted. The thematic analysis results revealed that both DALL-E 3 and Ideogram reinforce traditional gender stereotypes in AI-generated images, although to varying degrees. These findings emphasise that AI visualisation tools risk reinforcing narrow representations. In our discussion section, we propose suggestions for practitioners, individuals and researchers to increase representation when generating images with visible genders.
Authors: Jort Vincenti, Metod Jazbec, Guoxuan Xia
Abstract: Visual Autoregressive Models (VAR) offer efficient and high-quality image generation but suffer from computational redundancy due to repeated Transformer calls at increasing resolutions. We introduce a dynamic Mixture-of-Experts router integrated into VAR. The new architecture allows to trade compute for quality through scale-aware thresholding. This thresholding strategy balances expert selection based on token complexity and resolution, without requiring additional training. As a result, we achieve 20% fewer FLOPs, 11% faster inference and match the image quality achieved by the dense baseline.
Authors: Hanieh Shojaei Miandashti, Claus Brenner
Abstract: In addition to accurate scene understanding through precise semantic segmentation of LiDAR point clouds, detecting out-of-distribution (OOD) objects, instances not encountered during training, is essential to prevent the incorrect assignment of unknown objects to known classes. While supervised OOD detection methods depend on auxiliary OOD datasets, unsupervised methods avoid this requirement but typically rely on predictive entropy, the entropy of the predictive distribution obtained by averaging over an ensemble or multiple posterior weight samples. However, these methods often conflate epistemic (model) and aleatoric (data) uncertainties, misclassifying ambiguous in distribution regions as OOD. To address this issue, we present an unsupervised OOD detection approach that employs epistemic uncertainty derived from hierarchical Bayesian modeling of Gaussian Mixture Model (GMM) parameters in the feature space of a deep neural network. Without requiring auxiliary data or additional training stages, our approach outperforms existing uncertainty-based methods on the SemanticKITTI dataset, achieving an 18\% improvement in AUROC, 22\% increase in AUPRC, and 36\% reduction in FPR95 (from 76\% to 40\%), compared to the predictive entropy approach used in prior works.
Authors: Conor McCarthy, Loes Quirijnen, Jan Peter van Zandwijk, Zeno Geradts, Marcel Worring
Abstract: Within Human Activity Recognition (HAR), there is an insurmountable gap between the range of activities performed in life and those that can be captured in an annotated sensor dataset used in training. Failure to properly handle unseen activities seriously undermines any HAR classifier's reliability. Additionally within HAR, not all classes are equally dissimilar, some significantly overlap or encompass other sub-activities. Based on these observations, we arrange activity classes into a structured hierarchy. From there, we propose Hi-OSCAR: a Hierarchical Open-set Classifier for Activity Recognition, that can identify known activities at state-of-the-art accuracy while simultaneously rejecting unknown activities. This not only enables open-set classification, but also allows for unknown classes to be localized to the nearest internal node, providing insight beyond a binary "known/unknown" classification. To facilitate this and future open-set HAR research, we collected a new dataset: NFI_FARED. NFI_FARED contains data from multiple subjects performing nineteen activities from a range of contexts, including daily living, commuting, and rapid movements, which is fully public and available for download.
Authors: Mostafa Mohammadpour, Mehdi Zekriyapanah Gashti, Yusif S. Gasimov
Abstract: High-frequency oscillations (HFOs) are a new biomarker for identifying the epileptogenic zone. Mapping HFO-generating regions can improve the precision of resection sites in patients with refractory epilepsy. However, detecting HFOs remains challenging, and their clinical features are not yet fully defined. Visual identification of HFOs is time-consuming, labor-intensive, and subjective. As a result, developing automated methods to detect HFOs is critical for research and clinical use. In this study, we developed a novel method for detecting HFOs in the ripple and fast ripple frequency bands (80-500 Hz). We validated it using both controlled datasets and data from epilepsy patients. Our method employs an unsupervised clustering technique to categorize events extracted from the time-frequency domain using the S-transform. The proposed detector differentiates HFOs events from spikes, background activity, and artifacts. Compared to existing detectors, our method achieved a sensitivity of 97.67%, a precision of 98.57%, and an F-score of 97.78% on the controlled dataset. In epilepsy patients, our results showed a stronger correlation with surgical outcomes, with a ratio of 0.73 between HFOs rates in resected versus non-resected contacts. The study confirmed previous findings that HFOs are promising biomarkers of epileptogenicity in epileptic patients. Removing HFOs, especially fast ripple, leads to seizure freedom, while remaining HFOs lead to seizure recurrence.
Authors: Thomas Fel, Binxu Wang, Michael A. Lepori, Matthew Kowal, Andrew Lee, Randall Balestriero, Sonia Joseph, Ekdeep S. Lubana, Talia Konkle, Demba Ba, Martin Wattenberg
Abstract: DINOv2 is routinely deployed to recognize objects, scenes, and actions; yet the nature of what it perceives remains unknown. As a working baseline, we adopt the Linear Representation Hypothesis (LRH) and operationalize it using SAEs, producing a 32,000-unit dictionary that serves as the interpretability backbone of our study, which unfolds in three parts. In the first part, we analyze how different downstream tasks recruit concepts from our learned dictionary, revealing functional specialization: classification exploits "Elsewhere" concepts that fire everywhere except on target objects, implementing learned negations; segmentation relies on boundary detectors forming coherent subspaces; depth estimation draws on three distinct monocular depth cues matching visual neuroscience principles. Following these functional results, we analyze the geometry and statistics of the concepts learned by the SAE. We found that representations are partly dense rather than strictly sparse. The dictionary evolves toward greater coherence and departs from maximally orthogonal ideals (Grassmannian frames). Within an image, tokens occupy a low dimensional, locally connected set persisting after removing position. These signs suggest representations are organized beyond linear sparsity alone. Synthesizing these observations, we propose a refined view: tokens are formed by combining convex mixtures of archetypes (e.g., a rabbit among animals, brown among colors, fluffy among textures). This structure is grounded in Gardenfors' conceptual spaces and in the model's mechanism as multi-head attention produces sums of convex mixtures, defining regions bounded by archetypes. We introduce the Minkowski Representation Hypothesis (MRH) and examine its empirical signatures and implications for interpreting vision-transformer representations.
Authors: Zhe Dong, Yuzhe Sun, Haochen Jiang, Tianzhu Liu, Yanfeng Gu
Abstract: Remote sensing images inevitably suffer from various degradation factors during acquisition, including atmospheric interference, sensor limitations, and imaging conditions. These complex and heterogeneous degradations pose severe challenges to image quality and downstream interpretation tasks. Addressing limitations of existing all-in-one restoration methods that overly rely on implicit feature representations and lack explicit modeling of degradation physics, this paper proposes Physics-Guided Degradation-Adaptive Experts (PhyDAE). The method employs a two-stage cascaded architecture transforming degradation information from implicit features into explicit decision signals, enabling precise identification and differentiated processing of multiple heterogeneous degradations including haze, noise, blur, and low-light conditions. The model incorporates progressive degradation mining and exploitation mechanisms, where the Residual Manifold Projector (RMP) and Frequency-Aware Degradation Decomposer (FADD) comprehensively analyze degradation characteristics from manifold geometry and frequency perspectives. Physics-aware expert modules and temperature-controlled sparse activation strategies are introduced to enhance computational efficiency while ensuring imaging physics consistency. Extensive experiments on three benchmark datasets (MD-RSID, MD-RRSHID, and MDRS-Landsat) demonstrate that PhyDAE achieves superior performance across all four restoration tasks, comprehensively outperforming state-of-the-art methods. Notably, PhyDAE substantially improves restoration quality while achieving significant reductions in parameter count and computational complexity, resulting in remarkable efficiency gains compared to mainstream approaches and achieving optimal balance between performance and efficiency. Code is available at https://github.com/HIT-SIRS/PhyDAE.
Authors: Songtao Jiang, Yuan Wang, Sibo Song, Tianxiang Hu, Chenyi Zhou, Bin Pu, Yan Zhang, Zhibo Yang, Yang Feng, Joey Tianyi Zhou, Jin Hao, Zijian Chen, Ruijia Wu, Tao Tang, Junhui Lv, Hongxia Xu, Hongwei Wang, Jun Xiao, Bin Feng, Fudong Zhu, Kenli Li, Weidi Xie, Jimeng Sun, Jian Wu, Zuozhu Liu
Abstract: Real-world clinical decision-making grapples with integrating information from diverse data modalities, including medical text, 2D/3D images, and video, leading to inefficiencies and potential diagnostic oversights. While generalist vision-language models (VLMs) offer promise, their medical development faces challenges of opaque pipelines, data scarcity, and architectural inflexibility. Here we present Hulu-Med, a transparent medical VLM that unifies understanding across all these modalities. Built upon a unified patch-based vision encoder and an LLM decoder, Hulu-Med was progressively trained on 16.7 million (M) samples to scale from 2D to 3D and video comprehension. The medical-aware token reduction enables efficient training, requiring only 4,000 to 40,000 GPU hours for 7B to 32B parameter variants. Extensive evaluation across 30 benchmarks exhibits state-of-the-art performance, surpassing leading open-source models and competing with proprietary systems in tasks spanning visual question-answering, medical report generation, and complex reasoning in multilingual and rare disease scenarios. By open-sourcing our complete pipeline, we establish that high-performance medical VLM can be achieved transparently, providing a foundational tool for accessible and impactful clinical AI. Code is released on \href{https://github.com/ZJUI-AI4H/Hulu-Med}{https://github.com/ZJUI-AI4H/Hulu-Med}.
URLs: https://github.com/ZJUI-AI4H/Hulu-Med, https://github.com/ZJUI-AI4H/Hulu-Med
Authors: Kang Liao, Size Wu, Zhonghua Wu, Linyi Jin, Chao Wang, Yikai Wang, Fei Wang, Wei Li, Chen Change Loy
Abstract: Camera-centric understanding and generation are two cornerstones of spatial intelligence, yet they are typically studied in isolation. We present Puffin, a unified camera-centric multimodal model that extends spatial awareness along the camera dimension. Puffin integrates language regression and diffusion-based generation to interpret and create scenes from arbitrary viewpoints. To bridge the modality gap between cameras and vision-language, we introduce a novel paradigm that treats camera as language, enabling thinking with camera. This guides the model to align spatially grounded visual cues with photographic terminology while reasoning across geometric context. Puffin is trained on Puffin-4M, a large-scale dataset of 4 million vision-language-camera triplets. We incorporate both global camera parameters and pixel-wise camera maps, yielding flexible and reliable spatial generation. Experiments demonstrate Puffin superior performance over specialized models for camera-centric generation and understanding. With instruction tuning, Puffin generalizes to diverse cross-view tasks such as spatial imagination, world exploration, and photography guidance. We will release the code, models, dataset pipeline, and benchmark to advance multimodal spatial intelligence research.
Authors: Nicolas Ewen, Jairo Diaz-Rodriguez, Kelly Ramsay
Abstract: Traditional transfer learning typically reuses large pre-trained networks by freezing some of their weights and adding task-specific layers. While this approach is computationally efficient, it limits the model's ability to adapt to domain-specific features and can still lead to overfitting with very limited data. To address these limitations, we propose Structured Output Regularization (SOR), a simple yet effective framework that freezes the internal network structures (e.g., convolutional filters) while using a combination of group lasso and $L_1$ penalties. This framework tailors the model to specific data with minimal additional parameters and is easily applicable to various network components, such as convolutional filters or various blocks in neural networks enabling broad applicability for transfer learning tasks. We evaluate SOR on three few shot medical imaging classification tasks and we achieve competitive results using DenseNet121, and EfficientNetB4 bases compared to established benchmarks.
Authors: Yu Qi, Haibo Zhao, Ziyu Guo, Siyuan Ma, Ziyan Chen, Yaokun Han, Renrui Zhang, Zitiantao Lin, Shiji Xin, Yijian Huang, Kai Cheng, Peiheng Wang, Jiazheng Liu, Jiayi Zhang, Yizhe Zhu, Wenqing Wang, Yiran Qin, Xupeng Zhu, Haojie Huang, Lawson L. S. Wong
Abstract: Embodied capabilities refer to a suite of fundamental abilities for an agent to perceive, comprehend, and interact with the physical world. While multimodal large language models (MLLMs) show promise as embodied agents, a thorough and systematic evaluation of their embodied capabilities remains underexplored, as existing benchmarks primarily focus on specific domains such as planning or spatial understanding. To bridge this gap, we introduce BEAR, a comprehensive and fine-grained benchmark that evaluates MLLMs on atomic embodied capabilities. BEAR comprises 4,469 interleaved image-video-text entries across 14 domains in 6 categories, including tasks from low-level pointing, trajectory understanding, spatial reasoning, to high-level planning. Extensive evaluation results of 20 representative MLLMs reveal their persistent limitations across all domains of embodied capabilities. To tackle the shortfall, we propose BEAR-Agent, a multimodal conversable agent that integrates pretrained vision models to strengthen MLLM perception, 3D understanding, and planning capabilities. It substantially enhances MLLM performance across diverse embodied capabilities on BEAR, yielding a 9.12% absolute gain and a relative improvement of 17.5% on GPT-5. Furthermore, our experiments indicate that improving MLLM embodied capabilities can benefit embodied tasks in simulated environments. Project website: https://bear-official66.github.io/
Authors: Jiayang Liu, Daniel Tso, Yiming Bu, Qinru Qiu
Abstract: Adversarial attacks significantly challenge the safe deployment of deep learning models, particularly in real-world applications. Traditional defenses often rely on computationally intensive optimization (e.g., adversarial training or data augmentation) to improve robustness, whereas the human visual system achieves inherent robustness to adversarial perturbations through evolved biological mechanisms. We hypothesize that attention guided non-homogeneous sparse sampling and predictive coding plays a key role in this robustness. To test this hypothesis, we propose a novel defense framework incorporating three key biological mechanisms: foveal-peripheral processing, saccadic eye movements, and cortical filling-in. Our approach employs reinforcement learning-guided saccades to selectively capture multiple foveal-peripheral glimpses, which are integrated into a reconstructed image before classification. This biologically inspired preprocessing effectively mitigates adversarial noise, preserves semantic integrity, and notably requires no retraining or fine-tuning of downstream classifiers, enabling seamless integration with existing systems. Experiments on the ImageNet dataset demonstrate that our method improves system robustness across diverse classifiers and attack types, while significantly reducing training overhead compared to both biologically and non-biologically inspired defense techniques.
Authors: Gregory Yeghiyan, Jurius Azar, Devson Butani, Chan-Jin Chung
Abstract: This paper presents a real-time spill detection system that utilizes pretrained deep learning models with RGB and thermal imaging to classify spill vs. no-spill scenarios across varied environments. Using a balanced binary dataset (4,000 images), our experiments demonstrate the advantages of thermal imaging in inference speed, accuracy, and model size. We achieve up to 100% accuracy using lightweight models like VGG19 and NasNetMobile, with thermal models performing faster and more robustly across different lighting conditions. Our system runs on consumer-grade hardware (RTX 4080) and achieves inference times as low as 44 ms with model sizes under 350 MB, highlighting its deployability in safety-critical contexts. Results from experiments with a real robot and test datasets indicate that a VGG19 model trained on thermal imaging performs best.
Authors: Xiaohui Li, Shaobin Zhuang, Shuo Cao, Yang Yang, Yuandong Pu, Qi Qin, Siqi Luo, Bin Fu, Yihao Liu
Abstract: Generative models for Image Super-Resolution (SR) are increasingly powerful, yet their reliance on self-attention's quadratic complexity (O(N^2)) creates a major computational bottleneck. Linear Attention offers an O(N) solution, but its promise for photorealistic SR has remained largely untapped, historically hindered by a cascade of interrelated and previously unsolved challenges. This paper introduces LinearSR, a holistic framework that, for the first time, systematically overcomes these critical hurdles. Specifically, we resolve a fundamental, training instability that causes catastrophic model divergence using our novel "knee point"-based Early-Stopping Guided Fine-tuning (ESGF) strategy. Furthermore, we mitigate the classic perception-distortion trade-off with a dedicated SNR-based Mixture of Experts (MoE) architecture. Finally, we establish an effective and lightweight guidance paradigm, TAG, derived from our "precision-over-volume" principle. Our resulting LinearSR model simultaneously delivers state-of-the-art perceptual quality with exceptional efficiency. Its core diffusion forward pass (1-NFE) achieves SOTA-level speed, while its overall multi-step inference time remains highly competitive. This work provides the first robust methodology for applying Linear Attention in the photorealistic SR domain, establishing a foundational paradigm for future research in efficient generative super-resolution.
Authors: Paula Maddigan, Andrew Lensen, Rachael C. Shaw
Abstract: Accurate recognition and re-identification of individual animals is essential for successful wildlife population monitoring. Traditional methods, such as leg banding of birds, are time consuming and invasive. Recent progress in artificial intelligence, particularly computer vision, offers encouraging solutions for smart conservation and efficient automation. This study presents a unique pipeline for extracting high-quality key frames from videos of k\={a}k\={a} (Nestor meridionalis), a threatened forest-dwelling parrot in New Zealand. Key frame extraction is well-studied in person re-identification, however, its application to wildlife is limited. Using video recordings at a custom-built feeder, we extract key frames and evaluate the re-identification performance of our pipeline. Our unsupervised methodology combines object detection using YOLO and Grounding DINO, optical flow blur detection, image encoding with DINOv2, and clustering methods to identify representative key frames. The results indicate that our proposed key frame selection methods yield image collections which achieve high accuracy in k\={a}k\={a} re-identification, providing a foundation for future research using media collected in more diverse and challenging environments. Through the use of artificial intelligence and computer vision, our non-invasive and efficient approach provides a valuable alternative to traditional physical tagging methods for recognising k\={a}k\={a} individuals and therefore improving the monitoring of populations. This research contributes to developing fresh approaches in wildlife monitoring, with applications in ecology and conservation biology.
Authors: Shuo Xing, Soumik Dey, Mingyang Wu, Ashirbad Mishra, Hansi Wu, Binbin Li, Zhengzhong Tu
Abstract: Video quality assessment (VQA) is a fundamental computer vision task that aims to predict the perceptual quality of a given video in alignment with human judgments. Existing performant VQA models trained with direct score supervision suffer from (1) poor generalization across diverse content and tasks, ranging from user-generated content (UGC), short-form videos, to AI-generated content (AIGC), (2) limited interpretability, and (3) lack of extensibility to novel use cases or content types. We propose Q-Router, an agentic framework for universal VQA with a multi-tier model routing system. Q-Router integrates a diverse set of expert models and employs vision--language models (VLMs) as real-time routers that dynamically reason and then ensemble the most appropriate experts conditioned on the input video semantics. We build a multi-tiered routing system based on the computing budget, with the heaviest tier involving a specific spatiotemporal artifacts localization for interpretability. This agentic design enables Q-Router to combine the complementary strengths of specialized experts, achieving both flexibility and robustness in delivering consistent performance across heterogeneous video sources and tasks. Extensive experiments demonstrate that Q-Router matches or surpasses state-of-the-art VQA models on a variety of benchmarks, while substantially improving generalization and interpretability. Moreover, Q-Router excels on the quality-based question answering benchmark, Q-Bench-Video, highlighting its promise as a foundation for next-generation VQA systems. Finally, we show that Q-Router capably localizes spatiotemporal artifacts, showing potential as a reward function for post-training video generation models.
Authors: Yuanhao Zou, Zhaozheng Yin
Abstract: Medical Visual Question Answering (Med-VQA) is a challenging task that requires a deep understanding of both medical images and textual questions. Although recent works leveraging Medical Vision-Language Pre-training (Med-VLP) have shown strong performance on the Med-VQA task, there is still no unified solution for modality alignment, and the issue of hard negatives remains under-explored. Additionally, commonly used knowledge fusion techniques for Med-VQA may introduce irrelevant information. In this work, we propose a framework to address these challenges through three key contributions: (1) a unified solution for heterogeneous modality alignments across multiple levels, modalities, views, and stages, leveraging methods like contrastive learning and optimal transport theory; (2) a hard negative mining method that employs soft labels for multi-modality alignments and enforces the hard negative pair discrimination; and (3) a Gated Cross-Attention Module for Med-VQA that integrates the answer vocabulary as prior knowledge and selects relevant information from it. Our framework outperforms the previous state-of-the-art on widely used Med-VQA datasets like RAD-VQA, SLAKE, PathVQA and VQA-2019.
Authors: Rohan Choudhury, Shanchuan Lin, Jianyi Wang, Hao Chen, Qi Zhao, Feng Cheng, Lu Jiang, Kris Kitani, Laszlo A. Jeni
Abstract: Diffusion-based super-resolution (SR) is a key component in video generation and video restoration, but is slow and expensive, limiting scalability to higher resolutions and longer videos. Our key insight is that many regions in video are inherently low-detail and gain little from refinement, yet current methods process all pixels uniformly. To take advantage of this, we propose SkipSR, a simple framework for accelerating video SR by identifying low-detail regions directly from low-resolution input, then skipping computation on them entirely, only super-resolving the areas that require refinement. This simple yet effective strategy preserves perceptual quality in both standard and one-step diffusion SR models while significantly reducing computation. In standard SR benchmarks, our method achieves up to 60% faster end-to-end latency than prior models on 720p videos with no perceptible loss in quality. Video demos are available at https://rccchoudhury.github.io/skipsr/
Authors: Yiyang Huang, Yizhou Wang, Yun Fu
Abstract: Video large language models (Vid-LLMs), which excel in diverse video-language tasks, can be effectively constructed by adapting image-pretrained vision-language models (VLMs). However, this adaptation remains challenging, as it requires processing dense and temporally extended visual inputs that exceed the capacity of image-based models. This paper identifies the perception bottleneck and token overload as key challenges in extending image-based VLMs to the video domain. To address these issues, we propose D-CoDe, a training-free adaptation framework that incorporates dynamic compression and question decomposition. Specifically, dynamic compression alleviates the perception bottleneck through adaptive selection of representative frames and content-aware aggregation of spatial tokens, thereby reducing redundancy while preserving informative content. In parallel, question decomposition mitigates token overload by reformulating the original query into sub-questions, guiding the model to focus on distinct aspects of the video and enabling more comprehensive understanding. Experiments demonstrate that D-CoDe effectively improves video understanding across various benchmarks. Furthermore, strong performance on the challenging long-video benchmark highlights the potential of D-CoDe in handling complex video-language tasks. Code is available at https://github.com/hukcc/D-CoDe.
Authors: Hongrui Wu, Zhicheng Gao, Jin Cao, Kelu Yao, Wen Shen, Zhihua Wei
Abstract: Open-vocabulary 3D instance segmentation seeks to segment and classify instances beyond the annotated label space. Existing methods typically map 3D instances to 2D RGB-D images, and then employ vision-language models (VLMs) for classification. However, such a mapping strategy usually introduces noise from 2D occlusions and incurs substantial computational and memory costs during inference, slowing down the inference speed. To address the above problems, we propose a Fast Open-vocabulary 3D instance segmentation method via Label-guided Knowledge distillation (FOLK). Our core idea is to design a teacher model that extracts high-quality instance embeddings and distills its open-vocabulary knowledge into a 3D student model. In this way, during inference, the distilled 3D model can directly classify instances from the 3D point cloud, avoiding noise caused by occlusions and significantly accelerating the inference process. Specifically, we first design a teacher model to generate a 2D CLIP embedding for each 3D instance, incorporating both visibility and viewpoint diversity, which serves as the learning target for distillation. We then develop a 3D student model that directly produces a 3D embedding for each 3D instance. During training, we propose a label-guided distillation algorithm to distill open-vocabulary knowledge from label-consistent 2D embeddings into the student model. FOLK conducted experiments on the ScanNet200 and Replica datasets, achieving state-of-the-art performance on the ScanNet200 dataset with an AP50 score of 35.7, while running approximately 6.0x to 152.2x faster than previous methods. All codes will be released after the paper is accepted.
Authors: Ronan John, Anis Chihoub, Ryan Meegan, Gina Sidelli, Jeffery Neyhart, Peter Oudemans, Kristin Dana
Abstract: Change monitoring is an essential task for cranberry farming as it provides both breeders and growers with the ability to analyze growth, predict yield, and make treatment decisions. However, this task is often done manually, requiring significant time on the part of a cranberry grower or breeder. Deep learning based change monitoring holds promise, despite the caveat of hard-to-interpret high dimensional features and hand-annotations for fine-tuning. To address this gap, we introduce a method for modeling crop growth based on fine-tuning vision transformers (ViTs) using a self-supervised approach that avoids tedious image annotations. We use a two-fold pretext task (time regression and class prediction) to learn a latent space for the time-lapse evolution of plant and fruit appearance. The resulting 2D temporal tracks provide an interpretable time-series model of crop growth that can be used to: 1) predict growth over time and 2) distinguish temporal differences of cranberry varieties. We also provide a novel time-lapse dataset of cranberry fruit featuring eight distinct varieties, observed 52 times over the growing season (span of around four months), annotated with information about fungicide application, yield, and rot. Our approach is general and can be applied to other crops and applications (code and dataset can be found at https://github. com/ronan-39/tlt/).
URLs: https://github.
Authors: Daiki Yoshikawa, Takashi Matsubara
Abstract: Vision-language models have achieved remarkable success in multi-modal representation learning from large-scale pairs of visual scenes and linguistic descriptions. However, they still struggle to simultaneously express two distinct types of semantic structures: the hierarchy within a concept family (e.g., dog $\preceq$ mammal $\preceq$ animal) and the compositionality across different concept families (e.g., "a dog in a car" $\preceq$ dog, car). Recent works have addressed this challenge by employing hyperbolic space, which efficiently captures tree-like hierarchy, yet its suitability for representing compositionality remains unclear. To resolve this dilemma, we propose PHyCLIP, which employs an $\ell_1$-Product metric on a Cartesian product of Hyperbolic factors. With our design, intra-family hierarchies emerge within individual hyperbolic factors, and cross-family composition is captured by the $\ell_1$-product metric, analogous to a Boolean algebra. Experiments on zero-shot classification, retrieval, hierarchical classification, and compositional understanding tasks demonstrate that PHyCLIP outperforms existing single-space approaches and offers more interpretable structures in the embedding space.
Authors: Yufei Song, Ziqi Zhou, Qi Lu, Hangtao Zhang, Yifan Hu, Lulu Xue, Shengshan Hu, Minghui Li, Leo Yu Zhang
Abstract: Segmentation models exhibit significant vulnerability to adversarial examples in white-box settings, but existing adversarial attack methods often show poor transferability across different segmentation models. While some researchers have explored transfer-based adversarial attack (i.e., transfer attack) methods for segmentation models, the complex contextual dependencies within these models and the feature distribution gaps between surrogate and target models result in unsatisfactory transfer success rates. To address these issues, we propose SegTrans, a novel transfer attack framework that divides the input sample into multiple local regions and remaps their semantic information to generate diverse enhanced samples. These enhanced samples replace the original ones for perturbation optimization, thereby improving the transferability of adversarial examples across different segmentation models. Unlike existing methods, SegTrans only retains local semantic information from the original input, rather than using global semantic information to optimize perturbations. Extensive experiments on two benchmark datasets, PASCAL VOC and Cityscapes, four different segmentation models, and three backbone networks show that SegTrans significantly improves adversarial transfer success rates without introducing additional computational overhead. Compared to the current state-of-the-art methods, SegTrans achieves an average increase of 8.55% in transfer attack success rate and improves computational efficiency by more than 100%.
Authors: Han Hu, Zhuoran Zheng, Chen Lyu
Abstract: Knowledge distillation (KD) attacks pose a significant threat to deep model intellectual property by enabling adversaries to train student networks using a teacher model's outputs. While recent defenses in image classification have successfully disrupted KD by perturbing output probabilities, extending these methods to image restoration is difficult. Unlike classification, restoration is a generative task with continuous, high-dimensional outputs that depend on spatial coherence and fine details. Minor perturbations are often insufficient, as students can still learn the underlying mapping.To address this, we propose Adaptive Singular Value Perturbation (ASVP), a runtime defense tailored for image restoration models. ASVP operates on internal feature maps of the teacher using singular value decomposition (SVD). It amplifies the topk singular values to inject structured, high-frequency perturbations, disrupting the alignment needed for distillation. This hinders student learning while preserving the teacher's output quality.We evaluate ASVP across five image restoration tasks: super-resolution, low-light enhancement, underwater enhancement, dehazing, and deraining. Experiments show ASVP reduces student PSNR by up to 4 dB and SSIM by 60-75%, with negligible impact on the teacher's performance. Compared to prior methods, ASVP offers a stronger and more consistent defense.Our approach provides a practical solution to protect open-source restoration models from unauthorized knowledge distillation.
Authors: Zixi Yang, Jiapeng Li, Muxi Diao, Yinuo Jing, Kongming Liang
Abstract: Recently, Multi-modal Large Language Models (MLLMs) have demonstrated significant performance across various video understanding tasks. However, their robustness, particularly when faced with manipulated video content, remains largely unexplored. In this paper, we introduce Ro-Bench, the first benchmark for evaluating MLLMs on dynamic out-of-distribution (OOD) counterfactual video test sets. Ro-Bench incorporates high-quality, diverse and temporally relevant video data, by editing Style, Object, Background and their compositions. We evaluated eight recent video MLLMs and found that current models exhibit substantial performance degradation on Ro-Bench when exposed to counterfactual video content. Furthermore, we demonstrate that fine-tuning MLLMs with counterfactual data enhances robustness, achieving a 21.73% performance increase on Ro-Bench and a 12.78% improvement across 20 tasks in the MVBench dataset. These findings underscore the effectiveness of counterfactual data in enhancing the video understanding ability of MLLMs. The code and data will be released shortly.
Authors: Nisha Pillai, Aditi Virupakshaiah, Harrison W. Smith, Amanda J. Ashworth, Prasanna Gowda, Phillip R. Owens, Adam R. Rivers, Bindu Nanduri, Mahalingam Ramkumar
Abstract: Modern agricultural operations increasingly rely on integrated monitoring systems that combine multiple data sources for farm optimization. Aerial drone-based animal health monitoring serves as a key component but faces limited data availability, compounded by scene-specific issues such as small, occluded, or partially visible animals. Transfer learning approaches often fail to address this limitation due to the unavailability of large datasets that reflect specific farm conditions, including variations in animal breeds, environments, and behaviors. Therefore, there is a need for developing a problem-specific, animal-focused data augmentation strategy tailored to these unique challenges. To address this gap, we propose an object-focused data augmentation framework designed explicitly for animal health monitoring in constrained data settings. Our approach segments animals from backgrounds and augments them through transformations and diffusion-based synthesis to create realistic, diverse scenes that enhance animal detection and monitoring performance. Our initial experiments demonstrate that our augmented dataset yields superior performance compared to our baseline models on the animal detection task. By generating domain-specific data, our method empowers real-time animal health monitoring solutions even in data-scarce scenarios, bridging the gap between limited data and practical applicability.
Authors: Yifan Li, Zhenghao Chen, Ziheng Wu, Kun Zhou, Ruipu Luo, Can Zhang, Zhentao He, Yufei Zhan, Wayne Xin Zhao, Minghui Qiu
Abstract: Recent advances in inference-time scaling, particularly those leveraging reinforcement learning with verifiable rewards, have substantially enhanced the reasoning capabilities of Large Vision-Language Models (LVLMs). Inspired by this success, similar strategies have been applied to multimodal reasoning, yet their impact on visual perception remains unclear. To investigate this gap, we introduce DisTANCE, a perception-centric benchmark for visual estimation tasks. Evaluation results show that LVLMs exhibit limited estimation precision, and inference-time scaling offers only marginal gains. We attribute this to the fast perception paradigm of current LVLMs, where visual understanding is treated as a one-shot output without modeling the underlying perceptual process. To address this, we propose Perception-Time Scaling (PTS), a novel paradigm that encourages token-rich perception and decomposes complex perception problems into intermediate tractable sub-problems, thereby enabling perception to align with and benefit from inference-time scaling. Combined with reinforcement learning techniques, PTS significantly improves perception accuracy, raising high-precision performance on DisTANCE from 8.0% to 64.7%, and generalizes well to out-of-domain tasks. Surprisingly, even though PTS data are purely synthetic, combining them with math reasoning data yields consistent gains in both reasoning and real-world perception benchmarks. Further analysis reveals that PTS introduces more perception-related tokens and increases the model's attention to image tokens. Our code and data will be publicly released.
Authors: Zhenyu Wang, Mahathir Monjur, Shahriar Nirjon
Abstract: In mmWave-based pose estimation, sparse signals and weak reflections often cause models to infer body joints from statistical priors rather than sensor data. While prior knowledge helps in learning meaningful representations, over-reliance on it degrades performance in downstream tasks like gesture and activity recognition. In this paper, we introduce mmJoints, a framework that augments a pre-trained, black-box mmWave-based 3D pose estimator's output with additional joint descriptors. Rather than mitigating bias, mmJoints makes it explicit by estimating the likelihood of a joint being sensed and the reliability of its predicted location. These descriptors enhance interpretability and improve downstream task accuracy. Through extensive evaluations using over 115,000 signal frames across 13 pose estimation settings, we show that mmJoints estimates descriptors with an error rate below 4.2%. mmJoints also improves joint position accuracy by up to 12.5% and boosts activity recognition by up to 16% over state-of-the-art methods.
Authors: Maoliang Li, Ke Li, Yaoyang Liu, Jiayu Chen, Zihao Zheng, Yinjun Wu, Xiang Chen
Abstract: To effectively leverage user-specific data, retrieval augmented generation (RAG) is employed in multimodal large language model (MLLM) applications. However, conventional retrieval approaches often suffer from limited retrieval accuracy. Recent advances in multi-vector retrieval (MVR) improve accuracy by decomposing queries and matching against segmented images. They still suffer from sub-optimal accuracy and efficiency, overlooking alignment between the query and varying image objects and redundant fine-grained image segments. In this work, we present an efficient scheduling framework for image retrieval - HiMIR. First, we introduce a novel hierarchical paradigm, employing multiple intermediate granularities for varying image objects to enhance alignment. Second, we minimize redundancy in retrieval by leveraging cross-hierarchy similarity consistency and hierarchy sparsity to minimize unnecessary matching computation. Furthermore, we configure parameters for each dataset automatically for practicality across diverse scenarios. Our empirical study shows that, HiMIR not only achieves substantial accuracy improvements but also reduces computation by up to 3.5 times over the existing MVR system.
Authors: Zichuan Wang, Bo Peng, Songlin Yang, Zhenchen Tang, Jing Dong
Abstract: Although recent text-to-image (T2I) models have significantly improved the overall visual quality of generated images, they still struggle in the generation of accurate details in complex local regions, especially human hands. Generated hands often exhibit structural distortions and unrealistic textures, which can be very noticeable even when the rest of the body is well-generated. However, the quality assessment of hand regions remains largely neglected, limiting downstream task performance like human-centric generation quality optimization and AIGC detection. To address this, we propose the first quality assessment task targeting generated hand regions and showcase its abundant downstream applications. We first introduce the HandPair dataset for training hand quality assessment models. It consists of 48k images formed by high- and low-quality hand pairs, enabling low-cost, efficient supervision without manual annotation. Based on it, we develop HandEval, a carefully designed hand-specific quality assessment model. It leverages the powerful visual understanding capability of Multimodal Large Language Model (MLLM) and incorporates prior knowledge of hand keypoints, gaining strong perception of hand quality. We further construct a human-annotated test set with hand images from various state-of-the-art (SOTA) T2I models to validate its quality evaluation capability. Results show that HandEval aligns better with human judgments than existing SOTA methods. Furthermore, we integrate HandEval into image generation and AIGC detection pipelines, prominently enhancing generated hand realism and detection accuracy, respectively, confirming its universal effectiveness in downstream applications. Code and dataset will be available.
Authors: Yuki Nii, Futa Waseda, Ching-Chun Chang, Isao Echizen
Abstract: AI-based colorization has shown remarkable capability in generating realistic color images from grayscale inputs. However, it poses risks of copyright infringement -- for example, the unauthorized colorization and resale of monochrome manga and films. Despite these concerns, no effective method currently exists to prevent such misuse. To address this, we introduce the first defensive paradigm, Uncolorable Examples, which embed imperceptible perturbations into grayscale images to invalidate unauthorized colorization. To ensure real-world applicability, we establish four criteria: effectiveness, imperceptibility, transferability, and robustness. Our method, Perception-Aware Chroma-Restrictive Perturbation (PAChroma), generates Uncolorable Examples that meet these four criteria by optimizing imperceptible perturbations with a Laplacian filter to preserve perceptual quality, and applying diverse input transformations during optimization to enhance transferability across models and robustness against common post-processing (e.g., compression). Experiments on ImageNet and Danbooru datasets demonstrate that PAChroma effectively degrades colorization quality while maintaining the visual appearance. This work marks the first step toward protecting visual content from illegitimate AI colorization, paving the way for copyright-aware defenses in generative media.
Authors: Yao Teng, Fuyun Wang, Xian Liu, Zhekai Chen, Han Shi, Yu Wang, Zhenguo Li, Weiyang Liu, Difan Zou, Xihui Liu
Abstract: As a new paradigm of visual content generation, autoregressive text-to-image models suffer from slow inference due to their sequential token-by-token decoding process, often requiring thousands of model forward passes to generate a single image. To address this inefficiency, we propose Speculative Jacobi-Denoising Decoding (SJD2), a framework that incorporates the denoising process into Jacobi iterations to enable parallel token generation in autoregressive models. Our method introduces a next-clean-token prediction paradigm that enables the pre-trained autoregressive models to accept noise-perturbed token embeddings and predict the next clean tokens through low-cost fine-tuning. This denoising paradigm guides the model towards more stable Jacobi trajectories. During inference, our method initializes token sequences with Gaussian noise and performs iterative next-clean-token-prediction in the embedding space. We employ a probabilistic criterion to verify and accept multiple tokens in parallel, and refine the unaccepted tokens for the next iteration with the denoising trajectory. Experiments show that our method can accelerate generation by reducing model forward passes while maintaining the visual quality of generated images.
Authors: Hoigi Seo, Dong Un Kang, Hyunjin Cho, Joohoon Lee, Se Young Chun
Abstract: Large vision-language models (LVLMs), which integrate a vision encoder (VE) with a large language model, have achieved remarkable success across various tasks. However, there are still crucial challenges in LVLMs such as object hallucination, generating descriptions of objects that are not in the input image. Here, we argue that uncertain visual tokens within the VE is a key factor that contributes to object hallucination. Our statistical analysis found that there are positive correlations between visual tokens with high epistemic uncertainty and the occurrence of hallucinations. Furthermore, we show theoretically and empirically that visual tokens in early VE layers that exhibit large representation deviations under small adversarial perturbations indicate high epistemic uncertainty. Based on these findings, we propose a simple yet effective strategy to mitigate object hallucination by modifying the VE only. Our method comprises a proxy method with adversarial perturbations for identifying uncertain visual tokens efficiently and a method to mask these uncertain visual tokens during the self-attention process in the middle layers of the VE, suppressing their influence on visual encoding and thus alleviating hallucinations. Extensive experiments show that our method significantly reduces object hallucinations in LVLMs and can synergistically work with other prior arts.
Authors: Xiaoxiao Ma, Feng Zhao, Pengyang Ling, Haibo Qiu, Zhixiang Wei, Hu Yu, Jie Huang, Zhixiong Zeng, Lin Ma
Abstract: In this work, we first revisit the sampling issues in current autoregressive (AR) image generation models and identify that image tokens, unlike text tokens, exhibit lower information density and non-uniform spatial distribution. Accordingly, we present an entropy-informed decoding strategy that facilitates higher autoregressive generation quality with faster synthesis speed. Specifically, the proposed method introduces two main innovations: 1) dynamic temperature control guided by spatial entropy of token distributions, enhancing the balance between content diversity, alignment accuracy, and structural coherence in both mask-based and scale-wise models, without extra computational overhead, and 2) entropy-aware acceptance rules in speculative decoding, achieving near-lossless generation at about 85\% of the inference cost of conventional acceleration methods. Extensive experiments across multiple benchmarks using diverse AR image generation models demonstrate the effectiveness and generalizability of our approach in enhancing both generation quality and sampling speed.
Authors: Weitong Kong, Zichao Zeng, Di Wen, Jiale Wei, Kunyu Peng, June Moh Goo, Jan Boehm, Rainer Stiefelhagen
Abstract: Accurate perception is critical for vehicle safety, with LiDAR as a key enabler in autonomous driving. To ensure robust performance across environments, sensor types, and weather conditions without costly re-annotation, domain generalization in LiDAR-based 3D semantic segmentation is essential. However, LiDAR annotations are often noisy due to sensor imperfections, occlusions, and human errors. Such noise degrades segmentation accuracy and is further amplified under domain shifts, threatening system reliability. While noisy-label learning is well-studied in images, its extension to 3D LiDAR segmentation under domain generalization remains largely unexplored, as the sparse and irregular structure of point clouds limits direct use of 2D methods. To address this gap, we introduce the novel task Domain Generalization for LiDAR Semantic Segmentation under Noisy Labels (DGLSS-NL) and establish the first benchmark by adapting three representative noisy-label learning strategies from image classification to 3D segmentation. However, we find that existing noisy-label learning approaches adapt poorly to LiDAR data. We therefore propose DuNe, a dual-view framework with strong and weak branches that enforce feature-level consistency and apply cross-entropy loss based on confidence-aware filtering of predictions. Our approach shows state-of-the-art performance by achieving 56.86% mIoU on SemanticKITTI, 42.28% on nuScenes, and 52.58% on SemanticPOSS under 10% symmetric label noise, with an overall Arithmetic Mean (AM) of 49.57% and Harmonic Mean (HM) of 48.50%, thereby demonstrating robust domain generalization in DGLSS-NL tasks. The code is available on our project page.
Authors: Junhyeok Lee, Hyunwoong Kim, Hyungjin Chung, Heeseong Eom, Joon Jang, Chul-Ho Sohn, Kyu Sung Choi
Abstract: Image-to-Image translation models can help mitigate various challenges inherent to medical image acquisition. Latent diffusion models (LDMs) leverage efficient learning in compressed latent space and constitute the core of state-of-the-art generative image models. However, this efficiency comes with a trade-off, potentially compromising crucial pixel-level detail essential for high-fidelity medical images. This limitation becomes particularly critical when generating clinically significant structures, such as lesions, which often occupy only a small portion of the image. Failure to accurately reconstruct these regions can severely impact diagnostic reliability and clinical decision-making. To overcome this limitation, we propose a novel post-training framework for LDMs in medical image-to-image translation by incorporating lesion-aware medical pixel space objectives. This approach is essential, as it not only enhances overall image quality but also improves the precision of lesion delineation. We evaluate our framework on brain CT-to-MRI translation in acute ischemic stroke patients, where early and accurate diagnosis is critical for optimal treatment selection and improved patient outcomes. While diffusion MRI is the gold standard for stroke diagnosis, its clinical utility is often constrained by high costs and low accessibility. Using a dataset of 817 patients, we demonstrate that our framework improves overall image quality and enhances lesion delineation when synthesizing DWI and ADC images from CT perfusion scans, outperforming existing image-to-image translation models. Furthermore, our post-training strategy is easily adaptable to pre-trained LDMs and exhibits substantial potential for broader applications across diverse medical image translation tasks.
Authors: Yitong Chen, Xinyao Xu, Ping Zhu, Xinyong Han, Fangbo Qin, Shan Yu
Abstract: Flexible microelectrode (FME) implantation into brain cortex is challenging due to the deformable fiber-like structure of FME probe and the interaction with critical bio-tissue. To ensure reliability and safety, the implantation process should be monitored carefully. This paper develops an image-based anomaly detection framework based on the microscopic cameras of the robotic FME implantation system. The unified framework is utilized at four checkpoints to check the micro-needle, FME probe, hooking result, and implantation point, respectively. Exploiting the existing object localization results, the aligned regions of interest (ROIs) are extracted from raw image and input to a pretrained vision transformer (ViT). Considering the task specifications, we propose a progressive granularity patch feature sampling method to address the sensitivity-tolerance trade-off issue at different locations. Moreover, we select a part of feature channels with higher signal-to-noise ratios from the raw general ViT features, to provide better descriptors for each specific scene. The effectiveness of the proposed methods is validated with the image datasets collected from our implantation system.
Authors: Weijia Wang, Yuanzhi Su, Pei-Gen Ye, Yuan-Gen Wang, Xuequan Lu
Abstract: We present MambaH-Fit, a state space modelling framework tailored for hyper-surface fitting-based point cloud normal estimation. Existing normal estimation methods often fall short in modelling fine-grained geometric structures, thereby limiting the accuracy of the predicted normals. Recently, state space models (SSMs), particularly Mamba, have demonstrated strong modelling capability by capturing long-range dependencies with linear complexity and inspired adaptations to point cloud processing. However, existing Mamba-based approaches primarily focus on understanding global shape structures, leaving the modelling of local, fine-grained geometric details largely under-explored. To address the issues above, we first introduce an Attention-driven Hierarchical Feature Fusion (AHFF) scheme to adaptively fuse multi-scale point cloud patch features, significantly enhancing geometric context learning in local point cloud neighbourhoods. Building upon this, we further propose Patch-wise State Space Model (PSSM) that models point cloud patches as implicit hyper-surfaces via state dynamics, enabling effective fine-grained geometric understanding for normal prediction. Extensive experiments on benchmark datasets show that our method outperforms existing ones in terms of accuracy, robustness, and flexibility. Ablation studies further validate the contribution of the proposed components.
Authors: Juanqin Liu, Leonardo Plotegher, Eloy Roura, Shaoming He
Abstract: The extensive application of unmanned aerial vehicles (UAVs) in military reconnaissance, environmental monitoring, and related domains has created an urgent need for accurate and efficient multi-object tracking (MOT) technologies, which are also essential for UAV situational awareness. However, complex backgrounds, small-scale targets, and frequent occlusions and interactions continue to challenge existing methods in terms of detection accuracy and trajectory continuity. To address these issues, this paper proposes the Global-Local Detection and Tracking (GL-DT) framework. It employs a Spatio-Temporal Feature Fusion (STFF) module to jointly model motion and appearance features, combined with a global-local collaborative detection strategy, effectively enhancing small-target detection. Building upon this, the JPTrack tracking algorithm is introduced to mitigate common issues such as ID switches and trajectory fragmentation. Experimental results demonstrate that the proposed approach significantly improves the continuity and stability of MOT while maintaining real-time performance, providing strong support for the advancement of UAV detection and tracking technologies.
Authors: Youwei Zheng, Yuxi Ren, Xin Xia, Xuefeng Xiao, Xiaohua Xie
Abstract: Diffusion Transformer (DiT) has demonstrated remarkable performance in text-to-image generation; however, its large parameter size results in substantial inference overhead. Existing parameter compression methods primarily focus on pruning, but aggressive pruning often leads to severe performance degradation due to reduced model capacity. To address this limitation, we pioneer the transformation of a dense DiT into a Mixture of Experts (MoE) for structured sparsification, reducing the number of activated parameters while preserving model capacity. Specifically, we replace the Feed-Forward Networks (FFNs) in DiT Blocks with MoE layers, reducing the number of activated parameters in the FFNs by 62.5\%. Furthermore, we propose the Mixture of Blocks (MoB) to selectively activate DiT blocks, thereby further enhancing sparsity. To ensure an effective dense-to-MoE conversion, we design a multi-step distillation pipeline, incorporating Taylor metric-based expert initialization, knowledge distillation with load balancing, and group feature loss for MoB optimization. We transform large diffusion transformers (e.g., FLUX.1 [dev]) into an MoE structure, reducing activated parameters by 60\% while maintaining original performance and surpassing pruning-based approaches in extensive experiments. Overall, Dense2MoE establishes a new paradigm for efficient text-to-image generation.
Authors: Irash Perera (Department of Computer Science and Engineering, University of Moratuwa, Colombo, Sri Lanka), Uthayasanker Thayasivam (Department of Computer Science and Engineering, University of Moratuwa, Colombo, Sri Lanka)
Abstract: Intelligent analysis of medical imaging plays a crucial role in assisting clinical diagnosis, especially for identifying subtle pathological features. This paper introduces a novel multi-branch ConvNeXt architecture designed specifically for the nuanced challenges of medical image analysis. While applied here to the specific problem of COVID-19 diagnosis, the methodology offers a generalizable framework for classifying a wide range of pathologies from CT scans. The proposed model incorporates a rigorous end-to-end pipeline, from meticulous data preprocessing and augmentation to a disciplined two-phase training strategy that leverages transfer learning effectively. The architecture uniquely integrates features extracted from three parallel branches: Global Average Pooling, Global Max Pooling, and a new Attention-weighted Pooling mechanism. The model was trained and validated on a combined dataset of 2,609 CT slices derived from two distinct datasets. Experimental results demonstrate a superior performance on the validation set, achieving a final ROC-AUC of 0.9937, a validation accuracy of 0.9757, and an F1-score of 0.9825 for COVID-19 cases, outperforming all previously reported models on this dataset. These findings indicate that a modern, multi-branch architecture, coupled with careful data handling, can achieve performance comparable to or exceeding contemporary state-of-the-art models, thereby proving the efficacy of advanced deep learning techniques for robust medical diagnostics.
Authors: Weikai Huang, Jieyu Zhang, Taoyang Jia, Chenhao Zheng, Ziqi Gao, Jae Sung Park, Ranjay Krishna
Abstract: Visual grouping -- operationalized via instance segmentation, visual grounding, and object detection -- underpins applications from robotic perception to photo editing. Large annotated datasets are costly, biased in coverage, and hard to scale. Synthetic data are promising but often lack flexibility, accuracy, and compositional diversity. We present SOS, a simple and scalable data synthesis pipeline based on an object-centric composition strategy. It pastes high-quality synthetic object segments into new images using structured layout priors and generative relighting, producing accurate and diverse masks, boxes, and referring expressions. Models trained on 100000 synthetic images from SOS outperform those trained on larger real-image datasets such as GRIT (20M) and V3Det (200K) on detection and grounding tasks, achieving +10.9 AP on LVIS detection and +8.4 $N_{\text{Acc}}$ on gRefCOCO grounding. SOS enables controllable dataset construction and improves generalization in both low-data and closed-vocabulary settings. Augmenting LVIS and COCO with synthetic object segments yields strong performance across real-data scales and even larger gains under extremely limited real data (for example, +3.83 $AP_{\text{rare}}$ on LVIS instance segmentation and +6.59 AP with a 1 percent COCO setup). This controllability also supports targeted data generation for challenging intra-class referring in visual grounding.
Authors: Dominik Winter, Mai Bui, Monica Azqueta Gavaldon, Nicolas Triltsch, Marco Rosati, Nicolas Brieu
Abstract: Scarcity of annotated data, particularly for rare or atypical morphologies, present significant challenges for cell and nuclei segmentation in computational pathology. While manual annotation is labor-intensive and costly, synthetic data offers a cost-effective alternative. We introduce a Multimodal Semantic Diffusion Model (MSDM) for generating realistic pixel-precise image-mask pairs for cell and nuclei segmentation. By conditioning the generative process with cellular/nuclear morphologies (using horizontal and vertical maps), RGB color characteristics, and BERT-encoded assay/indication metadata, MSDM generates datasests with desired morphological properties. These heterogeneous modalities are integrated via multi-head cross-attention, enabling fine-grained control over the generated images. Quantitative analysis demonstrates that synthetic images closely match real data, with low Wasserstein distances between embeddings of generated and real images under matching biological conditions. The incorporation of these synthetic samples, exemplified by columnar cells, significantly improves segmentation model accuracy on columnar cells. This strategy systematically enriches data sets, directly targeting model deficiencies. We highlight the effectiveness of multimodal diffusion-based augmentation for advancing the robustness and generalizability of cell and nuclei segmentation models. Thereby, we pave the way for broader application of generative models in computational pathology.
Authors: Satya P. Singh, Rashmi Chaudhry, Anand Srivastava, Jagath C. Rajapakse
Abstract: Orthogonal moment-based image representations are fundamental in computer vision, but classical methods suffer from high computational complexity and numerical instability at large orders. Zernike and pseudo-Zernike moments, for instance, require coupled radial-angular processing that precludes efficient factorization, resulting in $\mathcal{O}(n^3N^2)$ to $\mathcal{O}(n^6N^2)$ complexity and $\mathcal{O}(N^4)$ condition number scaling for the $n$th-order moments on an $N\times N$ image. We introduce \textbf{PSepT} (Polar Separable Transform), a separable orthogonal transform that overcomes the non-separability barrier in polar coordinates. PSepT achieves complete kernel factorization via tensor-product construction of Discrete Cosine Transform (DCT) radial bases and Fourier harmonic angular bases, enabling independent radial and angular processing. This separable design reduces computational complexity to $\mathcal{O}(N^2 \log N)$, memory requirements to $\mathcal{O}(N^2)$, and condition number scaling to $\mathcal{O}(\sqrt{N})$, representing exponential improvements over polynomial approaches. PSepT exhibits orthogonality, completeness, energy conservation, and rotation-covariance properties. Experimental results demonstrate better numerical stability, computational efficiency, and competitive classification performance on structured datasets, while preserving exact reconstruction. The separable framework enables high-order moment analysis previously infeasible with classical methods, opening new possibilities for robust image analysis applications.
Authors: Aziz Bacha, Thomas George
Abstract: Deep neural networks are often considered opaque systems, prompting the need for explainability methods to improve trust and accountability. Existing approaches typically attribute test-time predictions either to input features (e.g., pixels in an image) or to influential training examples. We argue that both perspectives should be studied jointly. This work explores *training feature attribution*, which links test predictions to specific regions of specific training images and thereby provides new insights into the inner workings of deep models. Our experiments on vision datasets show that training feature attribution yields fine-grained, test-specific explanations: it identifies harmful examples that drive misclassifications and reveals spurious correlations, such as patch-based shortcuts, that conventional attribution methods fail to expose.
Authors: Clara Tomasini, Luis Riazuelo, Ana C. Murillo
Abstract: Video bronchoscopy is a fundamental procedure in respiratory medicine, where medical experts navigate through the bronchial tree of a patient to diagnose or operate the patient. Surgeons need to determine the position of the scope as they go through the airway until they reach the area of interest. This task is very challenging for practitioners due to the complex bronchial tree structure and varying doctor experience and training. Navigation assistance to locate the bronchoscope during the procedure can improve its outcome. Currently used techniques for navigational guidance commonly rely on previous CT scans of the patient to obtain a 3D model of the airway, followed by tracking of the scope with additional sensors or image registration. These methods obtain accurate locations but imply additional setup, scans and training. Accurate metric localization is not always required, and a topological localization with regard to a generic airway model can often suffice to assist the surgeon with navigation. We present an image-based bronchoscopy topological localization pipeline to provide navigation assistance during the procedure, with no need of patient CT scan. Our approach is trained only on phantom data, eliminating the high cost of real data labeling, and presents good generalization capabilities. The results obtained surpass existing methods, particularly on real data test sequences.
Authors: Yankun Wu, Zakaria Laskar, Giorgos Kordopatis-Zilos, Noa Garcia, Giorgos Tolias
Abstract: Instance-level recognition (ILR) focuses on identifying individual objects rather than broad categories, offering the highest granularity in image classification. However, this fine-grained nature makes creating large-scale annotated datasets challenging, limiting ILR's real-world applicability across domains. To overcome this, we introduce a novel approach that synthetically generates diverse object instances from multiple domains under varied conditions and backgrounds, forming a large-scale training set. Unlike prior work on automatic data synthesis, our method is the first to address ILR-specific challenges without relying on any real images. Fine-tuning foundation vision models on the generated data significantly improves retrieval performance across seven ILR benchmarks spanning multiple domains. Our approach offers a new, efficient, and effective alternative to extensive data collection and curation, introducing a new ILR paradigm where the only input is the names of the target domains, unlocking a wide range of real-world applications.
Authors: Yuchen Zhang, Yao Lu, Johannes Betz
Abstract: Modern object detectors are largely confined to a "closed-world" assumption, limiting them to a predefined set of classes and posing risks when encountering novel objects in real-world scenarios. While open-set detection methods aim to address this by identifying such instances as 'Unknown', this is often insufficient. Rather than treating all unknowns as a single class, assigning them more descriptive subcategories can enhance decision-making in safety-critical contexts. For example, identifying an object as an 'Unknown Animal' (requiring an urgent stop) versus 'Unknown Debris' (requiring a safe lane change) is far more useful than just 'Unknown' in autonomous driving. To bridge this gap, we introduce TARO, a novel detection framework that not only identifies unknown objects but also classifies them into coarse parent categories within a semantic hierarchy. TARO employs a unique architecture with a sparsemax-based head for modeling objectness, a hierarchy-guided relabeling component that provides auxiliary supervision, and a classification module that learns hierarchical relationships. Experiments show TARO can categorize up to 29.9% of unknowns into meaningful coarse classes, significantly reduce confusion between unknown and known classes, and achieve competitive performance in both unknown recall and known mAP. Code will be made available.
Authors: Johann-Friedrich Feiden, Tim K\"uchler, Denis Zavadski, Bogdan Savchynskyy, Carsten Rother
Abstract: Depth estimation from monocular video has become a key component of many real-world computer vision systems. Recently, Video Depth Anything (VDA) has demonstrated strong performance on long video sequences. However, it relies on batch-processing which prohibits its use in an online setting. In this work, we overcome this limitation and introduce online VDA (oVDA). The key innovation is to employ techniques from Large Language Models (LLMs), namely, caching latent features during inference and masking frames at training. Our oVDA method outperforms all competing online video depth estimation methods in both accuracy and VRAM usage. Low VRAM usage is particularly important for deployment on edge devices. We demonstrate that oVDA runs at 42 FPS on an NVIDIA A100 and at 20 FPS on an NVIDIA Jetson edge device. We will release both, code and compilation scripts, making oVDA easy to deploy on low-power hardware.
Authors: Sungwoo Kang
Abstract: Cricket shot classification from video sequences remains a challenging problem in sports video analysis, requiring effective modeling of both spatial and temporal features. This paper presents the first comprehensive baseline study comparing seven different deep learning approaches across four distinct research paradigms for cricket shot classification. We implement and systematically evaluate traditional CNN-LSTM architectures, attention-based models, vision transformers, transfer learning approaches, and modern EfficientNet-GRU combinations on a unified benchmark. A critical finding of our study is the significant performance gap between claims in academic literature and practical implementation results. While previous papers reported accuracies of 96\% (Balaji LRCN), 99.2\% (IJERCSE), and 93\% (Sensors), our standardized re-implementations achieve 46.0\%, 55.6\%, and 57.7\% respectively. Our modern SOTA approach, combining EfficientNet-B0 with a GRU-based temporal model, achieves 92.25\% accuracy, demonstrating that substantial improvements are possible with modern architectures and systematic optimization. All implementations follow modern MLOps practices with PyTorch Lightning, providing a reproducible research platform that exposes the critical importance of standardized evaluation protocols in sports video analysis research.
Authors: Mukilan Karuppasamy, Shankar Gangisetty, Shyam Nandan Rai, Carlo Masone, C V Jawahar
Abstract: Autonomous driving (AD) systems are becoming increasingly capable of handling complex tasks, mainly due to recent advances in deep learning and AI. As interactions between autonomous systems and humans increase, the interpretability of decision-making processes in driving systems becomes increasingly crucial for ensuring safe driving operations. Successful human-machine interaction requires understanding the underlying representations of the environment and the driving task, which remains a significant challenge in deep learning-based systems. To address this, we introduce the task of interpretability in maneuver prediction before they occur for driver safety, i.e., driver intent prediction (DIP), which plays a critical role in AD systems. To foster research in interpretable DIP, we curate the eXplainable Driving Action Anticipation Dataset (DAAD-X), a new multimodal, ego-centric video dataset to provide hierarchical, high-level textual explanations as causal reasoning for the driver's decisions. These explanations are derived from both the driver's eye-gaze and the ego-vehicle's perspective. Next, we propose Video Concept Bottleneck Model (VCBM), a framework that generates spatio-temporally coherent explanations inherently, without relying on post-hoc techniques. Finally, through extensive evaluations of the proposed VCBM on the DAAD-X dataset, we demonstrate that transformer-based models exhibit greater interpretability than conventional CNN-based models. Additionally, we introduce a multilabel t-SNE visualization technique to illustrate the disentanglement and causal correlation among multiple explanations. Our data, code and models are available at: https://mukil07.github.io/VCBM.github.io/
Authors: Huimin Liu, Jing Gao, Daria Baran, AxelX Montout, Neill W Campbell, Andrew W Dowsey
Abstract: Cattle behaviour is a crucial indicator of an individual animal health, productivity and overall well-being. Video-based monitoring, combined with deep learning techniques, has become a mainstream approach in animal biometrics, and it can offer high accuracy in some behaviour recognition tasks. We present Cattle-CLIP, a multimodal deep learning framework for cattle behaviour recognition, using semantic cues to improve the performance of video-based visual feature recognition. It is adapted from the large-scale image-language model CLIP by adding a temporal integration module. To address the domain gap between web data used for the pre-trained model and real-world cattle surveillance footage, we introduce tailored data augmentation strategies and specialised text prompts. Cattle-CLIP is evaluated under both fully-supervised and few-shot learning scenarios, with a particular focus on data-scarce behaviour recognition - an important yet under-explored goal in livestock monitoring. To evaluate the proposed method, we release the CattleBehaviours6 dataset, which comprises six types of indoor behaviours: feeding, drinking, standing-self-grooming, standing-ruminating, lying-self-grooming and lying-ruminating. The dataset consists of 1905 clips collected from our John Oldacre Centre dairy farm research platform housing 200 Holstein-Friesian cows. Experiments show that Cattle-CLIP achieves 96.1% overall accuracy across six behaviours in a supervised setting, with nearly 100% recall for feeding, drinking and standing-ruminating behaviours, and demonstrates robust generalisation with limited data in few-shot scenarios, highlighting the potential of multimodal learning in agricultural and animal behaviour analysis.
Authors: Yue Li, Shida Sun, Yu Hong, Feihu Xu, Zhiwei Xiong
Abstract: Transient measurements, captured by the timeresolved systems, are widely employed in photon-efficient reconstruction tasks, including line-of-sight (LOS) and non-line-of-sight (NLOS) imaging. However, challenges persist in their 3D reconstruction due to the low quantum efficiency of sensors and the high noise levels, particularly for long-range or complex scenes. To boost the 3D reconstruction performance in photon-efficient imaging, we propose a generic Time-Resolved Transformer (TRT) architecture. Different from existing transformers designed for high-dimensional data, TRT has two elaborate attention designs tailored for the spatio-temporal transient measurements. Specifically, the spatio-temporal self-attention encoders explore both local and global correlations within transient data by splitting or downsampling input features into different scales. Then, the spatio-temporal cross attention decoders integrate the local and global features in the token space, resulting in deep features with high representation capabilities. Building on TRT, we develop two task-specific embodiments: TRT-LOS for LOS imaging and TRT-NLOS for NLOS imaging. Extensive experiments demonstrate that both embodiments significantly outperform existing methods on synthetic data and real-world data captured by different imaging systems. In addition, we contribute a large-scale, high-resolution synthetic LOS dataset with various noise levels and capture a set of real-world NLOS measurements using a custom-built imaging system, enhancing the data diversity in this field. Code and datasets are available at https://github.com/Depth2World/TRT.
Authors: Wuyang Li, Wentao Pan, Po-Chien Luan, Yang Gao, Alexandre Alahi
Abstract: We propose Stable Video Infinity (SVI) that is able to generate infinite-length videos with high temporal consistency, plausible scene transitions, and controllable streaming storylines. While existing long-video methods attempt to mitigate accumulated errors via handcrafted anti-drifting (e.g., modified noise scheduler, frame anchoring), they remain limited to single-prompt extrapolation, producing homogeneous scenes with repetitive motions. We identify that the fundamental challenge extends beyond error accumulation to a critical discrepancy between the training assumption (seeing clean data) and the test-time autoregressive reality (conditioning on self-generated, error-prone outputs). To bridge this hypothesis gap, SVI incorporates Error-Recycling Fine-Tuning, a new type of efficient training that recycles the Diffusion Transformer (DiT)'s self-generated errors into supervisory prompts, thereby encouraging DiT to actively identify and correct its own errors. This is achieved by injecting, collecting, and banking errors through closed-loop recycling, autoregressively learning from error-injected feedback. Specifically, we (i) inject historical errors made by DiT to intervene on clean inputs, simulating error-accumulated trajectories in flow matching; (ii) efficiently approximate predictions with one-step bidirectional integration and calculate errors with residuals; (iii) dynamically bank errors into replay memory across discretized timesteps, which are resampled for new input. SVI is able to scale videos from seconds to infinite durations with no additional inference cost, while remaining compatible with diverse conditions (e.g., audio, skeleton, and text streams). We evaluate SVI on three benchmarks, including consistent, creative, and conditional settings, thoroughly verifying its versatility and state-of-the-art role.
Authors: Wangyu Wu, Xuhang Chen, Zhenhong Chen, Jing-En Jiang, Kim-Fung Tsang, Xiaowei Huang, Fei Ma, Jimin Xiao
Abstract: Cross-Domain Sequential Recommendation (CDSR) plays a crucial role in modern consumer electronics and e-commerce platforms, where users interact with diverse services such as books, movies, and online retail products. These systems must accurately capture both domain-specific and cross-domain behavioral patterns to provide personalized and seamless consumer experiences. To address this challenge, we propose \textbf{TEMA-LLM} (\textit{Tag-Enriched Multi-Attention with Large Language Models}), a practical and effective framework that integrates \textit{Large Language Models (LLMs)} for semantic tag generation and enrichment. Specifically, TEMA-LLM employs LLMs to assign domain-aware prompts and generate descriptive tags from item titles and descriptions. The resulting tag embeddings are fused with item identifiers as well as textual and visual features to construct enhanced item representations. A \textit{Tag-Enriched Multi-Attention} mechanism is then introduced to jointly model user preferences within and across domains, enabling the system to capture complex and evolving consumer interests. Extensive experiments on four large-scale e-commerce datasets demonstrate that TEMA-LLM consistently outperforms state-of-the-art baselines, underscoring the benefits of LLM-based semantic tagging and multi-attention integration for consumer-facing recommendation systems. The proposed approach highlights the potential of LLMs to advance intelligent, user-centric services in the field of consumer electronics.
Authors: Vijay M. Galshetwar, Praful Hambarde, Prashant W. Patil, Akshay Dudhane, Sachin Chaudhary, Santosh Kumar Vipparathi, Subrahmanyam Murala
Abstract: Adverse weather conditions such as haze, rain, and snow significantly degrade the quality of images and videos, posing serious challenges to intelligent transportation systems (ITS) that rely on visual input. These degradations affect critical applications including autonomous driving, traffic monitoring, and surveillance. This survey presents a comprehensive review of image and video restoration techniques developed to mitigate weather-induced visual impairments. We categorize existing approaches into traditional prior-based methods and modern data-driven models, including CNNs, transformers, diffusion models, and emerging vision-language models (VLMs). Restoration strategies are further classified based on their scope: single-task models, multi-task/multi-weather systems, and all-in-one frameworks capable of handling diverse degradations. In addition, we discuss day and night time restoration challenges, benchmark datasets, and evaluation protocols. The survey concludes with an in-depth discussion on limitations in current research and outlines future directions such as mixed/compound-degradation restoration, real-time deployment, and agentic AI frameworks. This work aims to serve as a valuable reference for advancing weather-resilient vision systems in smart transportation environments. Lastly, to stay current with rapid advancements in this field, we will maintain regular updates of the latest relevant papers and their open-source implementations at https://github.com/ChaudharyUPES/A-comprehensive-review-on-Multi-weather-restoration
URLs: https://github.com/ChaudharyUPES/A-comprehensive-review-on-Multi-weather-restoration
Authors: Jindong Hong, Wencheng Zhang, Shiqin Qiao, Jianhai Chen, Jianing Qiu, Chuanyang Zheng, Qian Xu, Yun Ji, Qianyue Wen, Weiwei Sun, Hao Li, Huizhen Li, Huichao Wang, Kai Wu, Meng Li, Yijun He, Lingjie Luo, Jiankai Sun
Abstract: Shoulder disorders, such as frozen shoulder (a.k.a., adhesive capsulitis), are common conditions affecting the health of people worldwide, and have a high incidence rate among the elderly and workers engaged in repetitive shoulder tasks. In regions with scarce medical resources, achieving early and accurate diagnosis poses significant challenges, and there is an urgent need for low-cost and easily scalable auxiliary diagnostic solutions. This research introduces videos captured by consumer-grade devices as the basis for diagnosis, reducing the cost for users. We focus on the innovative application of Multimodal Large Language Models (MLLMs) in the preliminary diagnosis of shoulder disorders and propose a Hybrid Motion Video Diagnosis framework (HMVDx). This framework divides the two tasks of action understanding and disease diagnosis, which are respectively completed by two MLLMs. In addition to traditional evaluation indicators, this work proposes a novel metric called Usability Index by the logical process of medical decision-making (action recognition, movement diagnosis, and final diagnosis). This index evaluates the effectiveness of MLLMs in the medical field from the perspective of the entire medical diagnostic pathway, revealing the potential value of low-cost MLLMs in medical applications for medical practitioners. In experimental comparisons, the accuracy of HMVDx in diagnosing shoulder joint injuries has increased by 79.6\% compared with direct video diagnosis, a significant technical contribution to future research on the application of MLLMs for video understanding in the medical field.
Authors: Alina Elena Baia, Alessio Xompero, Andrea Cavallaro
Abstract: While specialized learning-based models have historically dominated image privacy prediction, the current literature increasingly favours adopting large Vision-Language Models (VLMs) designed for generic tasks. This trend risks overlooking the performance ceiling set by purpose-built models due to a lack of systematic evaluation. To address this problem, we establish a zero-shot benchmark for image privacy classification, enabling a fair comparison. We evaluate the top-3 open-source VLMs, according to a privacy benchmark, using task-aligned prompts and we contrast their performance, efficiency, and robustness against established vision-only and multi-modal methods. Counter-intuitively, our results show that VLMs, despite their resource-intensive nature in terms of high parameter count and slower inference, currently lag behind specialized, smaller models in privacy prediction accuracy. We also find that VLMs exhibit higher robustness to image perturbations.
Authors: Patrick Wienholt, Sophie Caselitz, Robert Siepmann, Philipp Bruners, Keno Bressem, Christiane Kuhl, Jakob Nikolas Kather, Sven Nebelung, Daniel Truhn
Abstract: To determine whether using discrete semantic entropy (DSE) to reject questions likely to generate hallucinations can improve the accuracy of black-box vision-language models (VLMs) in radiologic image based visual question answering (VQA). This retrospective study evaluated DSE using two publicly available, de-identified datasets: (i) the VQA-Med 2019 benchmark (500 images with clinical questions and short-text answers) and (ii) a diagnostic radiology dataset (206 cases: 60 computed tomography scans, 60 magnetic resonance images, 60 radiographs, 26 angiograms) with corresponding ground-truth diagnoses. GPT-4o and GPT-4.1 answered each question 15 times using a temperature of 1.0. Baseline accuracy was determined using low-temperature answers (temperature 0.1). Meaning-equivalent responses were grouped using bidirectional entailment checks, and DSE was computed from the relative frequencies of the resulting semantic clusters. Accuracy was recalculated after excluding questions with DSE > 0.6 or > 0.3. p-values and 95% confidence intervals were obtained using bootstrap resampling and a Bonferroni-corrected threshold of p < .004 for statistical significance. Across 706 image-question pairs, baseline accuracy was 51.7% for GPT-4o and 54.8% for GPT-4.1. After filtering out high-entropy questions (DSE > 0.3), accuracy on the remaining questions was 76.3% (retained questions: 334/706) for GPT-4o and 63.8% (retained questions: 499/706) for GPT-4.1 (both p < .001). Accuracy gains were observed across both datasets and largely remained statistically significant after Bonferroni correction. DSE enables reliable hallucination detection in black-box VLMs by quantifying semantic inconsistency. This method significantly improves diagnostic answer accuracy and offers a filtering strategy for clinical VLM applications.
Authors: Ming Dai, Sen Yang, Boqiang Duan, Wankou Yang, Jingdong Wang
Abstract: Referring Video Object Segmentation (RefVOS) seeks to segment target objects in videos guided by natural language descriptions, demanding both temporal reasoning and fine-grained visual comprehension. Existing sampling strategies for LLM-based approaches typically rely on either handcrafted heuristics or external keyframe models. The former often overlooks essential temporal cues, while the latter increases system complexity. To address this, we propose a unified framework that jointly optimizes Temporal Sentence Grounding (TSG) and RefVOS, naturally incorporating key moment grounding capability. During training, we introduce a novel TSG paradigm that employs a dedicated \texttt{[FIND]} token for key moment identification through temporal token similarity matching, thereby avoiding the need for external timestamp encodings. For inference, we design a Moment-Centric Sampling (MCS) strategy that densely samples informative moments while sparsely sampling non-essential frames, preserving both motion details and global context. To further enhance tracking stability, we develop Bidirectional Anchor-updated Propagation (BAP), which leverages the most relevant moment as start point for high-quality mask initialization and dynamically updates at sampled points to mitigate accumulated errors. Code and model will be available at: https://github.com/Dmmm1997/MomentSeg
Authors: Siyuan Huang, Xiaoye Qu, Yafu Li, Yun Luo, Zefeng He, Daizong Liu, Yu Cheng
Abstract: While Reinforcement Learning with Verifiable Rewards (RLVR) has advanced the reasoning capabilities of Large Vision-Language Models (LVLMs), most existing methods in multimodal reasoning neglect the critical role of visual perception within the RLVR optimization process. In this paper, we undertake a pioneering exploration of multimodal RLVR through the novel perspective of token perception, which measures the visual dependency of each generated token. With a granular analysis of Chain-of-Thought (CoT) processes, we uncover two key insights: first, token perception in a rollout trajectory is sparsely distributed, where only a small fraction of tokens have high visual dependency for visually-grounded reasoning; second, different trajectories exhibit significant divergence in their overall visual dependency. Based on these observations, we propose Visually-Perceptive Policy Optimization (VPPO), a novel policy gradient algorithm that explicitly leverages token perception to refine the learning signal. Specifically, VPPO achieves this through a dual mechanism: it reweights a trajectory's advantage by its overall visual dependency, and focuses policy updates exclusively on perceptually pivotal tokens. On a comprehensive suite of eight perception and reasoning benchmarks, VPPO demonstrates substantial gains over leading open-source RL-tuned models, with its effectiveness consistently validated across 7B and 32B model scales. Our findings not only establish a new token-level perceptual perspective for analyzing multimodal RLVR but also present a novel and effective optimization strategy to significantly enhance the multimodal reasoning capabilities of LVLMs.
Authors: Tejaswi V. Panchagnula
Abstract: Animals often forage via Levy walks stochastic trajectories with heavy tailed step lengths optimized for sparse resource environments. We show that human visual gaze follows similar dynamics when scanning images. While traditional models emphasize image based saliency, the underlying spatiotemporal statistics of eye movements remain underexplored. Understanding these dynamics has broad applications in attention modeling and vision-based interfaces. In this study, we conducted a large scale human subject experiment involving 40 participants viewing 50 diverse images under unconstrained conditions, recording over 4 million gaze points using a high speed eye tracker. Analysis of these data shows that the gaze trajectory of the human eye also follows a Levy walk akin to animal foraging. This suggests that the human eye forages for visual information in an optimally efficient manner. Further, we trained a convolutional neural network (CNN) to predict fixation heatmaps from image input alone. The model accurately reproduced salient fixation regions across novel images, demonstrating that key components of gaze behavior are learnable from visual structure alone. Our findings present new evidence that human visual exploration obeys statistical laws analogous to natural foraging and open avenues for modeling gaze through generative and predictive frameworks.
Authors: Yuying Li, Siyi Qian, Hao Liang, Leqi Zheng, Ruichuan An, Yongzhen Guo, Wentao Zhang
Abstract: Geometric reasoning remains a core challenge for Multimodal Large Language Models (MLLMs). Even the most advanced closed-source systems, such as GPT-O3 and Gemini-2.5-Pro, still struggle to solve geometry problems reliably, despite exhibiting strong textual reasoning abilities on tasks like the International Mathematical Olympiad (IMO). This gap suggests that the bottleneck lies in understanding geometric diagrams rather than reasoning itself. Since geometric figures can often be faithfully described in concise textual form, converting visual content into captions offers a promising direction. Motivated by this insight, we introduce CapGeo, a caption-assisted reasoning framework that bridges visual and textual modalities. Experiments show substantial improvements when models are equipped with captions: Qwen2.5-VL-72B improves from 8.6% (vision-only) to 59.0%, while Claude-Opus-4 rises from 44.8% to 73.0%. To systematically evaluate and identify high-quality geometric captioning models, we further propose CapGeo-Bench, a dataset of 4,641 curated figure-caption pairs. Crucially, CapGeo-Bench incorporates a keypoint-based evaluation metric that correlates strongly with downstream CapGeo performance, enabling reliable assessment of geometric captioning ability. Together, our framework and benchmark highlight a new pathway toward advancing geometric reasoning in MLLMs.
Authors: Haozhe Jia, Wenshuo Chen, Xiucheng Wang, Nan Cheng, Hongbo Zhang, Kuimou Yu, Songning Lai, Nanjian Jia, Bowen Tian, Hongru Xiao, Yutao Yue
Abstract: Accurate and real-time radio map (RM) generation is crucial for next-generation wireless systems, yet diffusion-based approaches often suffer from large model sizes, slow iterative denoising, and high inference latency, which hinder practical deployment. To overcome these limitations, we propose \textbf{RadioFlow}, a novel flow-matching-based generative framework that achieves high-fidelity RM generation through single-step efficient sampling. Unlike conventional diffusion models, RadioFlow learns continuous transport trajectories between noise and data, enabling both training and inference to be significantly accelerated while preserving reconstruction accuracy. Comprehensive experiments demonstrate that RadioFlow achieves state-of-the-art performance with \textbf{up to 8$\times$ fewer parameters} and \textbf{over 4$\times$ faster inference} compared to the leading diffusion-based baseline (RadioDiff). This advancement provides a promising pathway toward scalable, energy-efficient, and real-time electromagnetic digital twins for future 6G networks. We release the code at \href{https://github.com/Hxxxz0/RadioFlow}{GitHub}.
Authors: Wenyao Zhang, Hongsi Liu, Bohan Li, Jiawei He, Zekun Qi, Yunnan Wang, Shengyang Zhao, Xinqiang Yu, Wenjun Zeng, Xin Jin
Abstract: Current self-supervised monocular depth estimation (MDE) approaches encounter performance limitations due to insufficient semantic-spatial knowledge extraction. To address this challenge, we propose Hybrid-depth, a novel framework that systematically integrates foundation models (e.g., CLIP and DINO) to extract visual priors and acquire sufficient contextual information for MDE. Our approach introduces a coarse-to-fine progressive learning framework: 1) Firstly, we aggregate multi-grained features from CLIP (global semantics) and DINO (local spatial details) under contrastive language guidance. A proxy task comparing close-distant image patches is designed to enforce depth-aware feature alignment using text prompts; 2) Next, building on the coarse features, we integrate camera pose information and pixel-wise language alignment to refine depth predictions. This module seamlessly integrates with existing self-supervised MDE pipelines (e.g., Monodepth2, ManyDepth) as a plug-and-play depth encoder, enhancing continuous depth estimation. By aggregating CLIP's semantic context and DINO's spatial details through language guidance, our method effectively addresses feature granularity mismatches. Extensive experiments on the KITTI benchmark demonstrate that our method significantly outperforms SOTA methods across all metrics, which also indeed benefits downstream tasks like BEV perception. Code is available at https://github.com/Zhangwenyao1/Hybrid-depth.
Authors: Zenan Lin, Wei Li, Jintao Chen, Zihao Wu, Wenxiong Kang, Changxin Gao, Liansheng Wang, Jin-Gang Yu
Abstract: Nuclei instance segmentation in pathological images is crucial for downstream tasks such as tumor microenvironment analysis. However, the high cost and scarcity of annotated data limit the applicability of fully supervised methods, while existing semi-supervised methods fail to adequately regularize consistency at the instance level, lack leverage of the inherent prior knowledge of pathological structures, and are prone to introducing noisy pseudo-labels during training. In this paper, we propose an Instance-Aware Robust Consistency Regularization Network (IRCR-Net) for accurate instance-level nuclei segmentation. Specifically, we introduce the Matching-Driven Instance-Aware Consistency (MIAC) and Prior-Driven Instance-Aware Consistency (PIAC) mechanisms to refine the nuclei instance segmentation result of the teacher and student subnetwork, particularly for densely distributed and overlapping nuclei. We incorporate morphological prior knowledge of nuclei in pathological images and utilize these priors to assess the quality of pseudo-labels generated from unlabeled data. Low-quality pseudo-labels are discarded, while high-quality predictions are enhanced to reduce pseudo-label noise and benefit the network's robust training. Experimental results demonstrate that the proposed method significantly enhances semi-supervised nuclei instance segmentation performance across multiple public datasets compared to existing approaches, even surpassing fully supervised methods in some scenarios.
Authors: Jinyuan Liu, Zihang Chen, Zhu Liu, Zhiying Jiang, Long Ma, Xin Fan, Risheng Liu
Abstract: We engage in the relatively underexplored task named thermal infrared image enhancement. Existing infrared image enhancement methods primarily focus on tackling individual degradations, such as noise, contrast, and blurring, making it difficult to handle coupled degradations. Meanwhile, all-in-one enhancement methods, commonly applied to RGB sensors, often demonstrate limited effectiveness due to the significant differences in imaging models. In sight of this, we first revisit the imaging mechanism and introduce a Progressive Prompt Fusion Network (PPFN). Specifically, the PPFN initially establishes prompt pairs based on the thermal imaging process. For each type of degradation, we fuse the corresponding prompt pairs to modulate the model's features, providing adaptive guidance that enables the model to better address specific degradations under single or multiple conditions. In addition, a Selective Progressive Training (SPT) mechanism is introduced to gradually refine the model's handling of composite cases to align the enhancement process, which not only allows the model to remove camera noise and retain key structural details, but also enhancing the overall contrast of the thermal image. Furthermore, we introduce the most high-quality, multi-scenarios infrared benchmark covering a wide range of scenarios. Extensive experiments substantiate that our approach not only delivers promising visual results under specific degradation but also significantly improves performance on complex degradation scenes, achieving a notable 8.76\% improvement. Code is available at https://github.com/Zihang-Chen/HM-TIR.
Authors: Qihang Ma, Shengyu Li, Jie Tang, Dingkang Yang, Shaodong Chen, Yingyi Zhang, Chao Feng, Jiao Ran
Abstract: Multi-modal keyphrase prediction (MMKP) aims to advance beyond text-only methods by incorporating multiple modalities of input information to produce a set of conclusive phrases. Traditional multi-modal approaches have been proven to have significant limitations in handling the challenging absence and unseen scenarios. Additionally, we identify shortcomings in existing benchmarks that overestimate model capability due to significant overlap in training tests. In this work, we propose leveraging vision-language models (VLMs) for the MMKP task. Firstly, we use two widely-used strategies, e.g., zero-shot and supervised fine-tuning (SFT) to assess the lower bound performance of VLMs. Next, to improve the complex reasoning capabilities of VLMs, we adopt Fine-tune-CoT, which leverages high-quality CoT reasoning data generated by a teacher model to finetune smaller models. Finally, to address the "overthinking" phenomenon, we propose a dynamic CoT strategy which adaptively injects CoT data during training, allowing the model to flexibly leverage its reasoning capabilities during the inference stage. We evaluate the proposed strategies on various datasets and the experimental results demonstrate the effectiveness of the proposed approaches. The code is available at https://github.com/bytedance/DynamicCoT.
Authors: Junyan Ye, Dongzhi Jiang, Jun He, Baichuan Zhou, Zilong Huang, Zhiyuan Yan, Hongsheng Li, Conghui He, Weijia Li
Abstract: Recently, Multimodal Large Language Models (MLLMs) have made rapid progress, particularly in enhancing their reasoning capabilities. However, existing reasoning benchmarks still primarily assess language-based reasoning, often treating visual input as replaceable context. To address this gap, we introduce BLINK-Twice, a vision-centric reasoning benchmark grounded in challenging perceptual tasks. Instead of relying on external knowledge, our tasks require models to reason from visual content alone, shifting the focus from language-based to image-grounded reasoning. Compared to prior perception benchmarks, it moves beyond shallow perception ("see") and requires fine-grained observation and analytical reasoning ("observe"). BLINK-Twice integrates three core components: seven types of visual challenges for testing visual reasoning, natural adversarial image pairs that enforce reliance on visual content, and annotated reasoning chains for fine-grained evaluation of the reasoning process rather than final answers alone. We evaluate 20 leading MLLMs, including 12 foundation models and 8 reasoning-enhanced models. BLINK-Twice poses a significant challenge to current models. While existing reasoning strategies in the language space-such as chain-of-thought or self-criticism can improve performance, they often result in unstable and redundant reasoning. We observe that repeated image observation improves performance across models, and active visual interaction, as demonstrated by models like o3, highlights the need for a new paradigm for vision reasoning. The dataset is publicly available at https://github.com/PicoTrex/BLINK-Twice
Authors: Yikang Zhang, Rui Fan
Abstract: 3D Gaussian splatting (3DGS) has demonstrated impressive performance in synthesizing high-fidelity novel views. Nonetheless, its effectiveness critically depends on the quality of the initialized point cloud. Specifically, achieving uniform and complete point coverage over the underlying scene structure requires overlapping observation frustums, an assumption that is often violated in unbounded, dynamic urban environments. Training Gaussian models with partially initialized point clouds often leads to distortions and artifacts, as camera rays may fail to intersect valid surfaces, resulting in incorrect gradient propagation to Gaussian primitives associated with occluded or invisible geometry. Additionally, existing densification strategies simply clone and split Gaussian primitives from existing ones, incapable of reconstructing missing structures. To address these limitations, we propose VAD-GS, a 3DGS framework tailored for geometry recovery in challenging urban scenes. Our method identifies unreliable geometry structures via voxel-based visibility reasoning, selects informative supporting views through diversity-aware view selection, and recovers missing structures via patch matching-based multi-view stereo reconstruction. This design enables the generation of new Gaussian primitives guided by reliable geometric priors, even in regions lacking initial points. Extensive experiments on the Waymo and nuScenes datasets demonstrate that VAD-GS outperforms state-of-the-art 3DGS approaches and significantly improves the quality of reconstructed geometry for both static and dynamic objects. Source code will be released upon publication.
Authors: Jinxiang Tu, Dayong Ren, Fei Shi, Zhenhong Jia, Yahong Ren, Jiwei Qin, Fang He
Abstract: Accurate forest biomass quantification is vital for carbon cycle monitoring. While airborne LiDAR excels at capturing 3D forest structure, directly estimating woody volume and Aboveground Biomass (AGB) from point clouds is challenging due to difficulties in modeling long-range dependencies needed to distinguish trees.We propose Minkowski-MambaNet, a novel deep learning framework that directly estimates volume and AGB from raw LiDAR. Its key innovation is integrating the Mamba model's Selective State Space Model (SSM) into a Minkowski network, enabling effective encoding of global context and long-range dependencies for improved tree differentiation. Skip connections are incorporated to enhance features and accelerate convergence.Evaluated on Danish National Forest Inventory LiDAR data, Minkowski-MambaNet significantly outperforms state-of-the-art methods, providing more accurate and robust estimates. Crucially, it requires no Digital Terrain Model (DTM) and is robust to boundary artifacts. This work offers a powerful tool for large-scale forest biomass analysis, advancing LiDAR-based forest inventories.
Authors: Reza Sedghi, Anand Subramoney, David Kappel
Abstract: Efficient inference with transformer-based models remains a challenge, especially in vision tasks like object detection. We analyze the inherent sparsity in the MLP layers of DETR and introduce two methods to exploit it without retraining. First, we propose Static Indicator-Based Sparsification (SIBS), a heuristic method that predicts neuron inactivity based on fixed activation patterns. While simple, SIBS offers limited gains due to the input-dependent nature of sparsity. To address this, we introduce Micro-Gated Sparsification (MGS), a lightweight gating mechanism trained on top of a pretrained DETR. MGS predicts dynamic sparsity using a small linear layer and achieves up to 85 to 95% activation sparsity. Experiments on the COCO dataset show that MGS maintains or even improves performance while significantly reducing computation. Our method offers a practical, input-adaptive approach to sparsification, enabling efficient deployment of pretrained vision transformers without full model retraining.
Authors: Jin-Chuan Shi, Chengye Su, Jiajun Wang, Ariel Shamir, Miao Wang
Abstract: Editing 4D scenes reconstructed from monocular videos based on text prompts is a valuable yet challenging task with broad applications in content creation and virtual environments. The key difficulty lies in achieving semantically precise edits in localized regions of complex, dynamic scenes, while preserving the integrity of unedited content. To address this, we introduce Mono4DEditor, a novel framework for flexible and accurate text-driven 4D scene editing. Our method augments 3D Gaussians with quantized CLIP features to form a language-embedded dynamic representation, enabling efficient semantic querying of arbitrary spatial regions. We further propose a two-stage point-level localization strategy that first selects candidate Gaussians via CLIP similarity and then refines their spatial extent to improve accuracy. Finally, targeted edits are performed on localized regions using a diffusion-based video editing model, with flow and scribble guidance ensuring spatial fidelity and temporal coherence. Extensive experiments demonstrate that Mono4DEditor enables high-quality, text-driven edits across diverse scenes and object types, while preserving the appearance and geometry of unedited areas and surpassing prior approaches in both flexibility and visual fidelity.
Authors: Ruirui Lin, Guoxi Huang, Nantheera Anantrasirichai
Abstract: Low-light video enhancement (LLVE) is challenging due to noise, low contrast, and color degradations. Learning-based approaches offer fast inference but still struggle with heavy noise in real low-light scenes, primarily due to limitations in effectively leveraging temporal information. In this paper, we address this issue with DWTA-Net, a novel two-stage framework that jointly exploits short- and long-term temporal cues. Stage I employs Visual State-Space blocks for multi-frame alignment, recovering brightness, color, and structure with local consistency. Stage II introduces a recurrent refinement module with dynamic weight-based temporal aggregation guided by optical flow, adaptively balancing static and dynamic regions. A texture-adaptive loss further preserves fine details while promoting smoothness in flat areas. Experiments on real-world low-light videos show that DWTA-Net effectively suppresses noise and artifacts, delivering superior visual quality compared with state-of-the-art methods.
Authors: David-Alexandre Duclos, William Guimont-Martin, Gabriel Jeanson, Arthur Larochelle-Tremblay, Th\'eo Defosse, Fr\'ed\'eric Moore, Philippe Nolet, Fran\c{c}ois Pomerleau, Philippe Gigu\`ere
Abstract: Interest in robotics for forest management is growing, but perception in complex, natural environments remains a significant hurdle. Conditions such as heavy occlusion, variable lighting, and dense vegetation pose challenges to automated systems, which are essential for precision forestry, biodiversity monitoring, and the automation of forestry equipment. These tasks rely on advanced perceptual capabilities, such as detection and fine-grained species classification of individual trees. Yet, existing datasets are inadequate to develop such perception systems, as they often focus on urban settings or a limited number of species. To address this, we present SilvaScenes, a new dataset for instance segmentation of tree species from under-canopy images. Collected across five bioclimatic domains in Quebec, Canada, SilvaScenes features 1476 trees from 24 species with annotations from forestry experts. We demonstrate the relevance and challenging nature of our dataset by benchmarking modern deep learning approaches for instance segmentation. Our results show that, while tree segmentation is easy, with a top mean average precision (mAP) of 67.65%, species classification remains a significant challenge with an mAP of only 35.69%. Our dataset and source code will be available at https://github.com/norlab-ulaval/SilvaScenes.
Authors: Jisu Han, Wonjun Hwang
Abstract: Test-time adaptation paradigm provides flexibility towards domain shifts by performing immediate adaptation on unlabeled target data from the source model. Vision-Language Models (VLMs) leverage their generalization capabilities for diverse downstream tasks, and test-time prompt tuning has emerged as a prominent solution for adapting VLMs. In this work, we explore contrastive VLMs and identify the modality gap caused by a single dominant feature dimension across modalities. We observe that the dominant dimensions in both text and image modalities exhibit high predictive sensitivity, and that constraining their influence can improve calibration error. Building on this insight, we propose dimensional entropy maximization that regularizes the distribution of textual features toward uniformity to mitigate the dependency of dominant dimensions. Our method alleviates the degradation of calibration performance in test-time prompt tuning, offering a simple yet effective solution to enhance the reliability of VLMs in real-world deployment scenarios.
Authors: Ruben Pascual, Mikel Sesma-Sara, Aranzazu Jurio, Daniel Paternain, Mikel Galar
Abstract: The audiovisual industry is undergoing a profound transformation as it is integrating AI developments not only to automate routine tasks but also to inspire new forms of art. This paper addresses the problem of producing a virtually unlimited number of novel characters that preserve the artistic style and shared visual traits of a small set of human-designed reference characters, thus broadening creative possibilities in animation, gaming, and related domains. Our solution builds upon DreamBooth, a well-established fine-tuning technique for text-to-image diffusion models, and adapts it to tackle two core challenges: capturing intricate character details beyond textual prompts and the few-shot nature of the training data. To achieve this, we propose a multi-token strategy, using clustering to assign separate tokens to individual characters and their collective style, combined with LoRA-based parameter-efficient fine-tuning. By removing the class-specific regularization set and introducing random tokens and embeddings during generation, our approach allows for unlimited character creation while preserving the learned style. We evaluate our method on five small specialized datasets, comparing it to relevant baselines using both quantitative metrics and a human evaluation study. Our results demonstrate that our approach produces high-quality, diverse characters while preserving the distinctive aesthetic features of the reference characters, with human evaluation further reinforcing its effectiveness and highlighting the potential of our method.
Authors: Parhom Esmaeili, Virginia Fernandez, Pedro Borges, Eli Gibson, Sebastien Ourselin, M. Jorge Cardoso
Abstract: Interactive segmentation is a promising strategy for building robust, generalisable algorithms for volumetric medical image segmentation. However, inconsistent and clinically unrealistic evaluation hinders fair comparison and misrepresents real-world performance. We propose a clinically grounded methodology for defining evaluation tasks and metrics, and built a software framework for constructing standardised evaluation pipelines. We evaluate state-of-the-art algorithms across heterogeneous and complex tasks and observe that (i) minimising information loss when processing user interactions is critical for model robustness, (ii) adaptive-zooming mechanisms boost robustness and speed convergence, (iii) performance drops if validation prompting behaviour/budgets differ from training, (iv) 2D methods perform well with slab-like images and coarse targets, but 3D context helps with large or irregularly shaped targets, (v) performance of non-medical-domain models (e.g. SAM2) degrades with poor contrast and complex shapes.
Authors: Zixin Zhang, Kanghao Chen, Xingwang Lin, Lutao Jiang, Xu Zheng, Yuanhuiyi Lyu, Litao Guo, Yinchuan Li, Ying-Cong Chen
Abstract: The ability to use, understand, and create tools is a hallmark of human intelligence, enabling sophisticated interaction with the physical world. For any general-purpose intelligent agent to achieve true versatility, it must also master these fundamental skills. While modern Multimodal Large Language Models (MLLMs) leverage their extensive common knowledge for high-level planning in embodied AI and in downstream Vision-Language-Action (VLA) models, the extent of their true understanding of physical tools remains unquantified. To bridge this gap, we present PhysToolBench, the first benchmark dedicated to evaluating the comprehension of physical tools by MLLMs. Our benchmark is structured as a Visual Question Answering (VQA) dataset comprising over 1,000 image-text pairs. It assesses capabilities across three distinct difficulty levels: (1) Tool Recognition: Requiring the recognition of a tool's primary function. (2) Tool Understanding: Testing the ability to grasp the underlying principles of a tool's operation. (3) Tool Creation: Challenging the model to fashion a new tool from surrounding objects when conventional options are unavailable. Our comprehensive evaluation of 32 MLLMs-spanning proprietary, open-source, specialized embodied, and backbones in VLAs-reveals a significant deficiency in tool understanding. Furthermore, we provide an in-depth analysis and propose preliminary solutions. Code and dataset are publicly available.
Authors: David V\'azquez-Pad\'in, Fernando P\'erez-Gonz\'alez, Alejandro Mart\'in-Del-R\'io
Abstract: We investigate diagonal artifacts present in images captured by several Samsung smartphones and their impact on PRNU-based camera source verification. We first show that certain Galaxy S series models share a common pattern causing fingerprint collisions, with a similar issue also found in some Galaxy A models. Next, we demonstrate that reliable PRNU verification remains feasible for devices supporting PRO mode with raw capture, since raw images bypass the processing pipeline that introduces artifacts. This option, however, is not available for the mid-range A series models or in forensic cases without access to raw images. Finally, we outline potential forensic applications of the diagonal artifacts, such as reducing misdetections in HDR images and localizing regions affected by synthetic bokeh in portrait-mode images.
Authors: PeiHuang Zheng, Yunlong Zhao, Zheng Cui, Yang Li
Abstract: Small object detection in aerial images suffers from severe information degradation during feature extraction due to limited pixel representations, where shallow spatial details fail to align effectively with semantic information, leading to frequent misses and false positives. Existing FPN-based methods attempt to mitigate these losses through post-processing enhancements, but the reconstructed details often deviate from the original image information, impeding their fusion with semantic content. To address this limitation, we propose PRNet, a real-time detection framework that prioritizes the preservation and efficient utilization of primitive shallow spatial features to enhance small object representations. PRNet achieves this via two modules:the Progressive Refinement Neck (PRN) for spatial-semantic alignment through backbone reuse and iterative refinement, and the Enhanced SliceSamp (ESSamp) for preserving shallow information during downsampling via optimized rearrangement and convolution. Extensive experiments on the VisDrone, AI-TOD, and UAVDT datasets demonstrate that PRNet outperforms state-of-the-art methods under comparable computational constraints, achieving superior accuracy-efficiency trade-offs.
Authors: Arthur Bizzi, Matias Grynberg, Vitor Matias, Daniel Perazzo, Jo\~ao Paulo Lima, Luiz Velho, Nuno Gon\c{c}alves, Jo\~ao Pereira, Guilherme Schardong, Tiago Novello
Abstract: Morphing is a long-standing problem in vision and computer graphics, requiring a time-dependent warping for feature alignment and a blending for smooth interpolation. Recently, multilayer perceptrons (MLPs) have been explored as implicit neural representations (INRs) for modeling such deformations, due to their meshlessness and differentiability; however, extracting coherent and accurate morphings from standard MLPs typically relies on costly regularizations, which often lead to unstable training and prevent effective feature alignment. To overcome these limitations, we propose FLOWING (FLOW morphING), a framework that recasts warping as the construction of a differential vector flow, naturally ensuring continuity, invertibility, and temporal coherence by encoding structural flow properties directly into the network architectures. This flow-centric approach yields principled and stable transformations, enabling accurate and structure-preserving morphing of both 2D images and 3D shapes. Extensive experiments across a range of applications - including face and image morphing, as well as Gaussian Splatting morphing - show that FLOWING achieves state-of-the-art morphing quality with faster convergence. Code and pretrained models are available at http://schardong.github.io/flowing.
Authors: Minkyoung Cho, Ruben Ohana, Christian Jacobsen, Adityan Jothi, Min-Hung Chen, Z. Morley Mao, Ethem Can
Abstract: Current controllable diffusion models typically rely on fixed architectures that modify intermediate activations to inject guidance conditioned on a new modality. This approach uses a static conditioning strategy for a dynamic, multi-stage denoising process, limiting the model's ability to adapt its response as the generation evolves from coarse structure to fine detail. We introduce TC-LoRA (Temporally Modulated Conditional LoRA), a new paradigm that enables dynamic, context-aware control by conditioning the model's weights directly. Our framework uses a hypernetwork to generate LoRA adapters on-the-fly, tailoring weight modifications for the frozen backbone at each diffusion step based on time and the user's condition. This mechanism enables the model to learn and execute an explicit, adaptive strategy for applying conditional guidance throughout the entire generation process. Through experiments on various data domains, we demonstrate that this dynamic, parametric control significantly enhances generative fidelity and adherence to spatial conditions compared to static, activation-based methods. TC-LoRA establishes an alternative approach in which the model's conditioning strategy is modified through a deeper functional adaptation of its weights, allowing control to align with the dynamic demands of the task and generative stage.
Authors: Shubham Trehan, Udhav Ramachandran, Akash Rao, Ruth Scimeca, Sathyanarayanan N. Aakur
Abstract: Object detection in biomedical settings is fundamentally constrained by the scarcity of labeled data and the frequent emergence of novel or rare categories. We present FSP-DETR, a unified detection framework that enables robust few-shot detection, open-set recognition, and generalization to unseen biomedical tasks within a single model. Built upon a class-agnostic DETR backbone, our approach constructs class prototypes from original support images and learns an embedding space using augmented views and a lightweight transformer decoder. Training jointly optimizes a prototype matching loss, an alignment-based separation loss, and a KL divergence regularization to improve discriminative feature learning and calibration under scarce supervision. Unlike prior work that tackles these tasks in isolation, FSP-DETR enables inference-time flexibility to support unseen class recognition, background rejection, and cross-task adaptation without retraining. We also introduce a new ova species detection benchmark with 20 parasite classes and establish standardized evaluation protocols. Extensive experiments across ova, blood cell, and malaria detection tasks demonstrate that FSP-DETR significantly outperforms prior few-shot and prototype-based detectors, especially in low-shot and open-set scenarios.
Authors: Fengming Lin
Abstract: We present a transparent, reproducible measurement of research trends across 26,104 accepted papers from CVPR, ICLR, and NeurIPS spanning 2023-2025. Titles and abstracts are normalized, phrase-protected, and matched against a hand-crafted lexicon to assign up to 35 topical labels and mine fine-grained cues about tasks, architectures, training regimes, objectives, datasets, and co-mentioned modalities. The analysis quantifies three macro shifts: (1) a sharp rise of multimodal vision-language-LLM work, which increasingly reframes classic perception as instruction following and multi-step reasoning; (2) steady expansion of generative methods, with diffusion research consolidating around controllability, distillation, and speed; and (3) resilient 3D and video activity, with composition moving from NeRFs to Gaussian splatting and a growing emphasis on human- and agent-centric understanding. Within VLMs, parameter-efficient adaptation like prompting/adapters/LoRA and lightweight vision-language bridges dominate; training practice shifts from building encoders from scratch to instruction tuning and finetuning strong backbones; contrastive objectives recede relative to cross-entropy/ranking and distillation. Cross-venue comparisons show CVPR has a stronger 3D footprint and ICLR the highest VLM share, while reliability themes such as efficiency or robustness diffuse across areas. We release the lexicon and methodology to enable auditing and extension. Limitations include lexicon recall and abstract-only scope, but the longitudinal signals are consistent across venues and years.
Authors: Peiwen Sun, Shiqiang Lang, Dongming Wu, Yi Ding, Kaituo Feng, Huadai Liu, Zhen Ye, Rui Liu, Yun-Hui Liu, Jianan Wang, Xiangyu Yue
Abstract: With the current surge in spatial reasoning explorations, researchers have made significant progress in understanding indoor scenes, but still struggle with diverse applications such as robotics and autonomous driving. This paper aims to advance all-scale spatial reasoning across diverse scenarios by tackling two key challenges: 1) the heavy reliance on indoor 3D scans and labor-intensive manual annotations for dataset curation; 2) the absence of effective all-scale scene modeling, which often leads to overfitting to individual scenes. In this paper, we introduce a holistic solution that integrates a structured spatial reasoning knowledge system, scale-aware modeling, and a progressive training paradigm, as the first attempt to broaden the all-scale spatial intelligence of MLLMs to the best of our knowledge. Using a task-specific, specialist-driven automated pipeline, we curate over 38K video scenes across 5 spatial scales to create SpaceVista-1M, a dataset comprising approximately 1M spatial QA pairs spanning 19 diverse task types. While specialist models can inject useful domain knowledge, they are not reliable for evaluation. We then build an all-scale benchmark with precise annotations by manually recording, retrieving, and assembling video-based data. However, naive training with SpaceVista-1M often yields suboptimal results due to the potential knowledge conflict. Accordingly, we introduce SpaceVista-7B, a spatial reasoning model that accepts dense inputs beyond semantics and uses scale as an anchor for scale-aware experts and progressive rewards. Finally, extensive evaluations across 5 benchmarks, including our SpaceVista-Bench, demonstrate competitive performance, showcasing strong generalization across all scales and scenarios. Our dataset, model, and benchmark will be released on https://peiwensun2000.github.io/mm2km .
Authors: Shaoqi Dong, Chaoyou Fu, Haihan Gao, Yi-Fan Zhang, Chi Yan, Chu Wu, Xiaoyu Liu, Yunhang Shen, Jing Huo, Deqiang Jiang, Haoyu Cao, Yang Gao, Xing Sun, Ran He, Caifeng Shan
Abstract: Vision-Language Action (VLA) models significantly advance robotic manipulation by leveraging the strong perception capabilities of pretrained vision-language models (VLMs). By integrating action modules into these pretrained models, VLA methods exhibit improved generalization. However, training them from scratch is costly. In this work, we propose a simple yet effective distillation-based framework that equips VLMs with action-execution capability by transferring knowledge from pretrained small action models. Our architecture retains the original VLM structure, adding only an action token and a state encoder to incorporate physical inputs. To distill action knowledge, we adopt a two-stage training strategy. First, we perform lightweight alignment by mapping VLM hidden states into the action space of the small action model, enabling effective reuse of its pretrained action decoder and avoiding expensive pretraining. Second, we selectively fine-tune the language model, state encoder, and action modules, enabling the system to integrate multimodal inputs with precise action generation. Specifically, the action token provides the VLM with a direct handle for predicting future actions, while the state encoder allows the model to incorporate robot dynamics not captured by vision alone. This design yields substantial efficiency gains over training large VLA models from scratch. Compared with previous state-of-the-art methods, our method achieves 97.3% average success rate on LIBERO (11.8% improvement) and 93.5% on LIBERO-LONG (24.5% improvement). In real-world experiments across five manipulation tasks, our method consistently outperforms the teacher model, achieving 82.0% success rate (17% improvement), which demonstrate that action distillation effectively enables VLMs to generate precise actions while substantially reducing training costs.
Authors: Ruyi Xu, Guangxuan Xiao, Yukang Chen, Liuning He, Kelly Peng, Yao Lu, Song Han
Abstract: Vision-language models (VLMs) could power real-time assistants and autonomous agents, but they face a critical challenge: understanding near-infinite video streams without escalating latency and memory usage. Processing entire videos with full attention leads to quadratic computational costs and poor performance on long videos. Meanwhile, simple sliding window methods are also flawed, as they either break coherence or suffer from high latency due to redundant recomputation. In this paper, we introduce StreamingVLM, a model designed for real-time, stable understanding of infinite visual input. Our approach is a unified framework that aligns training with streaming inference. During inference, we maintain a compact KV cache by reusing states of attention sinks, a short window of recent vision tokens, and a long window of recent text tokens. This streaming ability is instilled via a simple supervised fine-tuning (SFT) strategy that applies full attention on short, overlapped video chunks, which effectively mimics the inference-time attention pattern without training on prohibitively long contexts. For evaluation, we build Inf-Streams-Eval, a new benchmark with videos averaging over two hours that requires dense, per-second alignment between frames and text. On Inf-Streams-Eval, StreamingVLM achieves a 66.18% win rate against GPT-4O mini and maintains stable, real-time performance at up to 8 FPS on a single NVIDIA H100. Notably, our SFT strategy also enhances general VQA abilities without any VQA-specific fine-tuning, improving performance on LongVideoBench by +4.30 and OVOBench Realtime by +5.96. Code is available at https://github.com/mit-han-lab/streaming-vlm.
Authors: Rui Hu, Delai Qiu, Yining Wang, Shengping Liu, Jitao Sang
Abstract: Automatic speech recognition (ASR) systems often struggle with domain-specific terminology, especially in specialized settings such as academic lectures. To address this, we define the SlideASR task, which leverages the rich visual information from presentation slides to improve transcription accuracy. Existing pipeline methods for this task tend to be complex and underperform. Although omni-modal large language models (OLLMs) provide a promising end-to-end framework, they frequently fail in practice by degenerating into simple optical character recognition (OCR) systems. To overcome this, we propose Visually-Anchored Policy Optimization (VAPO), a novel post-training method designed to control the model's reasoning process. Drawing on the Chain-of-Thought reasoning paradigm, VAPO enforces a structured "Look before Transcription" procedure using a
Authors: Mathias Boulanger, Ericmoore Jossou
Abstract: In this work, we investigate the use of spatio-temporalImplicit Neural Representations (INRs) for dynamic X-ray computed tomography (XCT) reconstruction under interlaced acquisition schemes. The proposed approach combines ADMM-based optimization with INCODE, a conditioning framework incorporating prior knowledge, to enable efficient convergence. We evaluate our method under diverse acquisition scenarios, varying the severity of global undersampling, spatial complexity (quantified via spatial information), and noise levels. Across all settings, our model achieves strong performance and outperforms Time-Interlaced Model-Based Iterative Reconstruction (TIMBIR), a state-of-the-art model-based iterative method. In particular, we show that the inductive bias of the INR provides good robustness to moderate noise levels, and that introducing explicit noise modeling through a weighted least squares data fidelity term significantly improves performance in more challenging regimes. The final part of this work explores extensions toward a practical reconstruction framework. We demonstrate the modularity of our approach by explicitly modeling detector non-idealities, incorporating ring artifact correction directly within the reconstruction process. Additionally, we present a proof-of-concept 4D volumetric reconstruction by jointly optimizing over batched axial slices, an approach which opens up the possibilities for massive parallelization, a critical feature for processing large-scale datasets.
Authors: Xunyang Zhu, Hongfei Ye, Yifei Wang, Taoran Liu, Jianjun Chen
Abstract: The sizing field defined on a triangular background grid is pivotal for controlling the quality and efficiency of unstructured mesh generation. However, creating an optimal background grid that is geometrically conforming, computationally lightweight, and free from artifacts like banding is a significant challenge. This paper introduces a novel, adaptive background grid simplification (ABGS) framework based on a Graph Convolutional Network (GCN). We reformulate the grid simplification task as an edge score regression problem and train a GCN model to efficiently predict optimal edge collapse candidates. The model is guided by a custom loss function that holistically considers both geometric fidelity and sizing field accuracy. This data-driven approach replaces a costly procedural evaluation, accelerating the simplification process. Experimental results demonstrate the effectiveness of our framework across diverse and complex engineering models. Compared to the initial dense grids, our simplified background grids achieve an element reduction of 74%-94%, leading to a 35%-88% decrease in sizing field query times.
Authors: Yiming Liang, Huan Yu, Zili Wang, Shuyou Zhang, Guodong Yi, Jin Wang, Jianrong Tan
Abstract: Recent advancements in AI-driven 3D model generation have leveraged cross modality, yet generating models with smooth surfaces and minimizing storage overhead remain challenges. This paper introduces a novel multi-stage framework for generating 3D models composed of parameterized primitives, guided by textual and image inputs. In the framework, A model generation algorithm based on parameterized primitives, is proposed, which can identifies the shape features of the model constituent elements, and replace the elements with parameterized primitives with high quality surface. In addition, a corresponding model storage method is proposed, it can ensure the original surface quality of the model, while retaining only the parameters of parameterized primitives. Experiments on virtual scene dataset and real scene dataset demonstrate the effectiveness of our method, achieving a Chamfer Distance of 0.003092, a VIoU of 0.545, a F1-Score of 0.9139 and a NC of 0.8369, with primitive parameter files approximately 6KB in size. Our approach is particularly suitable for rapid prototyping of simple models.
Authors: Jiacheng Liu, Peiliang Cai, Qinming Zhou, Yuqi Lin, Deyang Kong, Benhao Huang, Yupei Pan, Haowen Xu, Chang Zou, Junshu Tang, Shikang Zheng, Linfeng Zhang
Abstract: The application of diffusion transformers is suffering from their significant inference costs. Recently, feature caching has been proposed to solve this problem by reusing features from previous timesteps, thereby skipping computation in future timesteps. However, previous feature caching assumes that features in adjacent timesteps are similar or continuous, which does not always hold in all settings. To investigate this, this paper begins with an analysis from the frequency domain, which reveal that different frequency bands in the features of diffusion models exhibit different dynamics across timesteps. Concretely, low-frequency components, which decide the structure of images, exhibit higher similarity but poor continuity. In contrast, the high-frequency bands, which decode the details of images, show significant continuity but poor similarity. These interesting observations motivate us to propose Frequency-aware Caching (FreqCa) which directly reuses features of low-frequency components based on their similarity, while using a second-order Hermite interpolator to predict the volatile high-frequency ones based on its continuity. Besides, we further propose to cache Cumulative Residual Feature (CRF) instead of the features in all the layers, which reduces the memory footprint of feature caching by 99%. Extensive experiments on FLUX.1-dev, FLUX.1-Kontext-dev, Qwen-Image, and Qwen-Image-Edit demonstrate its effectiveness in both generation and editing. Codes are available in the supplementary materials and will be released on GitHub.
Authors: Yifei Dong, Fengyi Wu, Guangyu Chen, Zhi-Qi Cheng, Qiyu Hu, Yuxuan Zhou, Jingdong Sun, Jun-Yan He, Qi Dai, Alexander G Hauptmann
Abstract: Enabling embodied agents to effectively imagine future states is critical for robust and generalizable visual navigation. Current state-of-the-art approaches, however, adopt modular architectures that separate navigation planning from visual world modeling, leading to state-action misalignment and limited adaptability in novel or dynamic scenarios. To overcome this fundamental limitation, we propose UniWM, a unified, memory-augmented world model integrating egocentric visual foresight and planning within a single multimodal autoregressive backbone. Unlike modular frameworks, UniWM explicitly grounds action decisions in visually imagined outcomes, ensuring tight alignment between prediction and control. A hierarchical memory mechanism further integrates detailed short-term perceptual cues with longer-term trajectory context, enabling stable, coherent reasoning over extended horizons. Extensive experiments across four challenging benchmarks (Go Stanford, ReCon, SCAND, HuRoN) demonstrate that UniWM substantially improves navigation success rates by up to 30%, significantly reduces trajectory errors compared to strong baselines, and exhibits impressive zero-shot generalization on the unseen TartanDrive dataset. These results highlight UniWM as a principled step toward unified, imagination-driven embodied navigation.
Authors: Motahare Mounesan, Sourya Saha, Houchao Gan, Md. Nurul Absur, Saptarshi Debroy
Abstract: Real-time multi-view 3D reconstruction is a mission-critical application for key edge-native use cases, such as fire rescue, where timely and accurate 3D scene modeling enables situational awareness and informed decision-making. However, the dynamic and unpredictable nature of edge resource availability introduces disruptions, such as degraded image quality, unstable network links, and fluctuating server loads, which challenge the reliability of the reconstruction pipeline. In this work, we present a reinforcement learning (RL)-based edge resource management framework for reliable 3D reconstruction to ensure high quality reconstruction within a reasonable amount of time, despite the system operating under a resource-constrained and disruption-prone environment. In particular, the framework adopts two cooperative Q-learning agents, one for camera selection and one for server selection, both of which operate entirely online, learning policies through interactions with the edge environment. To support learning under realistic constraints and evaluate system performance, we implement a distributed testbed comprising lab-hosted end devices and FABRIC infrastructure-hosted edge servers to emulate smart city edge infrastructure under realistic disruption scenarios. Results show that the proposed framework improves application reliability by effectively balancing end-to-end latency and reconstruction quality in dynamic environments.
Authors: Thai-Hoang Pham, Jiayuan Chen, Seungyeon Lee, Yuanlong Wang, Sayoko Moroi, Xueru Zhang, Ping Zhang
Abstract: As machine learning (ML) algorithms are increasingly used in medical image analysis, concerns have emerged about their potential biases against certain social groups. Although many approaches have been proposed to ensure the fairness of ML models, most existing works focus only on medical image diagnosis tasks, such as image classification and segmentation, and overlooked prognosis scenarios, which involve predicting the likely outcome or progression of a medical condition over time. To address this gap, we introduce FairTTE, the first comprehensive framework for assessing fairness in time-to-event (TTE) prediction in medical imaging. FairTTE encompasses a diverse range of imaging modalities and TTE outcomes, integrating cutting-edge TTE prediction and fairness algorithms to enable systematic and fine-grained analysis of fairness in medical image prognosis. Leveraging causal analysis techniques, FairTTE uncovers and quantifies distinct sources of bias embedded within medical imaging datasets. Our large-scale evaluation reveals that bias is pervasive across different imaging modalities and that current fairness methods offer limited mitigation. We further demonstrate a strong association between underlying bias sources and model disparities, emphasizing the need for holistic approaches that target all forms of bias. Notably, we find that fairness becomes increasingly difficult to maintain under distribution shifts, underscoring the limitations of existing solutions and the pressing need for more robust, equitable prognostic models.
Authors: Ammar I Marvi, Nancy G Kanwisher, Meenakshi Khosla
Abstract: The ventral, dorsal, and lateral streams in high-level human visual cortex are implicated in distinct functional processes. Yet, deep neural networks (DNNs) trained on a single task model the entire visual system surprisingly well, hinting at common computational principles across these pathways. To explore this inconsistency, we applied a novel sparse decomposition approach to identify the dominant components of visual representations within each stream. Consistent with traditional neuroscience research, we find a clear difference in component response profiles across the three visual streams -- identifying components selective for faces, places, bodies, text, and food in the ventral stream; social interactions, implied motion, and hand actions in the lateral stream; and some less interpretable components in the dorsal stream. Building on this, we introduce Sparse Component Alignment (SCA), a new method for measuring representational alignment between brains and machines that better captures the latent neural tuning of these two visual systems. Using SCA, we find that standard visual DNNs are more aligned with the ventral than either dorsal or lateral representations. SCA reveals these distinctions with greater resolution than conventional population-level geometry, offering a measure of representational alignment that is sensitive to a system's underlying axes of neural tuning.
Authors: Zhen Yang, Yansong Ma, Lei Chen
Abstract: Traditional Evidence Deep Learning (EDL) methods rely on static hyperparameter for uncertainty calibration, limiting their adaptability in dynamic data distributions, which results in poor calibration and generalization in high-risk decision-making tasks. To address this limitation, we propose the Meta-Policy Controller (MPC), a dynamic meta-learning framework that adjusts the KL divergence coefficient and Dirichlet prior strengths for optimal uncertainty modeling. Specifically, MPC employs a bi-level optimization approach: in the inner loop, model parameters are updated through a dynamically configured loss function that adapts to the current training state; in the outer loop, a policy network optimizes the KL divergence coefficient and class-specific Dirichlet prior strengths based on multi-objective rewards balancing prediction accuracy and uncertainty quality. Unlike previous methods with fixed priors, our learnable Dirichlet prior enables flexible adaptation to class distributions and training dynamics. Extensive experimental results show that MPC significantly enhances the reliability and calibration of model predictions across various tasks, improving uncertainty calibration, prediction accuracy, and performance retention after confidence-based sample rejection.
Authors: Zhen Yang, Yansong Ma, Lei Chen
Abstract: Trustworthy medical image segmentation aims at deliver accurate and reliable results for clinical decision-making. Most existing methods adopt the evidence deep learning (EDL) paradigm due to its computational efficiency and theoretical robustness. However, the EDL-based methods often neglect leveraging uncertainty maps rich in attention cues to refine ambiguous boundary segmentation. To address this, we propose a progressive evidence uncertainty guided attention (PEUA) mechanism to guide the model to focus on the feature representation learning of hard regions. Unlike conventional approaches, PEUA progressively refines attention using uncertainty maps while employing low-rank learning to denoise attention weights, enhancing feature learning for challenging regions. Concurrently, standard EDL methods suppress evidence of incorrect class indiscriminately via Kullback-Leibler (KL) regularization, impairing the uncertainty assessment in ambiguous areas and consequently distorts the corresponding attention guidance. We thus introduce a semantic-preserving evidence learning (SAEL) strategy, integrating a semantic-smooth evidence generator and a fidelity-enhancing regularization term to retain critical semantics. Finally, by embedding PEUA and SAEL with the state-of-the-art U-KAN, we proposes Evidential U-KAN, a novel solution for trustworthy medical image segmentation. Extensive experiments on 4 datasets demonstrate superior accuracy and reliability over the competing methods. The code is available at \href{https://anonymous.4open.science/r/Evidence-U-KAN-BBE8}{github}.
Authors: Yingtie Lei, Zimeng Li, Chi-Man Pun, Yupeng Liu, Xuhang Chen
Abstract: Ultra-high-field 7T MRI offers enhanced spatial resolution and tissue contrast that enables the detection of subtle pathological changes in neurological disorders. However, the limited availability of 7T scanners restricts widespread clinical adoption due to substantial infrastructure costs and technical demands. Computational approaches for synthesizing 7T-quality images from accessible 3T acquisitions present a viable solution to this accessibility challenge. Existing CNN approaches suffer from limited spatial coverage, while Transformer models demand excessive computational overhead. RWKV architectures offer an efficient alternative for global feature modeling in medical image synthesis, combining linear computational complexity with strong long-range dependency capture. Building on this foundation, we propose Frequency Spatial-RWKV (FS-RWKV), an RWKV-based framework for 3T-to-7T MRI translation. To better address the challenges of anatomical detail preservation and global tissue contrast recovery, FS-RWKV incorporates two key modules: (1) Frequency-Spatial Omnidirectional Shift (FSO-Shift), which performs discrete wavelet decomposition followed by omnidirectional spatial shifting on the low-frequency branch to enhance global contextual representation while preserving high-frequency anatomical details; and (2) Structural Fidelity Enhancement Block (SFEB), a module that adaptively reinforces anatomical structure through frequency-aware feature fusion. Comprehensive experiments on UNC and BNU datasets demonstrate that FS-RWKV consistently outperforms existing CNN-, Transformer-, GAN-, and RWKV-based baselines across both T1w and T2w modalities, achieving superior anatomical fidelity and perceptual quality.
Authors: Yeqing Yang, Le Xu, Lixia Tian
Abstract: Accurate segmentation of 3D medical images is critical for clinical applications like disease assessment and treatment planning. While the Segment Anything Model 2 (SAM2) has shown remarkable success in video object segmentation by leveraging temporal cues, its direct application to 3D medical images faces two fundamental domain gaps: 1) the bidirectional anatomical continuity between slices contrasts sharply with the unidirectional temporal flow in videos, and 2) precise boundary delineation, crucial for morphological analysis, is often underexplored in video tasks. To bridge these gaps, we propose SAM2-3dMed, an adaptation of SAM2 for 3D medical imaging. Our framework introduces two key innovations: 1) a Slice Relative Position Prediction (SRPP) module explicitly models bidirectional inter-slice dependencies by guiding SAM2 to predict the relative positions of different slices in a self-supervised manner; 2) a Boundary Detection (BD) module enhances segmentation accuracy along critical organ and tissue boundaries. Extensive experiments on three diverse medical datasets (the Lung, Spleen, and Pancreas in the Medical Segmentation Decathlon (MSD) dataset) demonstrate that SAM2-3dMed significantly outperforms state-of-the-art methods, achieving superior performance in segmentation overlap and boundary precision. Our approach not only advances 3D medical image segmentation performance but also offers a general paradigm for adapting video-centric foundation models to spatial volumetric data.
Authors: Wenyi Wu, Kun Zhou, Ruoxin Yuan, Vivian Yu, Stephen Wang, Zhiting Hu, Biwei Huang
Abstract: We study how to endow GUI agents with scalable memory that help generalize across unfamiliar interfaces and long-horizon tasks. Prior GUI agents compress past trajectories into text tokens, which balloons context length and misses decisive visual cues (e.g., exact widget size and position). We propose a continuous memory that encodes each GUI trajectory into a fixed-length sequence of continuous embeddings using the VLM itself as an encoder; these embeddings are plugged directly into the backbone's input layer, sharply reducing context cost while preserving fine-grained visual information. As memory size and retrieval depth increase, performance improves monotonically, unlike text memories that degrade with long prompts. To grow memory at low cost, we introduce an auto-scaling data flywheel that (i) discovers new environments via search, (ii) synthesizes tasks with an open-source VLM, (iii) rolls out trajectories with the agent, and (iv) verifies success with the same VLM. Using this pipeline, we collect 100k+ trajectories for about \$4000 and fine-tune only the memory encoder (LoRA on a Q-Former, 1.2\% parameters) with 1,500 samples. On real-world GUI benchmarks, our memory-augmented agent consistently improves success rates under long horizons and distribution shifts. Notably, Qwen-2.5-VL-7B + continuous memory achieves performance comparable to state-of-the-art closed-source models (e.g., GPT-4o, Claude-4).
Authors: Jingxuan Wu, Zhenglin Wan, Xingrui Yu, Yuzhe Yang, Bo An, Ivor Tsang
Abstract: Flow-based text-to-image models follow deterministic trajectories, forcing users to repeatedly sample to discover diverse modes, which is a costly and inefficient process. We present a training-free, inference-time control mechanism that makes the flow itself diversity-aware. Our method simultaneously encourages lateral spread among trajectories via a feature-space objective and reintroduces uncertainty through a time-scheduled stochastic perturbation. Crucially, this perturbation is projected to be orthogonal to the generation flow, a geometric constraint that allows it to boost variation without degrading image details or prompt fidelity. Our procedure requires no retraining or modification to the base sampler and is compatible with common flow-matching solvers. Theoretically, our method is shown to monotonically increase a volume surrogate while, due to its geometric constraints, approximately preserving the marginal distribution. This provides a principled explanation for why generation quality is robustly maintained. Empirically, across multiple text-to-image settings under fixed sampling budgets, our method consistently improves diversity metrics such as the Vendi Score and Brisque over strong baselines, while upholding image quality and alignment.
Authors: Akira Takahashi, Shusuke Takahashi, Yuki Mitsufuji
Abstract: We introduce MMAudioSep, a generative model for video/text-queried sound separation that is founded on a pretrained video-to-audio model. By leveraging knowledge about the relationship between video/text and audio learned through a pretrained audio generative model, we can train the model more efficiently, i.e., the model does not need to be trained from scratch. We evaluate the performance of MMAudioSep by comparing it to existing separation models, including models based on both deterministic and generative approaches, and find it is superior to the baseline models. Furthermore, we demonstrate that even after acquiring functionality for sound separation via fine-tuning, the model retains the ability for original video-to-audio generation. This highlights the potential of foundational sound generation models to be adopted for sound-related downstream tasks. Our code is available at https://github.com/sony/mmaudiosep.
Authors: Zirun Zhou, Zhengyang Xiao, Haochuan Xu, Jing Sun, Di Wang, Jingfeng Zhang
Abstract: Recent advances in vision-language-action (VLA) models have greatly improved embodied AI, enabling robots to follow natural language instructions and perform diverse tasks. However, their reliance on uncurated training datasets raises serious security concerns. Existing backdoor attacks on VLAs mostly assume white-box access and result in task failures instead of enforcing specific actions. In this work, we reveal a more practical threat: attackers can manipulate VLAs by simply injecting physical objects as triggers into the training dataset. We propose goal-oriented backdoor attacks (GoBA), where the VLA behaves normally in the absence of physical triggers but executes predefined and goal-oriented actions in the presence of physical triggers. Specifically, based on a popular VLA benchmark LIBERO, we introduce BadLIBERO that incorporates diverse physical triggers and goal-oriented backdoor actions. In addition, we propose a three-level evaluation that categorizes the victim VLA's actions under GoBA into three states: nothing to do, try to do, and success to do. Experiments show that GoBA enables the victim VLA to successfully achieve the backdoor goal in 97 percentage of inputs when the physical trigger is present, while causing zero performance degradation on clean inputs. Finally, by investigating factors related to GoBA, we find that the action trajectory and trigger color significantly influence attack performance, while trigger size has surprisingly little effect. The code and BadLIBERO dataset are accessible via the project page at https://goba-attack.github.io/.
Authors: Alemu Sisay Nigru, Michele Svanera, Austin Dibble, Connor Dalby, Mattia Savardi, Sergio Benini
Abstract: Accurate segmentation of infant brain MRI is critical for studying early neurodevelopment and diagnosing neurological disorders. Yet, it remains a fundamental challenge due to continuously evolving anatomy of the subjects, motion artifacts, and the scarcity of high-quality labeled data. In this work, we present LODi, a novel framework that utilizes prior knowledge from an adult brain MRI segmentation model to enhance the segmentation performance of infant scans. Given the abundance of publicly available adult brain MRI data, we pre-train a segmentation model on a large adult dataset as a starting point. Through transfer learning and domain adaptation strategies, we progressively adapt the model to the 0-2 year-old population, enabling it to account for the anatomical and imaging variability typical of infant scans. The adaptation of the adult model is carried out using weakly supervised learning on infant brain scans, leveraging silver-standard ground truth labels obtained with FreeSurfer. By introducing a novel training strategy that integrates hierarchical feature refinement and multi-level consistency constraints, our method enables fast, accurate, age-adaptive segmentation, while mitigating scanner and site-specific biases. Extensive experiments on both internal and external datasets demonstrate the superiority of our approach over traditional supervised learning and domain-specific models. Our findings highlight the advantage of leveraging adult brain priors as a foundation for age-flexible neuroimaging analysis, paving the way for more reliable and generalizable brain MRI segmentation across the lifespan.
Authors: Till Aczel, Lucas Theis, Wattenhofer Roger
Abstract: Evaluating generative models is challenging because standard metrics often fail to reflect human preferences. Human evaluations are more reliable but costly and noisy, as participants vary in expertise, attention, and diligence. Pairwise comparisons improve consistency, yet aggregating them into overall quality scores requires careful modeling. Bradley-Terry-based methods update item scores from comparisons, but existing approaches either ignore rater variability or lack convergence guarantees, limiting robustness and interpretability. We introduce BBQ, a Bayesian Bradley-Terry variant that explicitly models rater quality, downweighting or removing unreliable participants, and provides guaranteed monotonic likelihood convergence through an Expectation-Maximization algorithm. Empirical results show that BBQ achieves faster convergence, well-calibrated uncertainty estimates, and more robust, interpretable rankings compared to baseline Bradley-Terry models, even with noisy or crowdsourced raters. This framework enables more reliable and cost-effective human evaluation of generative models.
Authors: Valentin Biller, Lucas Zimmer, Can Erdur, Sandeep Nagar, Daniel R\"uckert, Niklas Bubeck, Jonas Weidner
Abstract: Magnetic resonance imaging (MRI) inpainting supports numerous clinical and research applications. We introduce the first generative model that conditions on voxel-level, continuous tumor concentrations to synthesize high-fidelity brain tumor MRIs. For the BraTS 2025 Inpainting Challenge, we adapt this architecture to the complementary task of healthy tissue restoration by setting the tumor concentrations to zero. Our latent diffusion model conditioned on both tissue segmentations and the tumor concentrations generates 3D spatially coherent and anatomically consistent images for both tumor synthesis and healthy tissue inpainting. For healthy inpainting, we achieve a PSNR of 18.5, and for tumor inpainting, we achieve 17.4. Our code is available at: https://github.com/valentin-biller/ldm.git
Authors: Mert \.Inan, Anthony Sicilia, Alex Xie, Saujas Vaduguru, Daniel Fried, Malihe Alikhani
Abstract: Establishing shared goals is a fundamental step in human-AI communication. However, ambiguities can lead to outputs that seem correct but fail to reflect the speaker's intent. In this paper, we explore this issue with a focus on the data visualization domain, where ambiguities in natural language impact the generation of code that visualizes data. The availability of multiple views on the contextual (e.g., the intended plot and the code rendering the plot) allows for a unique and comprehensive analysis of diverse ambiguity types. We develop a taxonomy of types of ambiguity that arise in this task and propose metrics to quantify them. Using Matplotlib problems from the DS-1000 dataset, we demonstrate that our ambiguity metrics better correlate with human annotations than uncertainty baselines. Our work also explores how multi-turn dialogue can reduce ambiguity, therefore, improve code accuracy by better matching user goals. We evaluate three pragmatic models to inform our dialogue strategies: Gricean Cooperativity, Discourse Representation Theory, and Questions under Discussion. A simulated user study reveals how pragmatic dialogues reduce ambiguity and enhance code accuracy, highlighting the value of multi-turn exchanges in code generation.
Authors: Xiao Yu, Baolin Peng, Michel Galley, Hao Cheng, Qianhui Wu, Janardhan Kulkarni, Suman Nath, Zhou Yu, Jianfeng Gao
Abstract: Reasoning models have recently shown remarkable progress in domains such as math and coding. However, their expert-level abilities in math and coding contrast sharply with their performance in long-horizon, interactive tasks such as web navigation and computer/phone-use. Inspired by literature on human cognition, we argue that current AI agents need ''vicarious trial and error'' - the capacity to mentally simulate alternative futures before acting - in order to enhance their understanding and performance in complex interactive environments. We introduce Dyna-Mind, a two-stage training framework that explicitly teaches (V)LM agents to integrate such simulation into their reasoning. In stage 1, we introduce Reasoning with Simulations (ReSim), which trains the agent to generate structured reasoning traces from expanded search trees built from real experience gathered through environment interactions. ReSim thus grounds the agent's reasoning in faithful world dynamics and equips it with the ability to anticipate future states in its reasoning. In stage 2, we propose Dyna-GRPO, an online reinforcement learning method to further strengthen the agent's simulation and decision-making ability by using both outcome rewards and intermediate states as feedback from real rollouts. Experiments on two synthetic benchmarks (Sokoban and ALFWorld) and one realistic benchmark (AndroidWorld) demonstrate that (1) ReSim effectively infuses simulation ability into AI agents, and (2) Dyna-GRPO leverages outcome and interaction-level signals to learn better policies for long-horizon, planning-intensive tasks. Together, these results highlight the central role of simulation in enabling AI agents to reason, plan, and act more effectively in the ever more challenging environments.
Authors: Disharee Bhowmick, Ranjith Ramanathan, Sathyanarayanan N. Aakur
Abstract: Time series data often contain latent temporal structure, transitions between locally stationary regimes, repeated motifs, and bursts of variability, that are rarely leveraged in standard representation learning pipelines. Existing models typically operate on raw or fixed-window sequences, treating all time steps as equally informative, which leads to inefficiencies, poor robustness, and limited scalability in long or noisy sequences. We propose STaTS, a lightweight, unsupervised framework for Structure-Aware Temporal Summarization that adaptively compresses both univariate and multivariate time series into compact, information-preserving token sequences. STaTS detects change points across multiple temporal resolutions using a BIC-based statistical divergence criterion, then summarizes each segment using simple functions like the mean or generative models such as GMMs. This process achieves up to 30x sequence compression while retaining core temporal dynamics. STaTS operates as a model-agnostic preprocessor and can be integrated with existing unsupervised time series encoders without retraining. Extensive experiments on 150+ datasets, including classification tasks on the UCR-85, UCR-128, and UEA-30 archives, and forecasting on ETTh1 and ETTh2, ETTm1, and Electricity, demonstrate that STaTS enables 85-90\% of the full-model performance while offering dramatic reductions in computational cost. Moreover, STaTS improves robustness under noise and preserves discriminative structure, outperforming uniform and clustering-based compression baselines. These results position STaTS as a principled, general-purpose solution for efficient, structure-aware time series modeling.
Authors: Luigi Celona, Simone Bianco, Marco Donzella, Paolo Napoletano
Abstract: State-of-The-Art (SoTA) image captioning models are often trained on the MicroSoft Common Objects in Context (MS-COCO) dataset, which contains human-annotated captions with an average length of approximately ten tokens. Although effective for general scene understanding, these short captions often fail to capture complex scenes and convey detailed information. Moreover, captioning models tend to exhibit bias towards the ``average'' caption, which captures only the more general aspects, thus overlooking finer details. In this paper, we present a novel approach to generate richer and more informative image captions by combining the captions generated from different SoTA captioning models. Our proposed method requires no additional model training: given an image, it leverages pre-trained models from the literature to generate the initial captions, and then ranks them using a newly introduced image-text-based metric, which we name BLIPScore. Subsequently, the top two captions are fused using a Large Language Model (LLM) to produce the final, more detailed description. Experimental results on the MS-COCO and Flickr30k test sets demonstrate the effectiveness of our approach in terms of caption-image alignment and hallucination reduction according to the ALOHa, CAPTURE, and Polos metrics. A subjective study lends additional support to these results, suggesting that the captions produced by our model are generally perceived as more consistent with human judgment. By combining the strengths of diverse SoTA models, our method enhances the quality and appeal of image captions, bridging the gap between automated systems and the rich and informative nature of human-generated descriptions. This advance enables the generation of more suitable captions for the training of both vision-language and captioning models.
Authors: Elena Camuffo, Umberto Michieli, Jijoong Moon, Daehyun Kim, Mete Ozay
Abstract: Improving model robustness in case of corrupted images is among the key challenges to enable robust vision systems on smart devices, such as robotic agents. Particularly, robust test-time performance is imperative for most of the applications. This paper presents a novel approach to improve robustness of any classification model, especially on severely corrupted images. Our method (FROST) employs high-frequency features to detect input image corruption type, and select layer-wise feature normalization statistics. FROST provides the state-of-the-art results for different models and datasets, outperforming competitors on ImageNet-C by up to 37.1% relative gain, improving baseline of 40.9% mCE on severe corruptions.
Authors: Qinhao Zhou, Yuwen Tan, Boqing Gong, Xiang Xiang
Abstract: Class-incremental learning (CIL) aims to enable models to continuously learn new classes while overcoming catastrophic forgetting. The introduction of pre-trained models has brought new tuning paradigms to CIL. In this paper, we revisit different parameter-efficient tuning (PET) methods within the context of continual learning. We observe that adapter tuning demonstrates superiority over prompt-based methods, even without parameter expansion in each learning session. Motivated by this, we propose incrementally tuning the shared adapter without imposing parameter update constraints, enhancing the learning capacity of the backbone. Additionally, we employ feature sampling from stored prototypes to retrain a unified classifier, further improving its performance. We estimate the semantic shift of old prototypes without access to past samples and update stored prototypes session by session. Our proposed method eliminates model expansion and avoids retaining any image samples. It surpasses previous pre-trained model-based CIL methods and demonstrates remarkable continual learning capabilities. Experimental results on five CIL benchmarks validate the effectiveness of our approach, achieving state-of-the-art (SOTA) performance.
Authors: Haojun Jiang, Teng Wang, Zhenguo Sun, Yulin Wang, Yang Yue, Yu Sun, Ning Jia, Meng Li, Shaqi Luo, Shiji Song, Gao Huang
Abstract: Echocardiography is an essential medical technique for diagnosing cardiovascular diseases, but its high operational complexity has led to a shortage of trained professionals. To address this issue, we introduce a novel probe movement guidance algorithm that has the potential to be applied in guiding robotic systems or novices with probe pose adjustment for high-quality standard plane image acquisition.Cardiac ultrasound faces two major challenges: (1) the inherently complex structure of the heart, and (2) significant individual variations. Previous works have only learned the population-averaged structure of the heart rather than personalized cardiac structures, leading to a performance bottleneck. Clinically, we observe that sonographers dynamically adjust their interpretation of a patient's cardiac anatomy based on prior scanning sequences, consequently refining their scanning strategies. Inspired by this, we propose a novel sequence-aware self-supervised pre-training method. Specifically, our approach learns personalized three-dimensional cardiac structural features by predicting the masked-out image features and probe movement actions in a scanning sequence. We hypothesize that if the model can predict the missing content it has acquired a good understanding of personalized cardiac structure. Extensive experiments on a large-scale expert scanning dataset with 1.67 million samples demonstrate that our proposed sequence-aware paradigm can effectively reduce probe guidance errors compared to other advanced baseline methods.
Authors: Minh Bui, Kostas Alexis
Abstract: Vision-based perception and reasoning is essential for scene understanding in any autonomous system. RGB and depth images are commonly used to capture both the semantic and geometric features of the environment. Developing methods to reliably interpret this data is critical for real-world applications, where noisy measurements are often unavoidable. In this work, we introduce a diffusion-based framework to address the RGB-D semantic segmentation problem. Additionally, we demonstrate that utilizing a Deformable Attention Transformer as the encoder to extract features from depth images effectively captures the characteristics of invalid regions in depth measurements. Our generative framework shows a greater capacity to model the underlying distribution of RGB-D images, achieving robust performance in challenging scenarios with significantly less training time compared to discriminative methods. Experimental results indicate that our approach achieves State-of-the-Art performance on both the NYUv2 and SUN-RGBD datasets in general and especially in the most challenging of their image data. Our project page will be available at https://diffusionmms.github.io/
Authors: Shubhi Bansal, Sreeharish A, Madhava Prasath J, Manikandan S, Sreekanth Madisetty, Mohammad Zia Ur Rehman, Chandravardhan Singh Raghaw, Gaurav Duggal, Nagendra Kumar
Abstract: Mamba, a special case of the State Space Model, is gaining popularity as an alternative to template-based deep learning approaches in medical image analysis. While transformers are powerful architectures, they have drawbacks, including quadratic computational complexity and an inability to address long-range dependencies efficiently. This limitation affects the analysis of large and complex datasets in medical imaging, where there are many spatial and temporal relationships. In contrast, Mamba offers benefits that make it well-suited for medical image analysis. It has linear time complexity, which is a significant improvement over transformers. Mamba processes longer sequences without attention mechanisms, enabling faster inference and requiring less memory. Mamba also demonstrates strong performance in merging multimodal data, improving diagnosis accuracy and patient outcomes. The organization of this paper allows readers to appreciate the capabilities of Mamba in medical imaging step by step. We begin by defining core concepts of SSMs and models, including S4, S5, and S6, followed by an exploration of Mamba architectures such as pure Mamba, U-Net variants, and hybrid models with convolutional neural networks, transformers, and Graph Neural Networks. We also cover Mamba optimizations, techniques and adaptations, scanning, datasets, applications, experimental results, and conclude with its challenges and future directions in medical imaging. This review aims to demonstrate the transformative potential of Mamba in overcoming existing barriers within medical imaging while paving the way for innovative advancements in the field. A comprehensive list of Mamba architectures applied in the medical field, reviewed in this work, is available at Github.
Authors: Soumava Paul, Prakhar Kaushik, Alan Yuille
Abstract: In this work, we introduce a generative approach for pose-free (without camera parameters) reconstruction of 360 scenes from a sparse set of 2D images. Pose-free scene reconstruction from incomplete, pose-free observations is usually regularized with depth estimation or 3D foundational priors. While recent advances have enabled sparse-view reconstruction of large complex scenes (with high degree of foreground and background detail) with known camera poses using view-conditioned generative priors, these methods cannot be directly adapted for the pose-free setting when ground-truth poses are not available during evaluation. To address this, we propose an image-to-image generative model designed to inpaint missing details and remove artifacts in novel view renders and depth maps of a 3D scene. We introduce context and geometry conditioning using Feature-wise Linear Modulation (FiLM) modulation layers as a lightweight alternative to cross-attention and also propose a novel confidence measure for 3D Gaussian splat representations to allow for better detection of these artifacts. By progressively integrating these novel views in a Gaussian-SLAM-inspired process, we achieve a multi-view-consistent 3D representation. Evaluations on the MipNeRF360 and DL3DV-10K benchmark dataset demonstrate that our method surpasses existing pose-free techniques and performs competitively with state-of-the-art posed (precomputed camera parameters are given) reconstruction methods in complex 360 scenes. Our project page provides additional results, videos, and code.
Authors: Xianfeng Tan, Yuhan Li, Wenxiang Shang, Yubo Wu, Jian Wang, Xuanhong Chen, Yi Zhang, Ran Lin, Bingbing Ni
Abstract: Standard clothing asset generation involves restoring forward-facing flat-lay garment images displayed on a clear background by extracting clothing information from diverse real-world contexts, which presents significant challenges due to highly standardized structure sampling distributions and clothing semantic absence in complex scenarios. Existing models have limited spatial perception, often exhibiting structural hallucinations and texture distortion in this high-specification generative task. To address this issue, we propose a novel Retrieval-Augmented Generation (RAG) framework, termed RAGDiffusion, to enhance structure determinacy and mitigate hallucinations by assimilating knowledge from language models and external databases. RAGDiffusion consists of two processes: (1) Retrieval-based structure aggregation, which employs contrastive learning and a Structure Locally Linear Embedding (SLLE) to derive global structure and spatial landmarks, providing both soft and hard guidance to counteract structural ambiguities; and (2) Omni-level faithful garment generation, which introduces a coarse-to-fine texture alignment that ensures fidelity in pattern and detail components within the diffusing. Extensive experiments on challenging real-world datasets demonstrate that RAGDiffusion synthesizes structurally and texture-faithful clothing assets with significant performance improvements, representing a pioneering effort in high-specification faithful generation with RAG to confront intrinsic hallucinations and enhance fidelity.
Authors: Ziqi Gao, Weikai Huang, Jieyu Zhang, Aniruddha Kembhavi, Ranjay Krishna
Abstract: Recent advances in text-to-vision generation excel in visual fidelity but struggle with compositional generalization and semantic alignment. Existing datasets are noisy and weakly compositional, limiting models' understanding of complex scenes, while scalable solutions for dense, high-quality annotations remain a challenge. We introduce Generate Any Scene, a data engine that systematically enumerates scene graphs representing the combinatorial array of possible visual scenes. Generate Any Scene dynamically constructs scene graphs of varying complexity from a structured taxonomy of objects, attributes, and relations. Given a sampled scene graph, Generate Any Scene translates it into a caption for text-to-image or text-to-video generation; it also translates it into a set of visual question answers that allow automatic evaluation and reward modeling of semantic alignment. Using Generate Any Scene, we first design a self-improving framework where models iteratively enhance their performance using generated data. Stable Diffusion v1.5 achieves an average 4% improvement over baselines and surpassing fine-tuning on CC3M. Second, we also design a distillation algorithm to transfer specific strengths from proprietary models to their open-source counterparts. Using fewer than 800 synthetic captions, we fine-tune Stable Diffusion v1.5 and achieve a 10% increase in TIFA score on compositional and hard concept generation. Third, we create a reward model to align model generation with semantic accuracy at a low cost. Using GRPO algorithm, we fine-tune SimpleAR-0.5B-SFT and surpass CLIP-based methods by +5% on DPG-Bench. Finally, we apply these ideas to the downstream task of content moderation where we train models to identify challenging cases by learning from synthetic data.
Authors: Bhanu Tokas, Rahul Nair, Hannah Kerner
Abstract: Most of the ML datasets we use today are biased. When we train models on these biased datasets, they often not only learn dataset biases but can also amplify them -- a phenomenon known as bias amplification. Several co-occurrence-based metrics have been proposed to measure bias amplification between a protected attribute A (e.g., gender) and a task T (e.g., cooking). However, these metrics fail to measure biases when A is balanced with T. To measure bias amplification in balanced datasets, recent work proposed a predictability-based metric called leakage amplification. However, leakage amplification cannot identify the direction in which biases are amplified. In this work, we propose a new predictability-based metric called directional predictability amplification (DPA). DPA measures directional bias amplification, even for balanced datasets. Unlike leakage amplification, DPA is easier to interpret and less sensitive to attacker models (a hyperparameter in predictability-based metrics). Our experiments on tabular and image datasets show that DPA is an effective metric for measuring directional bias amplification. The code will be available soon.
Authors: Robin Sch\"on, Julian Lorenz, Daniel Kienzle, Rainer Lienhart
Abstract: In this paper, we present a novel architecture for interactive segmentation in winter sports contexts. The field of interactive segmentation deals with the prediction of high-quality segmentation masks by informing the network about the objects position with the help of user guidance. In our case the guidance consists of click prompts. For this task, we first present a baseline architecture which is specifically geared towards quickly responding after each click. Afterwards, we motivate and describe a number of architectural modifications which improve the performance when tasked with segmenting winter sports equipment on the WSESeg dataset. With regards to the average NoC@85 metric on the WSESeg classes, we outperform SAM and HQ-SAM by 2.336 and 7.946 clicks, respectively. When applied to the HQSeg-44k dataset, our system delivers state-of-the-art results with a NoC@90 of 6.00 and NoC@95 of 9.89. In addition to that, we test our model on a novel dataset containing masks for humans during skiing.
Authors: Nicolas Deperrois, Hidetoshi Matsuo, Samuel Ruip\'erez-Campillo, Moritz Vandenhirtz, Sonia Laguna, Alain Ryser, Koji Fujimoto, Mizuho Nishio, Thomas M. Sutter, Julia E. Vogt, Jonas Kluckert, Thomas Frauenfelder, Christian Bl\"uthgen, Farhad Nooralahzadeh, Michael Krauthammer
Abstract: The widespread use of chest X-rays (CXRs), coupled with a shortage of radiologists, has driven growing interest in automated CXR analysis and AI-assisted reporting. While existing vision-language models (VLMs) show promise in specific tasks such as report generation or abnormality detection, they often lack support for interactive diagnostic capabilities. In this work we present RadVLM, a compact, multitask conversational foundation model designed for CXR interpretation. To this end, we curate a large-scale instruction dataset comprising over 1 million image-instruction pairs containing both single-turn tasks -- such as report generation, abnormality classification, and visual grounding -- and multi-turn, multi-task conversational interactions. After fine-tuning RadVLM on this instruction dataset, we evaluate it across different tasks along with re-implemented baseline VLMs. Our results show that RadVLM achieves state-of-the-art performance in conversational capabilities and visual grounding while remaining competitive in other radiology tasks. Ablation studies further highlight the benefit of joint training across multiple tasks, particularly for scenarios with limited annotated data. Together, these findings highlight the potential of RadVLM as a clinically relevant AI assistant, providing structured CXR interpretation and conversational capabilities to support more effective and accessible diagnostic workflows.
Authors: Francesco Pezone, Sergio Barbarossa, Giuseppe Caire
Abstract: This work introduces Semantically Masked Vector Quantized Generative Adversarial Network (SQ-GAN), a novel approach integrating semantically driven image coding and vector quantization to optimize image compression for semantic/task-oriented communications. The method only acts on source coding and is fully compliant with legacy systems. The semantics is extracted from the image computing its semantic segmentation map using off-the-shelf software. A new specifically developed semantic-conditioned adaptive mask module (SAMM) selectively encodes semantically relevant features of the image. The relevance of the different semantic classes is task-specific, and it is incorporated in the training phase by introducing appropriate weights in the loss function. SQ-GAN outperforms state-of-the-art image compression schemes such as JPEG2000, BPG, and deep-learning based methods across multiple metrics, including perceptual quality and semantic segmentation accuracy on the reconstructed image, at extremely low compression rates.
Authors: Fanhu Zeng, Haiyang Guo, Fei Zhu, Li Shen, Hao Tang
Abstract: Fine-tuning pre-trained models with custom data leads to numerous expert models on specific tasks. Merging models into one universal model to empower multi-task ability refraining from data leakage has gained popularity. With the expansion in data and model size, parameter-efficient tuning becomes the common practice for obtaining task-specific models efficiently. However, few methods are dedicated to efficient merging, and existing methods designed for full fine-tuning merging fail under efficient merging. To address the issue, we analyze from low-rank decomposition and reveal that direction robustness during merging is crucial for merging efficient modules. We furthermore uncover that compensating for the gap between stark singular values contributes to direction robustness. Therefore, we propose RobustMerge, a training-free parameter-efficient merging method with complementary parameter adaptation to maintain direction robustness. Specifically, we (1) prune parameters and scale coefficients from inter-parameter relation for singular values to maintain direction stability away from task interference, and (2) perform cross-task normalization to enhance unseen task generalization. We establish a benchmark consisting of diverse multimodal tasks, on which we conduct experiments to certify the outstanding performance and generalizability of our method. Additional studies and extensive analyses further showcase the effectiveness. Code is available at https://github.com/AuroraZengfh/RobustMerge.
Authors: Rahul Nair, Bhanu Tokas, Hannah Kerner
Abstract: When we train models on biased ML datasets, they not only learn these biases but can inflate them at test time - a phenomenon called bias amplification. To measure bias amplification in ML datasets, many co-occurrence-based metrics have been proposed. Co-occurrence-based metrics are effective in measuring bias amplification in simple problems like image classification. However, these metrics are ineffective for complex problems like image captioning as they cannot capture the semantics of a caption. To measure bias amplification in captions, prior work introduced a predictability-based metric called Leakage in Captioning (LIC). While LIC captures the semantics and context of captions, it has limitations. LIC cannot identify the direction in which bias is amplified, poorly estimates dataset bias due to a weak vocabulary substitution strategy, and is highly sensitive to attacker models (a hyperparameter in predictability-based metrics). To overcome these issues, we propose Directional Predictability Amplification in Captioning (DPAC). DPAC measures directional bias amplification in captions, provides a better estimate of dataset bias using an improved substitution strategy, and is less sensitive to attacker models. Our experiments on the COCO captioning dataset show how DPAC is the most reliable metric to measure bias amplification in captions.
Authors: Asifullah Khan, Laiba Asmatullah, Anza Malik, Shahzaib Khan, Hamna Asif
Abstract: Self-supervised learning is a machine learning approach that generates implicit labels by learning underlined patterns and extracting discriminative features from unlabeled data without manual labelling. Contrastive learning introduces the concept of "positive" and "negative" samples, where positive pairs (e.g., variation of the same image/object) are brought together in the embedding space, and negative pairs (e.g., views from different images/objects) are pushed farther away. This methodology has shown significant improvements in image understanding and image text analysis without much reliance on labeled data. In this paper, we comprehensively discuss the terminologies, recent developments and applications of contrastive learning with respect to text-image models. Specifically, we provide an overview of the approaches of contrastive learning in text-image models in recent years. Secondly, we categorize the approaches based on different model structures. Thirdly, we further introduce and discuss the latest advances of the techniques used in the process such as pretext tasks for both images and text, architectural structures, and key trends. Lastly, we discuss the recent state-of-art applications of self-supervised contrastive learning Text-Image based models.
Authors: Zhichao Sun, Huazhang Hu, Yidong Ma, Gang Liu, Yibo Chen, Xu Tang, Yao Hu, Yongchao Xu
Abstract: With the exponential growth of data, traditional object detection methods are increasingly struggling to handle vast vocabulary object detection tasks effectively. We analyze two key limitations of classification-based detectors: positive gradient dilution, where rare positive categories receive insufficient learning signals, and hard negative gradient dilution, where discriminative gradients are overwhelmed by numerous easy negatives. To address these challenges, we propose CQ-DINO, a category query-based object detection framework that reformulates classification as a contrastive task between object queries and learnable category queries. Our method introduces image-guided query selection, which reduces the negative space by adaptively retrieving top-K relevant categories per image via cross-attention, thereby rebalancing gradient distributions and facilitating implicit hard example mining. Furthermore, CQ-DINO flexibly integrates explicit hierarchical category relationships in structured datasets (e.g., V3Det) or learns implicit category correlations via self-attention in generic datasets (e.g., COCO). Experiments demonstrate that CQ-DINO achieves superior performance on the challenging V3Det benchmark (surpassing previous methods by 2.1% AP) while maintaining competitiveness in COCO. Our work provides a scalable solution for real-world detection systems requiring wide category coverage. The code is publicly at https://github.com/FireRedTeam/CQ-DINO.
Authors: Sanjoy Kundu, Shanmukha Vellamcheti, Sathyanarayanan N. Aakur
Abstract: Open-world egocentric activity recognition poses a fundamental challenge due to its unconstrained nature, requiring models to infer unseen activities from an expansive, partially observed search space. We introduce ProbRes, a Probabilistic Residual search framework based on jump-diffusion that efficiently navigates this space by balancing prior-guided exploration with likelihood-driven exploitation. Our approach integrates structured commonsense priors to construct a semantically coherent search space, adaptively refines predictions using Vision-Language Models (VLMs) and employs a stochastic search mechanism to locate high-likelihood activity labels while minimizing exhaustive enumeration efficiently. We systematically evaluate ProbRes across multiple openness levels (L0-L3), demonstrating its adaptability to increasing search space complexity. In addition to achieving state-of-the-art performance on benchmark datasets (GTEA Gaze, GTEA Gaze+, EPIC-Kitchens, and Charades-Ego), we establish a clear taxonomy for open-world recognition, delineating the challenges and methodological advancements necessary for egocentric activity understanding. Our results highlight the importance of structured search strategies, paving the way for scalable and efficient open-world activity recognition.
Authors: Kaushik Bhargav Sivangi, Paul Henderson, Fani Deligianni
Abstract: Human pose estimation (HPE) has become essential in numerous applications including healthcare, activity recognition, and human-computer interaction. However, the privacy implications of processing sensitive visual data present significant deployment barriers in critical domains. While traditional anonymization techniques offer limited protection and often compromise data utility for broader motion analysis, Differential Privacy (DP) provides formal privacy guarantees but typically degrades model performance when applied naively. In this work, we present the first comprehensive framework for differentially private 2D human pose estimation (2D-HPE) by applying Differentially Private Stochastic Gradient Descent (DP-SGD) to this task. To effectively balance privacy with performance, we adopt Projected DP-SGD (PDP-SGD), which projects the noisy gradients to a low-dimensional subspace. Next, we incorporate Feature Differential Privacy(FDP) to selectively privatize only sensitive features while retaining public visual cues. Finally, we propose a hybrid feature-projective DP framework that combines both approaches to balance privacy and accuracy for HPE. We evaluate our approach on the MPII dataset across varying privacy budgets, training strategies, and clipping norms. Our combined feature-projective method consistently outperforms vanilla DP-SGD and individual baselines, achieving up to 82.61\% mean PCKh@0.5 at $\epsilon = 0.8$, substantially closing the gap to the non-private performance. This work lays foundation for privacy-preserving human pose estimation in real-world, sensitive applications.
Authors: Haoran Chen, Junyan Lin, Xinghao Chen, Yue Fan, Jianfeng Dong, Xin Jin, Hui Su, Jinlan Fu, Xiaoyu Shen
Abstract: Multimodal large language models (MLLMs) typically extract visual features from the final layers of a pretrained Vision Transformer (ViT). This widespread deep-layer bias, however, is largely driven by empirical convention rather than principled analysis. While prior studies suggest that different ViT layers capture different types of information, with shallower layers focusing on fine visual details and deeper layers aligning more closely with textual semantics, the impact of this variation on MLLM performance remains underexplored. We present the first comprehensive study of visual layer selection for MLLMs, analyzing representation similarity across ViT layers to establish shallow, middle, and deep layer groupings. Through extensive evaluation of MLLMs (1.4B-7B parameters) across 10 benchmarks encompassing 60+ tasks, we find that while deep layers excel in semantic-rich tasks like OCR, shallow and middle layers significantly outperform them on fine-grained visual tasks including counting, positioning, and object localization. Building on these insights, we propose a lightweight feature fusion method that strategically incorporates shallower layers, achieving consistent improvements over both single-layer and specialized fusion baselines. Our work offers the first principled study of visual layer selection in MLLMs, showing that MLLMs can often see better when they look shallower.
Authors: Hao Xu, Arbind Agrahari Baniya, Sam Well, Mohamed Reda Bouadjenek, Richard Dazeley, Sunil Aryal
Abstract: Video event detection has become a cornerstone of modern sports analytics, powering automated performance evaluation, content generation, and tactical decision-making. Recent advances in deep learning have driven progress in related tasks such as Temporal Action Localization (TAL), which detects extended action segments; Action Spotting (AS), which identifies a representative timestamp; and Precise Event Spotting (PES), which pinpoints the exact frame of an event. Although closely connected, their subtle differences often blur the boundaries between them, leading to confusion in both research and practical applications. Furthermore, prior surveys either address generic video event detection or broader sports video tasks, but largely overlook the unique temporal granularity and domain-specific challenges of event spotting. In addition, most existing sports video surveys focus on elite-level competitions while neglecting the wider community of everyday practitioners. This survey addresses these gaps by: (i) clearly delineating TAL, AS, and PES and their respective use cases; (ii) introducing a structured taxonomy of state of the art approaches including temporal modeling strategies, multimodal frameworks, and data-efficient pipelines tailored for AS and PES; and (iii) critically assessing benchmark datasets and evaluation protocols, highlighting limitations such as reliance on broadcast quality footage and metrics that over reward permissive multilabel predictions. By synthesizing current research and exposing open challenges, this work provides a comprehensive foundation for developing temporally precise, generalizable, and practically deployable sports event detection systems for both the research and industry communities.
Authors: Wei Li, Dezhao Luo, Dongbao Yang, Zhenhang Li, Weiping Wang, Yu Zhou
Abstract: Video action understanding tasks in real-world scenarios always suffer data limitations. In this paper, we address the data-limited action understanding problem by bridging data scarcity. We propose a novel method that employs a text-to-video diffusion transformer to generate annotated data for model training. This paradigm enables the generation of realistic annotated data on an infinite scale without human intervention. We proposed the information enhancement strategy and the uncertainty-based label smoothing tailored to generate sample training. Through quantitative and qualitative analysis, we observed that real samples generally contain a richer level of information than generated samples. Based on this observation, the information enhancement strategy is proposed to enhance the informative content of the generated samples from two aspects: the environments and the characters. Furthermore, we observed that some low-quality generated samples might negatively affect model training. To address this, we devised the uncertainty-based label smoothing strategy to increase the smoothing of these samples, thus reducing their impact. We demonstrate the effectiveness of the proposed method on four datasets across five tasks and achieve state-of-the-art performance for zero-shot action recognition.
Authors: Kele Shao, Keda Tao, Can Qin, Haoxuan You, Yang Sui, Huan Wang
Abstract: Video large language models (video LLMs) excel at video comprehension but face significant computational inefficiency due to redundant video tokens. Existing token pruning methods offer solutions. However, approaches operating within the LLM (inner-LLM pruning), such as FastV, incur intrinsic computational overhead in shallow layers. In contrast, methods performing token pruning before the LLM (outer-LLM pruning) primarily address spatial redundancy within individual frames or limited temporal windows, neglecting the crucial global temporal dynamics and correlations across longer video sequences. This leads to sub-optimal spatio-temporal reduction and does not leverage video compressibility fully. Crucially, the synergistic potential and mutual influence of combining these strategies remain unexplored. To further reduce redundancy, we introduce HoliTom, a novel training-free holistic token merging framework. HoliTom employs outer-LLM pruning through global redundancy-aware temporal segmentation, followed by spatial-temporal merging to reduce visual tokens by over 90%, significantly alleviating the LLM's computational burden. Complementing this, we introduce a robust inner-LLM token similarity-based merging approach, designed for superior performance and compatibility with outer-LLM pruning. Evaluations demonstrate our method's promising efficiency-performance trade-off on LLaVA-OneVision-7B, reducing computational costs to 6.9% of FLOPs while maintaining 99.1% of the original performance. Furthermore, we achieve a 2.28x reduction in Time-To-First-Token (TTFT) and a 1.32x acceleration in decoding throughput, highlighting the practical benefits of our integrated pruning approach for efficient video LLMs inference.
Authors: Yang Yang, Siming Zheng, Qirui Yang, Jinwei Chen, Boxi Wu, Xiaofei He, Deng Cai, Bo Li, Peng-Tao Jiang
Abstract: Diffusion models have recently emerged as powerful tools for camera simulation, enabling both geometric transformations and realistic optical effects. Among these, image-based bokeh rendering has shown promising results, but diffusion for video bokeh remains unexplored. Existing image-based methods are plagued by temporal flickering and inconsistent blur transitions, while current video editing methods lack explicit control over the focus plane and bokeh intensity. These issues limit their applicability for controllable video bokeh. In this work, we propose a one-step diffusion framework for generating temporally coherent, depth-aware video bokeh rendering. The framework employs a multi-plane image (MPI) representation adapted to the focal plane to condition the video diffusion model, thereby enabling it to exploit strong 3D priors from pretrained backbones. To further enhance temporal stability, depth robustness, and detail preservation, we introduce a progressive training strategy. Experiments on synthetic and real-world benchmarks demonstrate superior temporal coherence, spatial accuracy, and controllability, outperforming prior baselines. This work represents the first dedicated diffusion framework for video bokeh generation, establishing a new baseline for temporally coherent and controllable depth-of-field effects.
Authors: Yu Sheng, Jiajun Deng, Xinran Zhang, Yu Zhang, Bei Hua, Yanyong Zhang, Jianmin Ji
Abstract: A major breakthrough in 3D reconstruction is the feedforward paradigm to generate pixel-wise 3D points or Gaussian primitives from sparse, unposed images. To further incorporate semantics while avoiding the significant memory and storage costs of high-dimensional semantic features, existing methods extend this paradigm by associating each primitive with a compressed semantic feature vector. However, these methods have two major limitations: (a) the naively compressed feature compromises expressiveness, affecting the model's ability to capture fine-grained semantics, and (b) the pixel-wise primitive prediction introduces redundancy in overlapping areas, causing unnecessary memory overhead. To this end, we introduce \textbf{SpatialSplat}, a feedforward framework that produces redundancy-aware Gaussians and capitalizes on a dual-field semantic representation. Particularly, with the insight that primitives within the same instance exhibit high semantic consistency, we decompose the semantic representation into a coarse feature field that encodes uncompressed semantics with minimal primitives, and a fine-grained yet low-dimensional feature field that captures detailed inter-instance relationships. Moreover, we propose a selective Gaussian mechanism, which retains only essential Gaussians in the scene, effectively eliminating redundant primitives. Our proposed Spatialsplat learns accurate semantic information and detailed instances prior with more compact 3D Gaussians, making semantic 3D reconstruction more applicable. We conduct extensive experiments to evaluate our method, demonstrating a remarkable 60\% reduction in scene representation parameters while achieving superior performance over state-of-the-art methods. The code is available at https://github.com/shengyuuu/SpatialSplat.git
Authors: Julius Erbach, Dominik Narnhofer, Andreas Dombos, Bernt Schiele, Jan Eric Lenssen, Konrad Schindler
Abstract: Flow-based latent generative models such as Stable Diffusion 3 are able to generate images with remarkable quality, even enabling photorealistic text-to-image generation. Their impressive performance suggests that these models should also constitute powerful priors for inverse imaging problems, but that approach has not yet led to comparable fidelity. There are several key obstacles: (i) the data likelihood term is usually intractable; (ii) learned generative models cannot be directly conditioned on the distorted observations, leading to conflicting objectives between data likelihood and prior; and (iii) the reconstructions can deviate from the observed data. We present FLAIR, a novel, training-free variational framework that leverages flow-based generative models as prior for inverse problems. To that end, we introduce a variational objective for flow matching that is agnostic to the type of degradation, and combine it with deterministic trajectory adjustments to guide the prior towards regions which are more likely under the posterior. To enforce exact consistency with the observed data, we decouple the optimization of the data fidelity and regularization terms. Moreover, we introduce a time-dependent calibration scheme in which the strength of the regularization is modulated according to off-line accuracy estimates. Results on standard imaging benchmarks demonstrate that FLAIR consistently outperforms existing diffusion- and flow-based methods in terms of reconstruction quality and sample diversity. Our code is available at https://inverseflair.github.io/.
Authors: Ziyi Wu, Anil Kag, Ivan Skorokhodov, Willi Menapace, Ashkan Mirzaei, Igor Gilitschenski, Sergey Tulyakov, Aliaksandr Siarohin
Abstract: Direct Preference Optimization (DPO) has recently been applied as a post-training technique for text-to-video diffusion models. To obtain training data, annotators are asked to provide preferences between two videos generated from independent noise. However, this approach prohibits fine-grained comparisons, and we point out that it biases the annotators towards low-motion clips as they often contain fewer visual artifacts. In this work, we introduce DenseDPO, a method that addresses these shortcomings by making three contributions. First, we create each video pair for DPO by denoising corrupted copies of a ground truth video. This results in aligned pairs with similar motion structures while differing in local details, effectively neutralizing the motion bias. Second, we leverage the resulting temporal alignment to label preferences on short segments rather than entire clips, yielding a denser and more precise learning signal. With only one-third of the labeled data, DenseDPO greatly improves motion generation over vanilla DPO, while matching it in text alignment, visual quality, and temporal consistency. Finally, we show that DenseDPO unlocks automatic preference annotation using off-the-shelf Vision Language Models (VLMs): GPT accurately predicts segment-level preferences similar to task-specifically fine-tuned video reward models, and DenseDPO trained on these labels achieves performance close to using human labels.
Authors: Sharang Kaul, Mario Berk, Thiemo Gerbich, Abhinav Valada
Abstract: Finding reliable matches is essential in multi-object tracking to ensure the accuracy and reliability of perception systems in safety-critical applications such as autonomous vehicles. Effective matching mitigates perception errors, enhancing object identification and tracking for improved performance and safety. However, traditional metrics such as Intersection over Union (IoU) and Center Point Distances (CPDs), which are effective in 2D image planes, often fail to find critical matches in complex 3D scenes. To address this limitation, we introduce Contour Errors (CEs), an ego or object-centric metric for identifying matches of interest in tracking scenarios from a functional perspective. By comparing bounding boxes in the ego vehicle's frame, contour errors provide a more functionally relevant assessment of object matches. Extensive experiments on the nuScenes dataset demonstrate that contour errors improve the reliability of matches over the state-of-the-art 2D IoU and CPD metrics in tracking-by-detection methods. In 3D car tracking, our results show that Contour Errors reduce functional failures (FPs/FNs) by 80% at close ranges and 60% at far ranges compared to IoU in the evaluation stage.
Authors: Pingyu Wu, Kai Zhu, Yu Liu, Longxiang Tang, Jian Yang, Yansong Peng, Wei Zhai, Yang Cao, Zheng-Jun Zha
Abstract: Autoregressive image generation aims to predict the next token based on previous ones. However, this process is challenged by the bidirectional dependencies inherent in conventional image tokenizations, which creates a fundamental misalignment with the unidirectional nature of autoregressive models. To resolve this, we introduce AliTok, a novel Aligned Tokenizer that alters the dependency structure of the token sequence. AliTok employs a bidirectional encoder constrained by a causal decoder, a design that compels the encoder to produce a token sequence with both semantic richness and forward-dependency. Furthermore, by incorporating prefix tokens and employing a two-stage tokenizer training process to enhance reconstruction performance, AliTok achieves high fidelity and predictability simultaneously. Building upon AliTok, a standard decoder-only autoregressive model with just 177M parameters achieves a gFID of 1.44 and an IS of 319.5 on the ImageNet-256 benchmark. Scaling up to 662M parameters, our model reaches a gFID of 1.28, surpassing the state-of-the-art diffusion method while achieving a 10x faster sampling speed. The code and weights are available at https://github.com/ali-vilab/alitok.
Authors: Lianming Huang, Haibo Hu, Yufei Cui, Jiacheng Zuo, Shangyu Wu, Nan Guan, Chun Jason Xue
Abstract: With the rapid advancement of autonomous driving, deploying Vision-Language Models (VLMs) to enhance perception and decision-making has become increasingly common. However, the real-time application of VLMs is hindered by high latency and computational overhead, limiting their effectiveness in time-critical driving scenarios. This challenge is particularly evident when VLMs exhibit over-inference, continuing to process unnecessary layers even after confident predictions have been reached. To address this inefficiency, we propose AD-EE, an Early Exit framework that incorporates domain characteristics of autonomous driving and leverages causal inference to identify optimal exit layers. We evaluate our method on large-scale real-world autonomous driving datasets, including Waymo and the corner-case-focused CODA, as well as on a real vehicle running the Autoware Universe platform. Extensive experiments across multiple VLMs show that our method significantly reduces latency, with maximum improvements reaching up to 57.58%, and enhances object detection accuracy, with maximum gains of up to 44%.
Authors: Yeongtak Oh, Dohyun Chung, Juhyeon Shin, Sangha Park, Johan Barthelemy, Jisoo Mok, Sungroh Yoon
Abstract: Recent multi-modal large language models (MLLMs) often struggle to generate personalized image captions, even when trained on high-quality captions. In this work, we observe that such limitations persist in existing post-training-based MLLM personalization methods. Specifically, despite being post-tuned with large-scale caption data through supervised fine-tuning (SFT), these models frequently fail to produce faithful descriptions in real-world scenarios, such as multi-concept image captioning. However, acquiring large-scale, high-quality captions for such complex settings is both costly and difficult. To address the data-centric nature of SFT, we propose a reinforcement learning (RL)-based post-training framework. To the best of our knowledge, this is the first RL-based approach to post-train MLLMs for personalized image captioning. Our method significantly enhances both visual recognition and personalized generation capabilities of MLLMs, and consistently outperforms existing SFT-based baselines, especially in the challenging multi-concept image captioning task. Project page: https://github.com/oyt9306/RePIC
Authors: Tahereh Zarrat Ehsan, Michael Tangermann, Ya\u{g}mur G\"u\c{c}l\"ut\"urk, Bastiaan R. Bloem, Luc J. W. Evers
Abstract: Accurately quantifying motor characteristics in Parkinson disease (PD) is crucial for monitoring disease progression and optimizing treatment strategies. The finger-tapping test is a standard motor assessment. Clinicians visually evaluate a patient's tapping performance and assign an overall severity score based on tapping amplitude, speed, and irregularity. However, this subjective evaluation is prone to inter- and intra-rater variability, and does not offer insights into individual motor characteristics captured during this test. This paper introduces a granular computer vision-based method for quantifying PD motor characteristics from video recordings. Four sets of clinically relevant features are proposed to characterize hypokinesia, bradykinesia, sequence effect, and hesitation-halts. We evaluate our approach on video recordings and clinical evaluations of 74 PD patients from the Personalized Parkinson Project. Principal component analysis with varimax rotation shows that the video-based features corresponded to the four deficits. Additionally, video-based analysis has allowed us to identify further granular distinctions within sequence effect and hesitation-halts deficits. In the following, we have used these features to train machine learning classifiers to estimate the Movement Disorder Society Unified Parkinson Disease Rating Scale (MDS-UPDRS) finger-tapping score. Compared to state-of-the-art approaches, our method achieves a higher accuracy in MDS-UPDRS score prediction, while still providing an interpretable quantification of individual finger-tapping motor characteristics. In summary, the proposed framework provides a practical solution for the objective assessment of PD motor characteristics, that can potentially be applied in both clinical and remote settings. Future work is needed to assess its responsiveness to symptomatic treatment and disease progression.
Authors: Lixuan He, Haoyu Dong, Zhenxing Chen, Yangcheng Yu, Jie Feng, Yong Li
Abstract: Vision-and-Language Navigation (VLN) in large-scale urban environments requires embodied agents to ground linguistic instructions in complex scenes and recall relevant experiences over extended time horizons. Prior modular pipelines offer interpretability but lack unified memory, while end-to-end (M)LLM agents excel at fusing vision and language yet remain constrained by fixed context windows and implicit spatial reasoning. We introduce \textbf{Mem4Nav}, a hierarchical spatial-cognition long-short memory system that can augment any VLN backbone. Mem4Nav fuses a sparse octree for fine-grained voxel indexing with a semantic topology graph for high-level landmark connectivity, storing both in trainable memory tokens embedded via a reversible Transformer. Long-term memory (LTM) compresses and retains historical observations at both octree and graph nodes, while short-term memory (STM) caches recent multimodal entries in relative coordinates for real-time obstacle avoidance and local planning. At each step, STM retrieval sharply prunes dynamic context, and, when deeper history is needed, LTM tokens are decoded losslessly to reconstruct past embeddings. Evaluated on Touchdown and Map2Seq across three backbones (modular, state-of-the-art VLN with prompt-based LLM, and state-of-the-art VLN with strided-attention MLLM), Mem4Nav yields 7-13 pp gains in Task Completion, sufficient SPD reduction, and >10 pp nDTW improvement. Ablations confirm the indispensability of both the hierarchical map and dual memory modules. Our codes are open-sourced via https://github.com/tsinghua-fib-lab/Mem4Nav.
Authors: Chen Sun, Haiyang Sun, Zhiqing Guo, Yunfeng Diao, Liejun Wang, Dan Ma, Gaobo Yang, Keqin Li
Abstract: Deepfakes pose significant security and privacy threats through malicious facial manipulations. While robust watermarking can aid in authenticity verification and source tracking, existing methods often lack the sufficient robustness against Deepfake manipulations. Diffusion models have demonstrated remarkable performance in image generation, enabling the seamless fusion of watermark with image during generation. In this study, we propose a novel robust watermarking framework based on diffusion model, called DiffMark. By modifying the training and sampling scheme, we take the facial image and watermark as conditions to guide the diffusion model to progressively denoise and generate corresponding watermarked image. In the construction of facial condition, we weight the facial image by a timestep-dependent factor that gradually reduces the guidance intensity with the decrease of noise, thus better adapting to the sampling process of diffusion model. To achieve the fusion of watermark condition, we introduce a cross information fusion (CIF) module that leverages a learnable embedding table to adaptively extract watermark features and integrates them with image features via cross-attention. To enhance the robustness of the watermark against Deepfake manipulations, we integrate a frozen autoencoder during training phase to simulate Deepfake manipulations. Additionally, we introduce Deepfake-resistant guidance that employs specific Deepfake model to adversarially guide the diffusion sampling process to generate more robust watermarked images. Experimental results demonstrate the effectiveness of the proposed DiffMark on typical Deepfakes. Our code will be available at https://github.com/vpsg-research/DiffMark.
Authors: Mohsi Jawaid, Marcus M\"artens, Tat-Jun Chin
Abstract: Spacecraft pose estimation is crucial for autonomous in-space operations, such as rendezvous, docking and on-orbit servicing. Vision-based pose estimation methods, which typically employ RGB imaging sensors, is a compelling solution for spacecraft pose estimation, but are challenged by harsh lighting conditions, which produce imaging artifacts such as glare, over-exposure, blooming and lens flare. Due to their much higher dynamic range, neuromorphic or event sensors are more resilient to extreme lighting conditions. However, event sensors generally have lower spatial resolution and suffer from reduced signal-to-noise ratio during periods of low relative motion. This work addresses these individual sensor limitations by introducing a sensor fusion approach combining RGB and event sensors. A beam-splitter prism was employed to achieve precise optical and temporal alignment. Then, a RANSAC-based technique was developed to fuse the information from the RGB and event channels to achieve pose estimation that leveraged the strengths of the two modalities. The pipeline was complemented by dropout uncertainty estimation to detect extreme conditions that affect either channel. To benchmark the performance of the proposed event-RGB fusion method, we collected a comprehensive real dataset of RGB and event data for satellite pose estimation in a laboratory setting under a variety of challenging illumination conditions. Encouraging results on the dataset demonstrate the efficacy of our event-RGB fusion approach and further supports the usage of event sensors for spacecraft pose estimation. To support community research on this topic, our dataset has been released publicly.
Authors: Sybelle Goedicke-Fritz (Department of General Pediatrics and Neonatology, Saarland University, Campus Homburg, Homburg/Saar, Germany), Michelle Bous (Department of General Pediatrics and Neonatology, Saarland University, Campus Homburg, Homburg/Saar, Germany), Annika Engel (Chair for Clinical Bioinformatics, Saarland Informatics Campus, Saarland University, Saarbr\"ucken, Germany), Matthias Flotho (Chair for Clinical Bioinformatics, Saarland Informatics Campus, Saarland University, Saarbr\"ucken, Germany, Helmholtz Institute for Pharmaceutical Research Saarland), Pascal Hirsch (Chair for Clinical Bioinformatics, Saarland Informatics Campus, Saarland University, Saarbr\"ucken, Germany), Hannah Wittig (Department of General Pediatrics and Neonatology, Saarland University, Campus Homburg, Homburg/Saar, Germany), Dino Milanovic (Chair for Clinical Bioinformatics, Saarland Informatics Campus, Saarland University, Saarbr\"ucken, Germany), Dominik Mohr (Department of General Pediatrics and Neonatology, Saarland University, Campus Homburg, Homburg/Saar, Germany), Mathias Kaspar (Digital Medicine, University Hospital of Augsburg, Augsburg, Germany), Sogand Nemat (Department of Radiology, and Interventional Radiology, University Hospital of Saarland, Homburg, Germany), Dorothea Kerner (Department of Radiology, and Interventional Radiology, University Hospital of Saarland, Homburg, Germany), Arno B\"ucker (Department of Radiology, and Interventional Radiology, University Hospital of Saarland, Homburg, Germany), Andreas Keller (Chair for Clinical Bioinformatics, Saarland Informatics Campus, Saarland University, Saarbr\"ucken, Germany, Helmholtz Institute for Pharmaceutical Research Saarland, Pharma Science Hub), Sascha Meyer (Clinical Centre Karlsruhe, Franz-Lust Clinic for Paediatrics, Karlsruhe, Germany), Michael Zemlin (Department of General Pediatrics and Neonatology, Saarland University, Campus Homburg, Homburg/Saar, Germany), Philipp Flotho (Chair for Clinical Bioinformatics, Saarland Informatics Campus, Saarland University, Saarbr\"ucken, Germany, Helmholtz Institute for Pharmaceutical Research Saarland)
Abstract: Bronchopulmonary dysplasia (BPD) is a chronic lung disease affecting 35% of extremely low birth weight infants. Defined by oxygen dependence at 36 weeks postmenstrual age, it causes lifelong respiratory complications. However, preventive interventions carry severe risks, including neurodevelopmental impairment, ventilator-induced lung injury, and systemic complications. Therefore, early BPD prognosis and prediction of BPD outcome is crucial to avoid unnecessary toxicity in low risk infants. Admission radiographs of extremely preterm infants are routinely acquired within 24h of life and could serve as a non-invasive prognostic tool. In this work, we developed and investigated a deep learning approach using chest X-rays from 163 extremely low-birth-weight infants ($\leq$32 weeks gestation, 401-999g) obtained within 24 hours of birth. We fine-tuned a ResNet-50 pretrained specifically on adult chest radiographs, employing progressive layer freezing with discriminative learning rates to prevent overfitting and evaluated a CutMix augmentation and linear probing. For moderate/severe BPD outcome prediction, our best performing model with progressive freezing, linear probing and CutMix achieved an AUROC of 0.78 $\pm$ 0.10, balanced accuracy of 0.69 $\pm$ 0.10, and an F1-score of 0.67 $\pm$ 0.11. In-domain pre-training significantly outperformed ImageNet initialization (p = 0.031) which confirms domain-specific pretraining to be important for BPD outcome prediction. Routine IRDS grades showed limited prognostic value (AUROC 0.57 $\pm$ 0.11), confirming the need of learned markers. Our approach demonstrates that domain-specific pretraining enables accurate BPD prediction from routine day-1 radiographs. Through progressive freezing and linear probing, the method remains computationally feasible for site-level implementation and future federated learning deployments.
Authors: Zhekai Chen, Ruihang Chu, Yukang Chen, Shiwei Zhang, Yujie Wei, Yingya Zhang, Xihui Liu
Abstract: Scaling visual generation models is essential for real-world content creation, yet requires substantial training and computational expenses. Alternatively, test-time scaling has garnered growing attention due to resource efficiency and promising performance. In this work, we present TTS-VAR, the first general test-time scaling framework for visual auto-regressive (VAR) models, modeling the generation process as a path searching problem. To dynamically balance computational efficiency with exploration capacity, we first introduce an adaptive descending batch size schedule throughout the causal generation process. Besides, inspired by VAR's hierarchical coarse-to-fine multi-scale generation, our framework integrates two key components: (i) At coarse scales, we observe that generated tokens are hard for evaluation, possibly leading to erroneous acceptance of inferior samples or rejection of superior samples. Noticing that the coarse scales contain sufficient structural information, we propose clustering-based diversity search. It preserves structural variety through semantic feature clustering, enabling later selection on samples with higher potential. (ii) In fine scales, resampling-based potential selection prioritizes promising candidates using potential scores, which are defined as reward functions incorporating multi-scale generation history. Experiments on the powerful VAR model Infinity show a notable 8.7% GenEval score improvement (from 0.69 to 0.75). Key insights reveal that early-stage structural features effectively influence final quality, and resampling efficacy varies across generation scales. Code is available at https://github.com/ali-vilab/TTS-VAR.
Authors: Fevziye Irem Eyiokur, Dogucan Yaman, Haz{\i}m Kemal Ekenel, Alexander Waibel
Abstract: We address the problem of Embodied Reference Understanding, which involves predicting the object that a person in the scene is referring to through both pointing gesture and language. Accurately identifying the referent requires multimodal understanding: integrating textual instructions, visual pointing, and scene context. However, existing methods often struggle to effectively leverage visual clues for disambiguation. We also observe that, while the referent is often aligned with the head-to-fingertip line, it occasionally aligns more closely with the wrist-to-fingertip line. Therefore, relying on a single line assumption can be overly simplistic and may lead to suboptimal performance. To address this, we propose a dual-model framework, where one model learns from the head-to-fingertip direction and the other from the wrist-to-fingertip direction. We further introduce a Gaussian ray heatmap representation of these lines and use them as input to provide a strong supervisory signal that encourages the model to better attend to pointing cues. To combine the strengths of both models, we present the CLIP-Aware Pointing Ensemble module, which performs a hybrid ensemble based on CLIP features. Additionally, we propose an object center prediction head as an auxiliary task to further enhance referent localization. We validate our approach through extensive experiments and analysis on the benchmark YouRefIt dataset, achieving an improvement of approximately 4 mAP at the 0.25 IoU threshold. We further evaluate our approach on the CAESAR and ISL Pointing datasets.
Authors: Hao Tang, Chenwei Xie, Xiaoyi Bao, Tingyu Weng, Pandeng Li, Yun Zheng, Liwei Wang
Abstract: In this paper, we propose UniLIP, a unified framework that adapts CLIP for multimodal understanding, generation and editing. Although CLIP excels at understanding, it lacks reconstruction abilities required to be a unified visual encoder. However, previous CLIP-based unified methods fail to balance understanding and reconstruction, leading to semantic degradation or inconsistent reconstructions. In contrast, we introduce a novel two-stage training scheme with a self-distillation strategy that progressively endows CLIP with high-fidelity reconstruction abilities while preserving its original comprehension performance. For enhanced reasoning and consistency in generation and editing, we further develop a dual-condition architecture built upon the MetaQuery framework. Our architecture jointly utilizes multimodal hidden states for rich contextual details and learnable query embeddings to harness the powerful reasoning abilities of Multimodal Large Language Models (MLLMs). Leveraging advanced image representation and architectural design, UniLIP demonstrates superior instruction following and edit fidelity. With only 1B and 3B parameters, UniLIP can outperform larger unified models such as BAGEL (7B) and Uniworld-V1 (12B), achieving state-of-the-art performance of 0.90 on GenEval, 0.63 on WISE, and 3.94 on ImgEdit. These results demonstrate that UniLIP successfully expands the application of CLIP, establishing its continuous features to not only serve as the optimal choice for understanding tasks but also achieve highly competitive performance in generation and editing tasks. Code and models are available at https://github.com/nnnth/UniLIP.
Authors: Ning Zhu, Xiaochuan Ma, Shaoting Zhang, Guotai Wang
Abstract: Cold-Start Active Learning (CSAL) aims to select informative samples for annotation without prior knowledge, which is important for improving annotation efficiency and model performance under a limited annotation budget in medical image analysis. Most existing CSAL methods rely on Self-Supervised Learning (SSL) on the target dataset for feature extraction, which is inefficient and limited by insufficient feature representation. Recently, pre-trained Foundation Models (FMs) have shown powerful feature extraction ability with a potential for better CSAL. However, this paradigm has been rarely investigated, with a lack of benchmarks for comparison of FMs in CSAL tasks. To this end, we propose MedCAL-Bench, the first systematic FM-based CSAL benchmark for medical image analysis. We evaluate 14 FMs and 7 CSAL strategies across 7 datasets under different annotation budgets, covering classification and segmentation tasks from diverse medical modalities. It is also the first CSAL benchmark that evaluates both the feature extraction and sample selection stages. Our experimental results reveal that: 1) Most FMs are effective feature extractors for CSAL, with DINO family performing the best in segmentation; 2) The performance differences of these FMs are large in segmentation tasks, while small for classification; 3) Different sample selection strategies should be considered in CSAL on different datasets, with Active Learning by Processing Surprisal (ALPS) performing the best in segmentation while RepDiv leading for classification. The code is available at https://github.com/HiLab-git/MedCAL-Bench.
Authors: Lefei Shen, Mouxiang Chen, Xu Liu, Han Fu, Xiaoxue Ren, Jianling Sun, Zhuo Li, Chenghao Liu
Abstract: Recent studies have indicated that vision models pre-trained on images can serve as time series foundation models (TSFMs) by reformulating time series forecasting (TSF) as image reconstruction. However, effective cross-modal transfer from vision to time series remains challenging due to three discrepancies: (1) the data-modality gap between structured, bounded image data and unbounded, heterogeneous time series; (2) the multivariate-forecasting gap between fixed RGB-three-channel vision models and time series with arbitrary numbers of variates; and (3) the probabilistic-forecasting gap between the deterministic outputs of vision models and the requirement for uncertainty-aware probabilistic predictions. To bridge these gaps, we propose VisonTS++, a TSFM based on continual pre-training of a vision model on large-scale time series. Our approach introduces three key innovations: (1) vision-model-based filtering to identify high-quality sequences to stabilize pre-training and mitigate modality gap; (2) colorized multivariate conversion, encoding multivariate series as multi-subfigure RGB images to enhance cross-variate modeling; (3) multi-quantile forecasting, using parallel reconstruction heads to generate quantile forecasts without parametric assumptions. Experiments show that VisionTS++ achieves state-of-the-art performance in both in-distribution and out-of-distribution forecasting, outperforming specialized TSFMs by 6%-44% in MSE reduction and ranking first in GIFT-Eval benchmark which comprises 23 datasets across 7 domains. Our work demonstrates that with appropriate adaptation, vision models can effectively generalize to TSF, thus advancing the pursuit of universal TSFMs. Code is available at https://github.com/HALF111/VisionTSpp.
Authors: Ke Ma, Jun Long, Hongxiao Fei, Liujie Hua, Zhen Dai, Yueyi Luo
Abstract: Pre-trained Vision-Language Models (VLMs) struggle with Zero-Shot Anomaly Detection (ZSAD) due to a critical adaptation gap: they lack the local inductive biases required for dense prediction and employ inflexible feature fusion paradigms. We address these limitations through an Architectural Co-Design framework that jointly refines feature representation and cross-modal fusion. Our method proposes a parameter-efficient Convolutional Low-Rank Adaptation (Conv-LoRA) adapter to inject local inductive biases for fine-grained representation, and introduces a Dynamic Fusion Gateway (DFG) that leverages visual context to adaptively modulate text prompts, enabling a powerful bidirectional fusion. Extensive experiments on diverse industrial and medical benchmarks demonstrate superior accuracy and robustness, validating that this synergistic co-design is critical for robustly adapting foundation models to dense perception tasks. The source code is available at https://github.com/cockmake/ACD-CLIP.
Authors: Peng Li, Yisheng He, Yingdong Hu, Yuan Dong, Weihao Yuan, Yuan Liu, Siyu Zhu, Gang Cheng, Zilong Dong, Yike Guo
Abstract: We present a feed-forward framework for Gaussian full-head synthesis from a single unposed image. Unlike previous work that relies on time-consuming GAN inversion and test-time optimization, our framework can reconstruct the Gaussian full-head model given a single unposed image in a single forward pass. This enables fast reconstruction and rendering during inference. To mitigate the lack of large-scale 3D head assets, we propose a large-scale synthetic dataset from trained 3D GANs and train our framework using only synthetic data. For efficient high-fidelity generation, we introduce a coarse-to-fine Gaussian head generation pipeline, where sparse points from the FLAME model interact with the image features by transformer blocks for feature extraction and coarse shape reconstruction, which are then densified for high-fidelity reconstruction. To fully leverage the prior knowledge residing in pretrained 3D GANs for effective reconstruction, we propose a dual-branch framework that effectively aggregates the structured spherical triplane feature and unstructured point-based features for more effective Gaussian head reconstruction. Experimental results show the effectiveness of our framework towards existing work. Project page at: https://panolam.github.io/.
Authors: Heeji Yoon, Jaewoo Jung, Junwan Kim, Hyungyu Choi, Heeseong Shin, Sangbeom Lim, Honggyu An, Chaehyun Kim, Jisang Han, Donghyun Kim, Chanho Eom, Sunghwan Hong, Seungryong Kim
Abstract: Multimodal large language models (MLLMs) trained with visual instruction tuning have achieved strong performance across diverse tasks, yet they remain limited in vision-centric tasks such as object counting or spatial reasoning. We attribute this gap to the prevailing text-only supervision paradigm, which provides only indirect guidance for the visual pathway and often leads MLLMs to discard fine-grained visual details during training. In this paper, we present VIsual Representation ALignment (VIRAL), a simple yet effective regularization strategy that aligns the internal visual representations of MLLMs with those of pre-trained vision foundation models (VFMs). By explicitly enforcing this alignment, VIRAL enables the model not only to retain critical visual details from the input vision encoder but also to complement additional visual knowledge from VFMs, thereby enhancing its ability to reason over complex visual inputs. Our experiments demonstrate consistent improvements across all tasks on widely adopted multimodal benchmarks. Furthermore, we conduct comprehensive ablation studies to validate the key design choices underlying our framework. We believe this simple finding opens up an important direction for the effective integration of visual information in training MLLMs.
Authors: Zhiyuan Yan, Kaiqing Lin, Zongjian Li, Junyan Ye, Hui Han, Zhendong Wang, Hao Liu, Bin Lin, Hao Li, Xue Xu, Xinyan Xiao, Jingdong Wang, Haifeng Wang, Li Yuan
Abstract: The pursuit of unified multimodal models (UMMs) has long been hindered by a fundamental schism between multimodal understanding and generation. Current approaches typically disentangle the two and treat them as separate endeavors with disjoint objectives, missing the mutual benefits. We argue that true unification requires more than just merging two tasks. It requires a unified, foundational objective that intrinsically links them. In this paper, we introduce an insightful paradigm through the Auto-Encoder lens, i.e., regarding understanding as the encoder (I2T) that compresses images into text, and generation as the decoder (T2I) that reconstructs images from that text. To implement this, we propose UAE, where we begin by pre-training the decoder with the proposed 700k long-context image-caption pairs to direct it to "understand" the fine-grained and complex semantics from the text. We then propose Unified-GRPO via reinforcement learning (RL) to unify the two, which covers two complementary stages: (1) Generation for Understanding, where the encoder is trained to generate informative captions that maximize the decoder's reconstruction quality, enhancing its visual perception; (2) Understanding for Generation, where the decoder is refined to reconstruct from these captions, forcing it to leverage every detail and improving its long-context instruction following and generation fidelity. Our empirical results suggest that understanding can largely enhance generation (verified on GenEval), while generation, in turn, notably strengthens fine-grained visual perception like small object and color recognition (verified on MMT-Bench). This bidirectional improvement reveals a deep synergy: under the unified reconstruction objective, generation and understanding can mutually benefit each other, moving closer to truly unified multimodal intelligence.
Authors: Jing Huang, Zhiya Tan, Shutao Gong, Fanwei Zeng, Joey Tianyi Zhou, Changtao Miao, Huazhe Tan, Weibin Yao, Jianshu Li
Abstract: As large vision language models (VLMs) advance, their capabilities in multilingual visual question answering (mVQA) have significantly improved. Chain-of-thought (CoT) reasoning has been proven to enhance interpretability and complex reasoning. However, most existing approaches rely primarily on textual CoT and provide limited support for multilingual multimodal reasoning, constraining their deployment in real-world applications. To address this gap, we introduce LaV-CoT, the first Language-aware Visual CoT framework with Multi-Aspect Reward Optimization. LaV-CoT incorporates an interpretable multi-stage reasoning pipeline consisting of Text Summary with Bounding Box (BBox), Language Identification, Spatial Object-level Captioning, and Step-by-step Logical Reasoning. Following this reasoning pipeline, we design an automated data curation method that generates multilingual CoT annotations through iterative generation, correction, and refinement, enabling scalable and high-quality training data. To improve reasoning and generalization, LaV-CoT adopts a two-stage training paradigm combining Supervised Fine-Tuning (SFT) with Language-aware Group Relative Policy Optimization (GRPO), guided by verifiable multi-aspect rewards including language consistency, structural accuracy, and semantic alignment. Extensive evaluations on public datasets including MMMB, Multilingual MMBench, and MTVQA show that LaV-CoT achieves up to ~9.5% accuracy improvements over open-source baselines of similar size and even surpasses models with 2$\times$ larger scales by ~2.6%. Moreover, LaV-CoT outperforms advanced proprietary models such as GPT-4o-0513 and Gemini-2.5-flash. We further conducted an online A/B test to validate our method on real-world data, highlighting its effectiveness for industrial deployment. Our code is available at this link: https://github.com/HJNVR/LaV-CoT
Authors: Ayhan Can Erdur, Christian Beischl, Daniel Scholz, Jiazhen Pan, Benedikt Wiestler, Daniel Rueckert, Jan C Peeken
Abstract: Missing input sequences are common in medical imaging data, posing a challenge for deep learning models reliant on complete input data. In this work, inspired by MultiMAE [2], we develop a masked autoencoder (MAE) paradigm for multi-modal, multi-task learning in 3D medical imaging with brain MRIs. Our method treats each MRI sequence as a separate input modality, leveraging a late-fusion-style transformer encoder to integrate multi-sequence information (multi-modal) and individual decoder streams for each modality for multi-task reconstruction. This pretraining strategy guides the model to learn rich representations per modality while also equipping it to handle missing inputs through cross-sequence reasoning. The result is a flexible and generalizable encoder for brain MRIs that infers missing sequences from available inputs and can be adapted to various downstream applications. We demonstrate the performance and robustness of our method against an MAE-ViT baseline in downstream segmentation and classification tasks, showing absolute improvement of $10.1$ overall Dice score and $0.46$ MCC over the baselines with missing input sequences. Our experiments demonstrate the strength of this pretraining strategy. The implementation is made available.
Authors: Elena Camuffo, Francesco Barbato, Mete Ozay, Simone Milani, Umberto Michieli
Abstract: We introduce MOCHA (Multi-modal Objects-aware Cross-arcHitecture Alignment), a knowledge distillation approach that transfers region-level multimodal semantics from a large vision-language teacher (e.g., LLaVa) into a lightweight vision-only object detector student (e.g., YOLO). A translation module maps student features into a joint space, where the training of the student and translator is guided by a dual-objective loss that enforces both local alignment and global relational consistency. Unlike prior approaches focused on dense or global alignment, MOCHA operates at the object level, enabling efficient transfer of semantics without modifying the teacher or requiring textual input at inference. We validate our method across four personalized detection benchmarks under few-shot regimes. Results show consistent gains over baselines, with a +10.1 average score improvement. Despite its compact architecture, MOCHA reaches performance on par with larger multimodal models, proving its suitability for real-world deployment.
Authors: Yue Cao, Quansong He, Kaishen Wang, Jianlong Xiong, Tao He
Abstract: U-like networks have become fundamental frameworks in medical image segmentation through skip connections that bridge high-level semantics and low-level spatial details. Despite their success, conventional skip connections exhibit two key limitations: inter-feature constraints and intra-feature constraints. The inter-feature constraint refers to the static nature of feature fusion in traditional skip connections, where information is transmitted along fixed pathways regardless of feature content. The intra-feature constraint arises from the insufficient modeling of multi-scale feature interactions, thereby hindering the effective aggregation of global contextual information. To overcome these limitations, we propose a novel Dynamic Skip Connection (DSC) block that fundamentally enhances cross-layer connectivity through adaptive mechanisms. The DSC block integrates two complementary components. (1) Test-Time Training (TTT) module. This module addresses the inter-feature constraint by enabling dynamic adaptation of hidden representations during inference, facilitating content-aware feature refinement. (2) Dynamic Multi-Scale Kernel (DMSK) module. To mitigate the intra-feature constraint, this module adaptively selects kernel sizes based on global contextual cues, enhancing the network capacity for multi-scale feature integration. The DSC block is architecture-agnostic and can be seamlessly incorporated into existing U-like network structures. Extensive experiments demonstrate the plug-and-play effectiveness of the proposed DSC block across CNN-based, Transformer-based, hybrid CNN-Transformer, and Mamba-based U-like networks.
Authors: Paul Julius K\"uhn, Duc Anh Nguyen, Arjan Kuijper, Holger Graf, Dieter Fellner, Saptarshi Neil Sinha
Abstract: Point cloud segmentation is central to autonomous driving and 3D scene understanding. While voxel- and point-based methods dominate recent research due to their compatibility with deep architectures and ability to capture fine-grained geometry, they often incur high computational cost, irregular memory access, and limited real-time efficiency. In contrast, range-view methods, though relatively underexplored - can leverage mature 2D semantic segmentation techniques for fast and accurate predictions. Motivated by the rapid progress in Visual Foundation Models (VFMs) for captioning, zero-shot recognition, and multimodal tasks, we investigate whether SAM2, the current state-of-the-art VFM for segmentation tasks, can serve as a strong backbone for LiDAR point cloud segmentation in the range view. We present , to our knowledge, the first range-view framework that adapts SAM2 to 3D segmentation, coupling efficient 2D feature extraction with standard projection/back-projection to operate on point clouds. To optimize SAM2 for range-view representations, we implement several architectural modifications to the encoder: (1) a novel module that emphasizes horizontal spatial dependencies inherent in LiDAR range images, (2) a customized configuration of tailored to the geometric properties of spherical projections, and (3) an adapted mechanism in the encoder backbone specifically designed to capture the unique spatial patterns and discontinuities present in range-view pseudo-images. Our approach achieves competitive performance on SemanticKITTI while benefiting from the speed, scalability, and deployment simplicity of 2D-centric pipelines. This work highlights the viability of VFMs as general-purpose backbones for 3D perception and opens a path toward unified, foundation-model-driven LiDAR segmentation. Results lets us conclude that range-view segmentation methods using VFMs leads to promising results.
Authors: William Muckelroy III, Mohammed Alsakabi, John Dolan, Ozan Tonguz
Abstract: LiDAR's dense, sharp point cloud (PC) representations of the surrounding environment enable accurate perception and significantly improve road safety by offering greater scene awareness and understanding. However, LiDAR's high cost continues to restrict the broad adoption of high-level Autonomous Driving (AD) systems in commercially available vehicles. Prior research has shown progress towards circumventing the need for LiDAR by training a neural network, using LiDAR point clouds as ground truth (GT), to produce LiDAR-like 3D point clouds using only 4D Radars. One of the best examples is a neural network created to train a more efficient radar target detector with a modular 2D convolutional neural network (CNN) backbone and a temporal coherence network at its core that uses the RaDelft dataset for training (see arXiv:2406.04723). In this work, we investigate the impact of higher-capacity segmentation backbones on the quality of the produced point clouds. Our results show that while very high-capacity models may actually hurt performance, an optimal segmentation backbone can provide a 23.7% improvement over the state-of-the-art (SOTA).
Authors: Dwip Dalal, Gautam Vashishtha, Anku Rani, Aishwarya Reganti, Parth Patwa, Mohd Sarique, Chandan Gupta, Keshav Nath, Viswanatha Reddy, Vinija Jain, Aman Chadha, Amitava Das, Amit Sheth, Asif Ekbal
Abstract: The rise in harmful online content not only distorts public discourse but also poses significant challenges to maintaining a healthy digital environment. In response to this, we introduce a multimodal dataset uniquely crafted for identifying hate in digital content. Central to our methodology is the innovative application of watermarked, stability-enhanced, stable diffusion techniques combined with the Digital Attention Analysis Module (DAAM). This combination is instrumental in pinpointing the hateful elements within images, thereby generating detailed hate attention maps, which are used to blur these regions from the image, thereby removing the hateful sections of the image. We release this data set as a part of the dehate shared task. This paper also describes the details of the shared task. Furthermore, we present DeHater, a vision-language model designed for multimodal dehatification tasks. Our approach sets a new standard in AI-driven image hate detection given textual prompts, contributing to the development of more ethical AI applications in social media.
Authors: Zikun Guo, Xinyue Xu, Pei Xiang, Shu Yang, Xin Han, Di Wang, Lijie Hu
Abstract: Vision language models(VLMs) are increasingly integrated into clinical workflows, but they often exhibit sycophantic behavior prioritizing alignment with user phrasing social cues or perceived authority over evidence based reasoning. This study evaluate clinical sycophancy in medical visual question answering through a novel clinically grounded benchmark. We propose a medical sycophancy dataset construct from PathVQA, SLAKE, and VQA-RAD stratified by different type organ system and modality. Using psychologically motivated pressure templates including various sycophancy. In our adversarial experiments on various VLMs, we found that these models are generally vulnerable, exhibiting significant variations in the occurrence of adversarial responses, with weak correlations to the model accuracy or size. Imitation and expert provided corrections were found to be the most effective triggers, suggesting that the models possess a bias mechanism independent of visual evidence. To address this, we propose Visual Information Purification for Evidence based Response (VIPER) a lightweight mitigation strategy that filters non evidentiary content for example social pressures and then generates constrained evidence first answers. This framework reduces sycophancy by an average amount outperforming baselines while maintaining interpretability. Our benchmark analysis and mitigation framework lay the groundwork for robust deployment of medical VLMs in real world clinician interactions emphasizing the need for evidence anchored defenses.
Authors: Mingyu Kang, Yong Suk Choi
Abstract: Text-to-image diffusion models have achieved remarkable success in generating high-quality and diverse images. Building on these advancements, diffusion models have also demonstrated exceptional performance in text-guided image editing. A key strategy for effective image editing involves inverting the source image into editable noise maps associated with the target image. However, previous inversion methods face challenges in adhering closely to the target text prompt. The limitation arises because inverted noise maps, while enabling faithful reconstruction of the source image, restrict the flexibility needed for desired edits. To overcome this issue, we propose Editable Noise Map Inversion (ENM Inversion), a novel inversion technique that searches for optimal noise maps to ensure both content preservation and editability. We analyze the properties of noise maps for enhanced editability. Based on this analysis, our method introduces an editable noise refinement that aligns with the desired edits by minimizing the difference between the reconstructed and edited noise maps. Extensive experiments demonstrate that ENM Inversion outperforms existing approaches across a wide range of image editing tasks in both preservation and edit fidelity with target prompts. Our approach can also be easily applied to video editing, enabling temporal consistency and content manipulation across frames.
Authors: Xiangrui Liu, Minghao Qin, Yan Shu, Zhengyang Liang, Yang Tian, Chen Jason Zhang, Bo Zhao, Zheng Liu
Abstract: Identifying key moments in long videos is essential for downstream understanding and reasoning tasks. In this paper, we introduce a new problem, Taskoriented Temporal Grounding ToTG, which aims to localize time intervals containing the necessary information based on a task's natural description. Along with the definition, we also present ToTG Bench, a comprehensive benchmark for evaluating the performance on ToTG. ToTG is particularly challenging for traditional approaches due to their limited generalizability and difficulty in handling long videos. To address these challenges, we propose TimeScope, a novel framework built upon progressive reasoning. TimeScope first identifies a coarse-grained temporal scope in the long video that likely contains the key moments, and then refines this scope through finegrained moment partitioning. Additionally, we curate a highquality dataset, namely ToTG Pile, to enhance TimeScope's ability to perform progressive temporal grounding effectively. Extensive experiments demonstrate that TimeScope consistently outperforms both existing temporalgrounding methods and popular MLLMs across various settings, highlighting its effectiveness in addressing this new challenging problem.
Authors: Xin Zou, Di Lu, Yizhou Wang, Yibo Yan, Yuanhuiyi Lyu, Xu Zheng, Linfeng Zhang, Xuming Hu
Abstract: Despite their powerful capabilities, Multimodal Large Language Models (MLLMs) suffer from considerable computational overhead due to their reliance on massive visual tokens. Recent studies have explored token pruning to alleviate this problem, which typically uses text-vision cross-attention or [\texttt{CLS}] attention to assess and discard redundant visual tokens. In this work, we identify a critical limitation of such attention-first pruning approaches, i.e., they tend to preserve semantically similar tokens, resulting in pronounced performance drops under high pruning ratios. To this end, we propose {HoloV}, a simple yet effective, plug-and-play visual token pruning framework for efficient inference. Distinct from previous attention-first schemes, HoloV rethinks token retention from a holistic perspective. By adaptively distributing the pruning budget across different spatial crops, HoloV ensures that the retained tokens capture the global visual context rather than isolated salient features. This strategy minimizes representational collapse and maintains task-relevant information even under aggressive pruning. Experimental results demonstrate that our HoloV achieves superior performance across various tasks, MLLM architectures, and pruning ratios compared to SOTA methods. For instance, LLaVA1.5 equipped with HoloV preserves 95.8\% of the original performance after pruning 88.9\% of visual tokens, achieving superior efficiency-accuracy trade-offs.
Authors: Jincan Lou, Jingkun Chen, Haoquan Li, Hang Li, Wenjian Huang, Weihua Chen, Fan Wang, Jianguo Zhang
Abstract: Accurate liver segmentation from contrast-enhanced MRI is essential for diagnosis, treatment planning, and disease monitoring. However, it remains challenging due to limited annotated data, heterogeneous enhancement protocols, and significant domain shifts across scanners and institutions. Traditional image-to-image translation frameworks have made great progress in domain generalization, but their application is not straightforward. For example, Pix2Pix requires image registration, and cycle-GAN cannot be integrated seamlessly into segmentation pipelines. Meanwhile, these methods are originally used to deal with cross-modality scenarios, and often introduce structural distortions and suffer from unstable training, which may pose drawbacks in our single-modality scenario. To address these challenges, we propose CoSSeg-TTA, a compact segmentation framework for the GED4 (Gd-EOB-DTPA enhanced hepatobiliary phase MRI) modality built upon nnU-Netv2 and enhanced with a semi-supervised mean teacher scheme to exploit large amounts of unlabeled volumes. A domain adaptation module, incorporating a randomized histogram-based style appearance transfer function and a trainable contrast-aware network, enriches domain diversity and mitigates cross-center variability. Furthermore, a continual test-time adaptation strategy is employed to improve robustness during inference. Extensive experiments demonstrate that our framework consistently outperforms the nnU-Netv2 baseline, achieving superior Dice score and Hausdorff Distance while exhibiting strong generalization to unseen domains under low-annotation conditions.
Authors: Hao Fang, Zechao Zhan, Weixin Feng, Ziwei Huang, Xubin Li, Tiezheng Ge
Abstract: Recent advances in image generation and editing technologies have enabled state-of-the-art models to achieve impressive results in general domains. However, when applied to e-commerce scenarios, these general models often encounter consistency limitations. To address this challenge, we introduce TBStar-Edit, an new image editing model tailored for the e-commerce domain. Through rigorous data engineering, model architecture design and training strategy, TBStar-Edit achieves precise and high-fidelity image editing while maintaining the integrity of product appearance and layout. Specifically, for data engineering, we establish a comprehensive data construction pipeline, encompassing data collection, construction, filtering, and augmentation, to acquire high-quality, instruction-following, and strongly consistent editing data to support model training. For model architecture design, we design a hierarchical model framework consisting of a base model, pattern shifting modules, and consistency enhancement modules. For model training, we adopt a two-stage training strategy to enhance the consistency preservation: first stage for editing pattern shifting, and second stage for consistency enhancement. Each stage involves training different modules with separate datasets. Finally, we conduct extensive evaluations of TBStar-Edit on a self-proposed e-commerce benchmark, and the results demonstrate that TBStar-Edit outperforms existing general-domain editing models in both objective metrics (VIE Score) and subjective user preference.
Authors: Yongxuan Lyu, Guangfeng Jiang, Hongsi Liu, Jun Liu
Abstract: The manual annotation of outdoor LiDAR point clouds for instance segmentation is extremely costly and time-consuming. Current methods attempt to reduce this burden but still rely on some form of human labeling. To completely eliminate this dependency, we introduce ALISE, a novel framework that performs LiDAR instance segmentation without any annotations. The central challenge is to generate high-quality pseudo-labels in a fully unsupervised manner. Our approach starts by employing Vision Foundation Models (VFMs), guided by text and images, to produce initial pseudo-labels. We then refine these labels through a dedicated spatio-temporal voting module, which combines 2D and 3D semantics for both offline and online optimization. To achieve superior feature learning, we further introduce two forms of semantic supervision: a set of 2D prior-based losses that inject visual knowledge into the 3D network, and a novel prototype-based contrastive loss that builds a discriminative feature space by exploiting 3D semantic consistency. This comprehensive design results in significant performance gains, establishing a new state-of-the-art for unsupervised 3D instance segmentation. Remarkably, our approach even outperforms MWSIS, a method that operates with supervision from ground-truth (GT) 2D bounding boxes by a margin of 2.53% in mAP (50.95% vs. 48.42%).
Authors: Shih-Yao Lin, Sibendu Paul, Caren Chen
Abstract: Selecting informative keyframes is critical for efficient video understanding, yet existing approaches often rely on heuristics, ignore semantics, or produce redundant frames. We propose KeyScore, a caption-aware frame scoring method that combines three complementary signals: semantic similarity to captions, temporal representativeness, and contextual drop impact. Applied to large-scale video-caption datasets, KeyScore generates frame-level importance scores that enable training keyframe extractors or guiding video-language models. To support this, we also propose STACFP, a Spatio-Temporal Adaptive Clustering method that generates diverse and compact frame proposals across long videos. Together, KeyScore and STACFP reduce uninformative frames while preserving critical content, resulting in faster and more accurate inference. Our experiments on three standard video-language benchmarks (MSRVTT, MSVD, DiDeMo) show that combining STACFP and KeyScore enables up to 99% frame reduction compared to full-frame processing, while outperforming uniform 8-frame encoders in video-text retrieval, keyframe extraction, and action recognition tasks. By focusing on semantically relevant frames, our method enhances both efficiency and performance, enabling scalable and caption-grounded video understanding.
Authors: Fei Zhang, Rob Chancia, Josie Clapp, Amirhossein Hassanzadeh, Dimah Dera, Richard MacKenzie, Jan van Aardt
Abstract: Accurate semantic segmentation of terrestrial laser scanning (TLS) point clouds is limited by costly manual annotation. We propose a semi-automated, uncertainty-aware pipeline that integrates spherical projection, feature enrichment, ensemble learning, and targeted annotation to reduce labeling effort, while sustaining high accuracy. Our approach projects 3D points to a 2D spherical grid, enriches pixels with multi-source features, and trains an ensemble of segmentation networks to produce pseudo-labels and uncertainty maps, the latter guiding annotation of ambiguous regions. The 2D outputs are back-projected to 3D, yielding densely annotated point clouds supported by a three-tier visualization suite (2D feature maps, 3D colorized point clouds, and compact virtual spheres) for rapid triage and reviewer guidance. Using this pipeline, we build Mangrove3D, a semantic segmentation TLS dataset for mangrove forests. We further evaluate data efficiency and feature importance to address two key questions: (1) how much annotated data are needed and (2) which features matter most. Results show that performance saturates after ~12 annotated scans, geometric features contribute the most, and compact nine-channel stacks capture nearly all discriminative power, with the mean Intersection over Union (mIoU) plateauing at around 0.76. Finally, we confirm the generalization of our feature-enrichment strategy through cross-dataset tests on ForestSemantic and Semantic3D. Our contributions include: (i) a robust, uncertainty-aware TLS annotation pipeline with visualization tools; (ii) the Mangrove3D dataset; and (iii) empirical guidance on data efficiency and feature importance, thus enabling scalable, high-quality segmentation of TLS point clouds for ecological monitoring and beyond. The dataset and processing scripts are publicly available at https://fz-rit.github.io/through-the-lidars-eye/.
Authors: Qinghongbing Xie, Zhaoyuan Xia, Feng Zhu, Lijun Gong, Ziyue Li, Rui Zhao, Long Zeng
Abstract: Recently spatial-temporal intelligence of Visual-Language Models (VLMs) has attracted much attention due to its importance for Autonomous Driving, Embodied AI and General Artificial Intelligence. Existing spatial-temporal benchmarks mainly focus on egocentric perspective reasoning with images/video context, or geographic perspective reasoning with graphics context (eg. a map), thus fail to assess VLMs' geographic spatial-temporal intelligence with both images/video and graphics context, which is important for areas like traffic management and emergency response. To address the gaps, we introduce Geo-Temporal Reasoning benchmark (GTR-Bench), a novel challenge for geographic temporal reasoning of moving targets in a large-scale camera network. GTR-Bench is more challenging as it requires multiple perspective switches between maps and videos, joint reasoning across multiple videos with non-overlapping fields of view, and inference over spatial-temporal regions that are unobserved by any video context. Evaluations of more than 10 popular VLMs on GTR-Bench demonstrate that even the best proprietary model, Gemini-2.5-Pro (34.9%), significantly lags behind human performance (78.61%) on geo-temporal reasoning. Moreover, our comprehensive analysis on GTR-Bench reveals three primary deficiencies of current models for geo-temporal reasoning. (1) VLMs' reasoning is impaired by an imbalanced utilization of spatial-temporal context. (2) VLMs are weak in temporal forecasting, which leads to worse performance on temporal-emphasized tasks than on spatial-emphasized tasks. (3) VLMs lack the proficiency to comprehend or align the map data with multi-view video inputs. We believe GTR-Bench offers valuable insights and opens up new opportunities for research and applications in spatial-temporal intelligence. Benchmark and code will be released at https://github.com/X-Luffy/GTR-Bench.
Authors: Moon Ye-Bin, Roy Miles, Tae-Hyun Oh, Ismail Elezi, Jiankang Deng
Abstract: Image retouching not only enhances visual quality but also serves as a means of expressing personal preferences and emotions. However, existing learning-based approaches require large-scale paired data and operate as black boxes, making the retouching process opaque and limiting their adaptability to handle diverse, user- or image-specific adjustments. In this work, we propose RetouchLLM, a training-free white-box image retouching system, which requires no training data and performs interpretable, code-based retouching directly on high-resolution images. Our framework progressively enhances the image in a manner similar to how humans perform multi-step retouching, allowing exploration of diverse adjustment paths. It comprises of two main modules: a visual critic that identifies differences between the input and reference images, and a code generator that produces executable codes. Experiments demonstrate that our approach generalizes well across diverse retouching styles, while natural language-based user interaction enables interpretable and controllable adjustments tailored to user intent.
Authors: Haipeng Liu, Yang Wang, Meng Wang
Abstract: Text-guided image inpainting aims at reconstructing the masked regions as per text prompts, where the longstanding challenges lie in the preservation for unmasked regions, while achieving the semantics consistency between unmasked and inpainted masked regions. Previous arts failed to address both of them, always with either of them to be remedied. Such facts, as we observed, stem from the entanglement of the hybrid (e.g., mid-and-low) frequency bands that encode varied image properties, which exhibit different robustness to text prompts during the denoising process. In this paper, we propose a null-text-null frequency-aware diffusion models, dubbed \textbf{NTN-Diff}, for text-guided image inpainting, by decomposing the semantics consistency across masked and unmasked regions into the consistencies as per each frequency band, while preserving the unmasked regions, to circumvent two challenges in a row. Based on the diffusion process, we further divide the denoising process into early (high-level noise) and late (low-level noise) stages, where the mid-and-low frequency bands are disentangled during the denoising process. As observed, the stable mid-frequency band is progressively denoised to be semantically aligned during text-guided denoising process, which, meanwhile, serves as the guidance to the null-text denoising process to denoise low-frequency band for the masked regions, followed by a subsequent text-guided denoising process at late stage, to achieve the semantics consistency for mid-and-low frequency bands across masked and unmasked regions, while preserve the unmasked regions. Extensive experiments validate the superiority of NTN-Diff over the state-of-the-art diffusion models to text-guided diffusion models. Our code can be accessed from https://github.com/htyjers/NTN-Diff.
Authors: Fevziye Irem Eyiokur, Dogucan Yaman, Haz{\i}m Kemal Ekenel, Alexander Waibel
Abstract: Embodied Reference Understanding requires identifying a target object in a visual scene based on both language instructions and pointing cues. While prior works have shown progress in open-vocabulary object detection, they often fail in ambiguous scenarios where multiple candidate objects exist in the scene. To address these challenges, we propose a novel ERU framework that jointly leverages LLM-based data augmentation, depth-map modality, and a depth-aware decision module. This design enables robust integration of linguistic and embodied cues, improving disambiguation in complex or cluttered environments. Experimental results on two datasets demonstrate that our approach significantly outperforms existing baselines, achieving more accurate and reliable referent detection.
Authors: Michael Niemeyer, Fabian Manhardt, Marie-Julie Rakotosaona, Michael Oechsle, Christina Tsalicoglou, Keisuke Tateno, Jonathan T. Barron, Federico Tombari
Abstract: Recent advances in neural scene representations have led to unprecedented quality in 3D reconstruction and view synthesis. Despite achieving high-quality results for common benchmarks with curated data, outputs often degrade for data that contain per image variations such as strong exposure changes, present, e.g., in most scenes with indoor and outdoor areas or rooms with windows. In this paper, we introduce Neural Exposure Fields (NExF), a novel technique for robustly reconstructing 3D scenes with high quality and 3D-consistent appearance from challenging real-world captures. In the core, we propose to learn a neural field predicting an optimal exposure value per 3D point, enabling us to optimize exposure along with the neural scene representation. While capture devices such as cameras select optimal exposure per image/pixel, we generalize this concept and perform optimization in 3D instead. This enables accurate view synthesis in high dynamic range scenarios, bypassing the need of post-processing steps or multi-exposure captures. Our contributions include a novel neural representation for exposure prediction, a system for joint optimization of the scene representation and the exposure field via a novel neural conditioning mechanism, and demonstrated superior performance on challenging real-world data. We find that our approach trains faster than prior works and produces state-of-the-art results on several benchmarks improving by over 55% over best-performing baselines.
Authors: Darshan Thaker, Abhishek Goyal, Ren\'e Vidal
Abstract: Image restoration aims to recover high-quality images from degraded observations. When the degradation process is known, the recovery problem can be formulated as an inverse problem, and in a Bayesian context, the goal is to sample a clean reconstruction given the degraded observation. Recently, modern pretrained diffusion models have been used for image restoration by modifying their sampling procedure to account for the degradation process. However, these methods often rely on certain approximations that can lead to significant errors and compromised sample quality. In this paper, we provide the first rigorous analysis of this approximation error for linear inverse problems under distributional assumptions on the space of natural images, demonstrating cases where previous works can fail dramatically. Motivated by our theoretical insights, we propose a simple modification to existing diffusion-based restoration methods. Our approach introduces a time-varying low-pass filter in the frequency domain of the measurements, progressively incorporating higher frequencies during the restoration process. We develop an adaptive curriculum for this frequency schedule based on the underlying data distribution. Our method significantly improves performance on challenging image restoration tasks including motion deblurring and image dehazing.
Authors: Dipam Goswami, Simone Magistri, Kai Wang, Bart{\l}omiej Twardowski, Andrew D. Bagdanov, Joost van de Weijer
Abstract: Using pre-trained models has been found to reduce the effect of data heterogeneity and speed up federated learning algorithms. Recent works have explored training-free methods using first- and second-order statistics to aggregate local client data distributions at the server and achieve high performance without any training. In this work, we propose a training-free method based on an unbiased estimator of class covariance matrices which only uses first-order statistics in the form of class means communicated by clients to the server. We show how these estimated class covariances can be used to initialize the global classifier, thus exploiting the covariances without actually sharing them. We also show that using only within-class covariances results in a better classifier initialization. Our approach improves performance in the range of 4-26% with exactly the same communication cost when compared to methods sharing only class means and achieves performance competitive or superior to methods sharing second-order statistics with dramatically less communication overhead. The proposed method is much more communication-efficient than federated prompt-tuning methods and still outperforms them. Finally, using our method to initialize classifiers and then performing federated fine-tuning or linear probing again yields better performance. Code is available at https://github.com/dipamgoswami/FedCOF.
Authors: Yifei Dong, Fengyi Wu, Qi He, Zhi-Qi Cheng, Heng Li, Minghan Li, Zebang Cheng, Yuxuan Zhou, Jingdong Sun, Qi Dai, Alexander G Hauptmann
Abstract: Vision-and-Language Navigation (VLN) has been studied mainly in either discrete or continuous settings, with little attention to dynamic, crowded environments. We present HA-VLN 2.0, a unified benchmark introducing explicit social-awareness constraints. Our contributions are: (i) a standardized task and metrics capturing both goal accuracy and personal-space adherence; (ii) HAPS 2.0 dataset and simulators modeling multi-human interactions, outdoor contexts, and finer language-motion alignment; (iii) benchmarks on 16,844 socially grounded instructions, revealing sharp performance drops of leading agents under human dynamics and partial observability; and (iv) real-world robot experiments validating sim-to-real transfer, with an open leaderboard enabling transparent comparison. Results show that explicit social modeling improves navigation robustness and reduces collisions, underscoring the necessity of human-centric approaches. By releasing datasets, simulators, baselines, and protocols, HA-VLN 2.0 provides a strong foundation for safe, socially responsible navigation research.
Authors: Feihan Feng, Jingxin Nie
Abstract: Decoding sensory experiences from neural activity to reconstruct human-perceived visual stimuli and semantic content remains a challenge in neuroscience and artificial intelligence. Despite notable progress in current brain decoding models, a critical gap still persists in their systematic integration with established neuroscientific theories and the exploration of underlying neural mechanisms. Here, we present a novel framework that directly decodes fMRI signals into textual descriptions of viewed natural images. Our novel deep learning model, trained without visual information, achieves state-of-the-art semantic decoding performance, generating meaningful captions that capture the core semantic content of complex scenes. Neuroanatomical analysis reveals the critical role of higher-level visual cortices, including MT+ complex, ventral stream visual cortex, and inferior parietal cortex, in visual semantic processing. Furthermore, category-specific analysis demonstrates nuanced neural representations for semantic dimensions like animacy and motion. This work provides a more direct and interpretable framework to the brain's semantic decoding, offering a powerful new methodology for probing the neural basis of complex semantic processing, refining the understanding of the distributed semantic network, and potentially developing brain-sinpired language models.
Authors: Sanjeev Muralikrishnan, Niladri Shekhar Dutt, Niloy J. Mitra
Abstract: Animation retargetting applies sparse motion description (e.g., keypoint sequences) to a character mesh to produce a semantically plausible and temporally coherent full-body mesh sequence. Existing approaches come with restrictions -- they require access to template-based shape priors or artist-designed deformation rigs, suffer from limited generalization to unseen motion and/or shapes, or exhibit motion jitter. We propose Self-supervised Motion Fields (SMF), a self-supervised framework that is trained with only sparse motion representations, without requiring dataset-specific annotations, templates, or rigs. At the heart of our method are Kinetic Codes, a novel autoencoder-based sparse motion encoding, that exposes a semantically rich latent space, simplifying large-scale training. Our architecture comprises dedicated spatial and temporal gradient predictors, which are jointly trained in an end-to-end fashion. The combined network, regularized by the Kinetic Codes' latent space, has good generalization across both unseen shapes and new motions. We evaluated our method on unseen motion sampled from AMASS, D4D, Mixamo, and raw monocular video for animation transfer on various characters with varying shapes and topology. We report a new SoTA on the AMASS dataset in the context of generalization to unseen motion. Code, weights, and supplementary are available on the project webpage at https://motionfields.github.io/
Authors: Tri Ton, Ji Woo Hong, Chang D. Yoo
Abstract: This paper introduces Timestep-Adaptive Representation Alignment with Onset-Aware Conditioning (TARO), a novel framework for high-fidelity and temporally coherent video-to-audio synthesis. Built upon flow-based transformers, which offer stable training and continuous transformations for enhanced synchronization and audio quality, TARO introduces two key innovations: (1) Timestep-Adaptive Representation Alignment (TRA), which dynamically aligns latent representations by adjusting alignment strength based on the noise schedule, ensuring smooth evolution and improved fidelity, and (2) Onset-Aware Conditioning (OAC), which integrates onset cues that serve as sharp event-driven markers of audio-relevant visual moments to enhance synchronization with dynamic visual events. Extensive experiments on the VGGSound and Landscape datasets demonstrate that TARO outperforms prior methods, achieving relatively 53% lower Frechet Distance (FD), 29% lower Frechet Audio Distance (FAD), and a 97.19% Alignment Accuracy, highlighting its superior audio quality and synchronization precision.
Authors: Simone Azeglio, Arianna Di Bernardo
Abstract: Recent advances in diffusion models have demonstrated their remarkable ability to capture complex image distributions, but the geometric properties of the learned data manifold remain poorly understood. We address this gap by introducing a score-based Riemannian metric that leverages the Stein score function from diffusion models to characterize the intrinsic geometry of the data manifold without requiring explicit parameterization. Our approach defines a metric tensor in the ambient space that stretches distances perpendicular to the manifold while preserving them along tangential directions, effectively creating a geometry where geodesics naturally follow the manifold's contours. We develop efficient algorithms for computing these geodesics and demonstrate their utility for both interpolation between data points and extrapolation beyond the observed data distribution. Through experiments on synthetic data with known geometry, Rotated MNIST, and complex natural images via Stable Diffusion, we show that our score-based geodesics capture meaningful transformations that respect the underlying data distribution. Our method consistently outperforms baseline approaches on perceptual metrics (LPIPS) and distribution-level metrics (FID, KID), producing smoother, more realistic image transitions. These results reveal the implicit geometric structure learned by diffusion models and provide a principled way to navigate the manifold of natural images through the lens of Riemannian geometry.
Authors: Elias Abad Rocamora, Christian Schlarmann, Naman Deep Singh, Yongtao Wu, Matthias Hein, Volkan Cevher
Abstract: Adversarial input attacks can cause a significant shift of CLIP embeddings. This can affect the downstream robustness of models incorporating CLIP in the pipeline, such as text-to-image generative models or large vision language models. While some efforts have been done towards making the CLIP image encoders robust, the robustness of text encoders remains unexplored. In this work, we cover this gap in the literature. We propose LEAF: an efficient adversarial finetuning method for the text domain, with the ability to scale to large CLIP models. Our models significantly improve the zero-shot adversarial accuracy in the text domain, while maintaining the vision performance provided by robust image encoders. When combined with text-to-image diffusion models, we can improve the generation quality under adversarial noise. In multimodal retrieval tasks, LEAF improves the recall under adversarial noise over standard CLIP models. Finally, we show that robust text encoders facilitate better reconstruction of input text from its embedding via direct optimization. We open-source our code ( https://github.com/LIONS-EPFL/LEAF ) and models ( https://huggingface.co/LEAF-CLIP ).
URLs: https://github.com/LIONS-EPFL/LEAF, https://huggingface.co/LEAF-CLIP
Authors: Wenrui Zhou, Mohamed Hendy, Shu Yang, Qingsong Yang, Zikun Guo, Yuyu Luo, Lijie Hu, Di Wang
Abstract: As video large language models (Video-LLMs) become increasingly integrated into real-world applications that demand grounded multimodal reasoning, ensuring their factual consistency and reliability is of critical importance. However, sycophancy, the tendency of these models to align with user input even when it contradicts the visual evidence, undermines their trustworthiness in such contexts. Current sycophancy research has largely overlooked its specific manifestations in the video-language domain, resulting in a notable absence of systematic benchmarks and targeted evaluations to understand how Video-LLMs respond under misleading user input. To fill this gap, we propose VISE (Video-LLM Sycophancy Benchmarking and Evaluation), the first benchmark designed to evaluate sycophantic behavior in state-of-the-art Video-LLMs across diverse question formats, prompt biases, and visual reasoning tasks. Specifically, VISE pioneeringly brings linguistic perspectives on sycophancy into the video domain, enabling fine-grained analysis across multiple sycophancy types and interaction patterns. Furthermore, we propose two potential training-free mitigation strategies, revealing potential paths for reducing sycophantic bias: (i) enhancing visual grounding through interpretable key-frame selection and (ii) steering model behavior away from sycophancy via targeted, inference-time intervention on its internal neural representations. Our code is available at https://github.com/William030422/Video-Sycophancy.
Authors: Ahmed Jaheen, Abdelrahman Elsayed, Damir Kim, Daniil Tikhonov, Matheus Scatolin, Mohor Banerjee, Qiankun Ji, Mostafa Salem, Hu Wang, Sarim Hashmi, Mohammad Yaqub
Abstract: Brain cancer affects millions worldwide, and in nearly every clinical setting, doctors rely on magnetic resonance imaging (MRI) to diagnose and monitor gliomas. However, the current standard for tumor quantification through manual segmentation of multi-parametric MRI is time-consuming, requires expert radiologists, and is often infeasible in under-resourced healthcare systems. This problem is especially pronounced in low-income regions, where MRI scanners are of lower quality and radiology expertise is scarce, leading to incorrect segmentation and quantification. In addition, the number of acquired MRI scans in Africa is typically small. To address these challenges, the BraTS-Lighthouse 2025 Challenge focuses on robust tumor segmentation in sub-Saharan Africa (SSA), where resource constraints and image quality degradation introduce significant shifts. In this study, we present EMedNeXt -- an enhanced brain tumor segmentation framework based on MedNeXt V2 with deep supervision and optimized post-processing pipelines tailored for SSA. EMedNeXt introduces three key contributions: a larger region of interest, an improved nnU-Net v2-based architectural skeleton, and a robust model ensembling system. Evaluated on the hidden validation set, our solution achieved an average LesionWise DSC of 0.897 with an average LesionWise NSD of 0.541 and 0.84 at a tolerance of 0.5 mm and 1.0 mm, respectively.
Authors: Lixuan He, Jie Feng, Yong Li
Abstract: Large Language Models (LLMs) are typically fine-tuned for reasoning tasks through a two-stage pipeline of Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL), a process fraught with catastrophic forgetting and suboptimal trade-offs between imitation and exploration. Recent single-stage methods attempt to unify SFT and RL using heuristics, but lack a principled mechanism for dynamically balancing the two paradigms. In this paper, we reframe this challenge through the theoretical lens of \textbf{implicit rewards}, viewing SFT and RL not as distinct methods but as complementary reward signals. We introduce \textbf{Adaptive Meta Fine-Tuning (AMFT)}, a novel single-stage algorithm that learns the optimal balance between SFT's implicit, path-level reward and RL's explicit, outcome-based reward. The core of AMFT is a \textbf{meta-gradient adaptive weight controller} that treats the SFT-RL balance as a learnable parameter, dynamically optimizing it to maximize long-term task performance. This forward-looking approach, regularized by policy entropy for stability, autonomously discovers an effective training curriculum. We conduct a comprehensive evaluation on challenging benchmarks spanning mathematical reasoning, abstract visual reasoning (General Points), and vision-language navigation (V-IRL). AMFT consistently establishes a new state-of-the-art and demonstrats superior generalization on out-of-distribution (OOD) tasks. Ablation studies and training dynamic analysis confirm that the meta-learning controller is crucial for AMFT's stability, sample efficiency, and performance, offering a more principled and effective paradigm for LLM alignment. Our codes are open-sourced via https://github.com/hlxtsyj/AMFT.
Authors: Joseph Ko, Hariprasath Govindarajan, Fredrik Lindsten, Vanessa Przybylo, Kara Sulia, Marcus van Lier-Walqui, Kara Lamb
Abstract: Ice-containing clouds strongly impact climate, but they are hard to model due to ice crystal habit (i.e., shape) diversity. We use self-supervised learning (SSL) to learn latent representations of crystals from ice crystal imagery. By pre-training a vision transformer with many cloud particle images, we learn robust representations of crystal morphology, which can be used for various science-driven tasks. Our key contributions include (1) validating that our SSL approach can be used to learn meaningful representations, and (2) presenting a relevant application where we quantify ice crystal diversity with these latent representations. Our results demonstrate the power of SSL-driven representations to improve the characterization of ice crystals and subsequently constrain their role in Earth's climate system.
Authors: Yuxiao Yi, Qingyao Zhuang, Zhi-Qin John Xu, Xiaowen Wang, Yan Ren, Tianming Qiu
Abstract: Pediatric brain tumor segmentation presents unique challenges due to the rarity and heterogeneity of these malignancies, yet remains critical for clinical diagnosis and treatment planning. We propose an ensemble approach integrating nnU-Net, Swin UNETR, and HFF-Net for the BraTS-PED 2025 challenge. Our method incorporates three key extensions: adjustable initialization scales for optimal nnU-Net complexity control, transfer learning from BraTS 2021 pre-trained models to enhance Swin UNETR's generalization on pediatric dataset, and frequency domain decomposition for HFF-Net to separate low-frequency tissue contours from high-frequency texture details. Our final ensemble framework combines nnU-Net ($\gamma=0.7$), fine-tuned Swin UNETR, and HFF-Net, achieving Dice scores of 62.7% (CC), 83.2% (ED), 72.9% (ET), 85.7% (NET), 91.8% (TC), and 92.6% (WT) on the unseen test dataset, respectively. Our proposed method achieves first place (rank 1st) in the BraTS 2025 Pediatric Brain Tumor Segmentation Challenge.
Authors: Berker Demirel, Marco Fumero, Theofanis Karaletsos, Francesco Locatello
Abstract: Simulating in silico cellular responses to interventions is a promising direction to accelerate high-content image-based assays, critical for advancing drug discovery and gene editing. To support this, we introduce MorphGen, a state-of-the-art diffusion-based generative model for fluorescent microscopy that enables controllable generation across multiple cell types and perturbations. To capture biologically meaningful patterns consistent with known cellular morphologies, MorphGen is trained with an alignment loss to match its representations to the phenotypic embeddings of OpenPhenom, a state-of-the-art biological foundation model. Unlike prior approaches that compress multichannel stains into RGB images -- thus sacrificing organelle-specific detail -- MorphGen generates the complete set of fluorescent channels jointly, preserving per-organelle structures and enabling a fine-grained morphological analysis that is essential for biological interpretation. We demonstrate biological consistency with real images via CellProfiler features, and MorphGen attains an FID score over 35% lower than the prior state-of-the-art MorphoDiff, which only generates RGB images for a single cell type. Code is available at https://github.com/czi-ai/MorphGen.
Authors: Yujie Zhou, Pengyang Ling, Jiazi Bu, Yibin Wang, Yuhang Zang, Jiaqi Wang, Li Niu, Guangtao Zhai
Abstract: The integration of online reinforcement learning (RL) into diffusion and flow models has recently emerged as a promising approach for aligning generative models with human preferences. Stochastic sampling via Stochastic Differential Equations (SDE) is employed during the denoising process to generate diverse denoising directions for RL exploration. While existing methods effectively explore potential high-value samples, they suffer from sub-optimal preference alignment due to sparse and narrow reward signals. To address these challenges, we propose a novel Granular-GRPO (G$^2$RPO) framework that achieves precise and comprehensive reward assessments of sampling directions in reinforcement learning of flow models. Specifically, a Singular Stochastic Sampling strategy is introduced to support step-wise stochastic exploration while enforcing a high correlation between the reward and the injected noise, thereby facilitating a faithful reward for each SDE perturbation. Concurrently, to eliminate the bias inherent in fixed-granularity denoising, we introduce a Multi-Granularity Advantage Integration module that aggregates advantages computed at multiple diffusion scales, producing a more comprehensive and robust evaluation of the sampling directions. Experiments conducted on various reward models, including both in-domain and out-of-domain evaluations, demonstrate that our G$^2$RPO significantly outperforms existing flow-based GRPO baselines,highlighting its effectiveness and robustness.
Authors: Seunghyeon Kim, Jinho Kim, Taesun Yeom, Wonpyo Park, Kyuyeun Kim, Jaeho Lee
Abstract: Transformer-based vision encoders -- such as CLIP -- are central to multimodal intelligence, powering applications from autonomous web agents to robotic control. Since these applications often demand real-time processing of massive visual data, reducing the inference cost of vision encoders is critical. Post-training quantization offers a practical path, but remains challenging even at 8-bit precision due to massive-scale activations (i.e., outliers). In this work, we propose $\textit{RegCache}$, a training-free algorithm to mitigate outliers in vision encoders, enabling quantization with significantly smaller accuracy drops. The proposed RegCache introduces outlier-prone yet semantically meaningless prefix tokens to the target vision encoder, which prevents other tokens from having outliers. Notably, we observe that outliers in vision encoders behave differently from those in language models, motivating two technical innovations: middle-layer prefixing and token deletion. Experiments show that our method consistently improves the accuracy of quantized models across both text-supervised and self-supervised vision encoders.
Authors: Masih Aminbeidokhti, Heitor Rapela Medeiros, Srikanth Muralidharan, Eric Granger, Marco Pedersoli
Abstract: Ensembling fine-tuned models initialized from powerful pre-trained weights is a common strategy to improve robustness under distribution shifts, but it comes with substantial computational costs due to the need to train and store multiple models. Dropout offers a lightweight alternative by simulating ensembles through random neuron deactivation; however, when applied to pre-trained models, it tends to over-regularize and disrupt critical representations necessary for generalization. In this work, we investigate Mixout, a stochastic regularization technique that provides an alternative to Dropout for domain generalization. Rather than deactivating neurons, Mixout mitigates overfitting by probabilistically swapping a subset of fine-tuned weights with their pre-trained counterparts during training, thereby maintaining a balance between adaptation and retention of prior knowledge. Our study reveals that achieving strong performance with Mixout on domain generalization benchmarks requires a notably high masking probability of 0.9 for ViTs and 0.8 for ResNets. While this may seem like a simple adjustment, it yields two key advantages for domain generalization: (1) higher masking rates more strongly penalize deviations from the pre-trained parameters, promoting better generalization to unseen domains; and (2) high-rate masking substantially reduces computational overhead, cutting gradient computation by up to 45% and gradient memory usage by up to 90%. Experiments across five domain generalization benchmarks, PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet, using ResNet and ViT architectures, show that our approach, High-rate Mixout, achieves out-of-domain accuracy comparable to ensemble-based methods while significantly reducing training costs.