论文检索

Matthew Raffel, Lizhong Chen

The Kolmogorov-Arnold Network (KAN) has been gaining popularity as an alternative to the multilayer perceptron (MLP) due to its greater expressiveness and interpretability. Even so, KAN suffers from training instability and being orders of magnitude slower due to its increased computational cost, limiting its applicability to large-scale tasks. Recently, the Kolmogorov-Arnold Transformer (KAT) has been proposed, achieving FLOPs comparable to traditional Transformer models with MLPs by leveraging Group-Rational KAN (GR-KAN). Unfortunately, despite the comparable FLOPs, our testing shows that KAT remains 123x slower during training, indicating that there are other performance bottlenecks beyond FLOPs. In this paper, we conduct a series of experiments to understand the root cause of the slowdown in KAT. We uncover that the slowdown can be isolated to memory stalls, linked more specifically to inefficient gradient accumulations in the backward pass of GR-KAN. To address this memory bottleneck, we propose FlashKAT, which minimizes accesses to slow memory and the usage of atomic adds through a restructured kernel. Evaluations show that FlashKAT achieves up to an 86.5x training speedup over state-of-the-art KAT while reducing rounding errors in gradient computation.

Karim Radouane, Sylvie Ranwez, Julien Lagarde, Andon Tchechmedjiev

In this paper, we address a challenging task, synchronous motion captioning, that aim to generate a language description synchronized with human motion sequences. This task pertains to numerous applications, such as aligned sign language transcription and unsupervised action segmentation and temporal grounding. Our method introduces mechanisms to control self- and cross-attention distributions of the Transformer, allowing interpretability and aligned text generation. We achieve this through masking strategies and structuring losses that push the model to maximize attention only on the most important frames contributing to the generation of a motion word. These constraints aim to prevent undesired mixing of information in attention maps and to provide a monotonic attention distribution across tokens. Thus, the cross attentions of tokens are used for progressive text generation in synchronization with human motion sequences. We demonstrate the superior performance of our approach through evaluation on the two available benchmark datasets, KIT-ML and HumanML3D. As visual evaluation is essential for this task, we provide a comprehensive set of animated visual illustrations of the output of synchronous text generation in the code repository.

Yansong Qu, Shaohui Dai, Xinyang Li, Yuze Wang, You Shen, Shengchuan Zhang, Liujuan Cao

Reconstructing 3D objects from a single image is a long-standing challenge, particularly under real-world occlusions. While recent diffusion-based view synthesis models can generate consistent novel views from a single RGB image, they generally assume fully visible inputs and struggle when parts of the object are occluded, leading to inconsistent views and degraded 3D reconstruction quality. To address this limitation, we propose DeOcc-1-to-3, an end-to-end framework for occlusion-aware multi-view generation. Our method directly synthesizes six structurally consistent novel views from a single partially occluded image, enabling downstream 3D reconstruction without requiring prior inpainting or manual annotations. We design a self-supervised training pipeline that leverages occluded–unoccluded image pairs and pseudo-ground-truth views to guide structure-aware completion and view consistency. Without modifying the original architecture, we fully fine-tune the diffusion model to jointly learn completion and multi-view generation. Additionally, we introduce the first benchmark for occlusion-aware reconstruction, covering diverse occlusion levels, object categories, and mask patterns, providing a standardized evaluation protocol.

Wentao Qu, Guofeng Mei, Jing Wang, Yujiao Wu, Xiaoshui Huang, Liang Xiao

Denoising Diffusion Probabilistic Models (DDPMs) have shown success in robust 3D object detection tasks. Existing methods often rely on the score matching from 3D boxes or pre-trained diffusion priors. However, they typically require multi-step iterations in inference, which limits efficiency. To address this, we propose a Robust single-stage fully Sparse 3D object Detection Network with a Detachable Latent Framework (DLF) of DDPMs, named RSDNet. Specifically, RSDNet learns the denoising process in latent feature spaces through lightweight denoising networks like multi-level denoising autoencoders (DAEs). This enables RSDNet to effectively understand scene distributions under multi-level perturbations, achieving robust and reliable detection. Meanwhile, we reformulate the noising and denoising mechanisms of DDPMs, enabling DLF to construct multi-type and multi-level noise samples and targets, enhancing RSDNet robustness to multiple perturbations. Furthermore, a semantic-geometric conditional guidance is introduced to perceive the object boundaries and shapes, alleviating the center feature missing problem in sparse representations, enabling RSDNet to perform in a fully sparse detection pipeline. Moreover, the detachable denoising network design of DLF enables RSDNet to perform single-step detection in inference, further enhancing detection efficiency. Extensive experiments on public benchmarks show that RSDNet can outperform existing methods, achieving state-of-the-art detection.

Kanglin Qu, Pan Gao, Qun Dai, Zhanzhi Ye, Rui Ye, Yuanhao Sun

Due to the long-range modeling ability and linear complexity property, Mamba has attracted considerable attention in point cloud analysis. Despite some interesting progress, related work still suffers from imperfect point cloud serialization, insufficient high-level geometric perception, and overfitting of the selective state space model (S6) at the core of Mamba. To this end, we resort to an SSM-based point cloud network termed CloudMamba to address the above challenges. Specifically, we propose sequence expanding and sequence merging, where the former serializes points along each axis separately and the latter serves to fuse the corresponding higher-order features causally inferred from different sequences, enabling unordered point sets to adapt more stably to the causal nature of Mamba without parameters. Meanwhile, we design chainedMamba that chains the forward and backward processes in the parallel bidirectional Mamba, capturing high-level geometric information during scanning. In addition, we propose a grouped selective state space model (GS6) via parameter sharing on S6, alleviating the overfitting problem caused by the computational mode in S6. Experiments on various point cloud tasks validate CloudMamba's ability to achieve state-of-the-art results with significantly less complexity.

Jinyuan Qu, Hongyang Li, Xingyu Chen, Shilong Liu, Yukai Shi, Tianhe Ren, Ruitao Jing, Lei Zhang

In this paper, we present SegDINO3D, a novel Transformer encoder-decoder framework for 3D instance segmentation. As 3D training data is generally not as sufficient as 2D training images, SegDINO3D is designed to fully leverage 2D representation from a pre-trained 2D detection model, including both image-level and object-level features, for improving 3D representation. SegDINO3D takes both a point cloud and its associated 2D images as input. In the encoder stage, it first enriches each 3D point by retrieving 2D image features from its corresponding image views and then leverages a 3D encoder for 3D context fusion. In the decoder stage, it formulates 3D object queries as 3D anchor boxes and performs cross-attention from 3D queries to 2D object queries obtained from 2D images using the 2D detection model. These 2D object queries serve as a compact object-level representation of 2D images, effectively avoiding the challenge of keeping thousands of image feature maps in the memory while faithfully preserving the knowledge of the pre-trained 2D model. The introducing of 3D box queries also enables the model to modulate cross-attention using the predicted boxes for more precise querying. SegDINO3D achieves the state-of-the-art performance on the ScanNetV2 and ScanNet200 3D instance segmentation benchmarks. Notably, on the challenging ScanNet200 dataset, SegDINO3D significantly outperforms prior methods by +8.7 and +6.8 mAP on the validation and hidden test sets, respectively, demonstrating its superiority.

Jierui Qu, Jianchun Zhao

Accurate whole-heart segmentation is a critical component in the precise diagnosis and interventional planning of cardiovascular diseases. Integrating complementary information from modalities such as computed tomography (CT) and magnetic resonance imaging (MRI) can significantly enhance segmentation accuracy and robustness. However, existing multi-modal segmentation methods face several limitations: severe spatial inconsistency between modalities hinders effective feature fusion; fusion strategies are often static and lack adaptability; and the processes of feature alignment and segmentation are decoupled and inefficient. To address these challenges, we propose a dual-branch U-Net architecture enhanced by reinforcement learning for feature alignment, termed RL-U2Net, designed for precise and efficient multi-modal 3D whole-heart segmentation. The model employs a dual-branch U-shaped network to process CT and MRI patches in parallel, and introduces a novel RL-XAlign module between the encoders. The module employs a cross‑modal attention mechanism to capture semantic correspondences between modalities and a reinforcement learning agent learns an optimal rotation strategy that consistently aligns anatomical pose and texture features. The aligned features are then reconstructed through their respective decoders. Finally, an ensemble‑learning–based decision module integrates the predictions from individual patches to produce the final segmentation result. Experimental results on the publicly available MM-WHS 2017 dataset demonstrate that the proposed RL-U2Net outperforms existing state-of-the-art methods, achieving Dice coefficients of 93.1% on CT and 87.0% on MRI, thereby validating the effectiveness and superiority of the proposed approach.

Gang Qu, Ping Wang, Siming Zheng, Xin Yuan

Deep learning methods have achieved remarkable success in image compressed sensing (CS) task, namely reconstructing a high-fidelity image from its compressed measurement. However, existing methods are deficient in incoherent compressed measurement at sensing phase and implicit measurement representations at reconstruction phase, limiting the overall performance. In this work, we answer two questions: (i) how to improve the measurement incoherence for decreasing the ill-posedness; (ii) how to learn informative representations from measurements. To this end, we propose a novel asymmetric Kronecker CS (AKCS) model and theoretically present its better incoherence than previous Kronecker CS with minimal increase of complexity. Moreover, apart from the explicit measurement representations in gradient descent projection in unfolding networks, we further propose a measurement-aware cross attention (MACA) mechanism to learn implicit measurement representations. We integrate AKCS and MACA into a widely-used unfolding architecture to get a measurement-enhanced unfolding network (MEUNet). Extensive experiments demonstrate that the proposed MEUNet achieves state-of-the-art (SOTA) performance in reconstruction accuracy with high efficiency.

Chenfan Qu, Yiwu Zhong, Jian Liu, Xuekang Zhu, Bohan Yu, Lianwen Jin

The growing prevalence of tampered images poses serious security threats, highlighting the urgent need for reliable detection methods. Multimodal large language models (MLLMs) demonstrate strong potential in analyzing tampered images and generating interpretations. However, they still struggle with identifying micro-level artifacts, exhibit low accuracy in localizing tampered text regions, and heavily rely on expensive annotations for forgery interpretation. To this end, we introduce TextShield-R1, the first reinforcement learning based MLLM solution for tampered text detection and reasoning. Specifically, our approach introduces Forensic Continual pre-training, an easy-to-hard curriculum that well prepares the MLLM for tampered text detection by harnessing the large-scale cheap data from natural image forensic and OCR tasks. During fine-tuning, we perform Group Relative Policy Optimization with novel reward functions to reduce annotation dependency and improve reasoning capabilities. At inference time, we enhance localization accuracy via OCR Rectification, a method that leverages the MLLM’s strong text recognition abilities to refine its predictions. Furthermore, to support rigorous evaluation, we introduce Text Forensics Reasoning (TFR) benchmark, comprising over 45k real and tampered images across 16 languages, 10 tampering techniques, and diverse domains. Rich reasoning-style annotations are included, allowing for comprehensive assessment. Our TFR benchmark simultaneously addresses seven major limitations of existing benchmarks and enables robust evaluation under cross-style, cross-method, and cross-language conditions. Extensive experiments demonstrate that TextShield-R1 significantly advances the state of the art in interpretable tampered text detection.

Zongyang Qiu, Bingyuan Wang, Xingbei Chen, Yingqing He, Zeyu Wang

Emotion plays a pivotal role in video-based expression, but existing video generation systems predominantly focus on low-level visual metrics while neglecting affective dimensions. Although emotion analysis has made progress in the visual domain, the video community lacks dedicated resources to bridge emotion understanding with generative tasks, particularly for stylized and non-realistic contexts. To address this gap, we introduce EmoVid, the first multimodal, emotion-annotated video dataset specifically designed for artistic media, which includes cartoon animations, movie clips, and animated stickers. Each video is annotated with emotion labels, visual attributes (brightness, colorfulness, hue), and text captions. Through systematic analysis, we uncover spatial and temporal patterns linking visual features to emotional perceptions across diverse video forms. Building on these insights, we develop an emotion-conditioned video generation technique by fine-tuning the Wan2.1 model. The results show a significant improvement in both quantitative metrics and the visual quality of generated videos for text-to-video and image-to-video tasks. EmoVid establishes a new benchmark and protocol for affective video computing. Our work not only offers valuable insights into visual emotion analysis in artistic videos but also provides practical methods for enhancing emotional expression in video generation. The extended version and the dataset are available on our project page.

Zhijie Qiu, Shuaibo Li, Laixin Zhang, Xuming Hu, Wei Ma

Accurately recognizing distracted driving activities in real-world scenarios is essential for improving road and pedestrian safety. However, existing approaches are prone to attending to irrelevant scene context and are susceptible to interference from redundant frames, compromising their robustness in complex driving environments. To overcome these limitations, we propose DualScope, a novel framework that captures behaviorally critical information from both spatial and temporal perspectives. In the spatial domain, we introduce a Synergistic Behavior-Centric Distillation mechanism that leverages two key information sources: (1) position-aware knowledge derived from the SAM model, which enhances the perception of critical regions and their semantic interaction structures; and (2) fine-grained visual details obtained from cropped key regions, which improve the model's ability to capture detailed patterns within behavior-relevant areas. In the temporal domain, we present the Saliency-Aware Fine-to-Coarse Temporal Modeling module, comprising three components: a Fine-Grained Motion Encoder for capturing local inter-frame dependencies; a Dynamic Difference Extractor for generating salient motion dynamics; and a Saliency-Aware Temporal Pyramid Mamba for integrating these representations to enable multi-scale temporal modeling. This design effectively captures both short-term motions and long-term behavioral patterns. Furthermore, incorporating salient dynamics enhances the model's focus on significant behavioral variations. Extensive experiments on seven publicly available DDAR datasets demonstrate that DualScope consistently outperforms state-of-the-art methods, validating its effectiveness in capturing behavioral cues across spatial and temporal dimensions.

Linwei Qiu, Gongzhe Li, Xiaozhe Zhang, Qilin Sun, Fengying Xie

Image correction and rectangling are valuable tasks in practical photography systems such as smartphones. Recent remarkable advancements in deep learning have undeniably brought about substantial performance improvements in these fields. Nevertheless, existing methods mainly rely on task-specific architectures. This significantly restricts their generalization ability and effective application across a wide range of different tasks. In this paper, we introduce the Unified Rectification Framework (UniRect), a comprehensive approach that addresses these practical tasks from a consistent distortion rectification perspective. Our approach incorporates various task-specific inverse problems into a general distortion model by simulating different types of lenses. To handle diverse distortions, UniRect adopts one task-agnostic rectification framework with a dual-component structure: a Deformation Module, which utilizes a novel Residual Progressive Thin-Plate Spline (RP-TPS) model to address complex geometric deformations, and a subsequent Restoration Module, which employs Residual Mamba Blocks (RMBs) to counteract the degradation caused by the deformation process and enhance the fidelity of the output image. Moreover, a Sparse Mixture-of-Experts (SMoEs) structure is designed to circumvent heavy task competition in multi-task learning due to varying distortions. Extensive experiments demonstrate that our models have achieved state-of-the-art performance compared with other up-to-date methods.

Junhui Qiu, Xiang Xiang, Hongyun Wang, Jiaqi Gui

Aerial multi-modal visual streams registration and fusion can generate more comprehensive scene information representations for UAVs' cross-modal perception. However, current challenges lie primarily in the essential difficulty of joint spatiotemporal representation learning from dynamic background and moving targets, and a critical shortage exists in large-scale, well-annotated multi-modal visual streams benchmark for UAV platforms. In this paper, we propose AerialFusion, a co-motion-driven unified UAVs visual streams registration and fusion that fully mines modality-invariant common features based on motion-aware, enabling spatiotemporally coherent registration and fusion. Specifically, 1) a Skewed Motion Distribution Field Co-Motion-Driven Image Registration, 2) a Co-Motion Generative Fusion, 3) a Streams-based Unified Learning. Furthermore, we introduce EUM3D, a registration and fusion benchmark for UAVs cross-modal perception. This benchmark contains 60 synchronized visible-infrared visual streams, or 122k spatially and temporally aligned pairs, most of which were taken at low-light scenes. And EUM3D provides pixel-level alignment guarantees via perspective-transform ground-truth. Extensive experiments reveal that AerialFusion surpasses current focus on image and static background fusion methods in aerial sequence scenarios, addressing spatiotemporal mismatches while suppressing cross-modal interference.

Jie Qiu, Dizuo Cao, Linwei Dai, Xin Li, Fan Yang, Dong Yu, Changying Wang, Zongheng Wen, Youqin Chen, Jianzhang Chen

Remote sensing imagery poses a distinct challenge for semantic segmentation due to its inherent fractal complexity and the diversity of geometric structures present in real-world geospatial scenes. Euclidean-based models typically assume spatial uniformity; however, such assumptions often break down when confronted with objects exhibiting markedly different structural characteristics—such as roads versus vegetation—thereby complicating the feature representation process. Hyperbolic space offers a theoretically grounded alternative for modeling such hierarchical and heterogeneous patterns, yet fully replacing Euclidean geometry incurs significant computational overhead. We therefore introduce Geometry-Aware Adaptive Routing (GAAR), a novel module that facilitates geometry-aware routing by dynamically allocating high-level features to either Euclidean or Hyperbolic subspaces through a learnable binary gating mechanism, informed by structural priors learned during training. To further promote routing stability and geometric consistency, we introduce Geometry-Aware Deterministic Regularization (GADR), a regularization strategy that encourages confident, structure-aligned assignments. GAAR is plug-and-play and integrates seamlessly into existing segmentation architectures. Experiments on three challenging Remote Sensing Image Semantic Segmentation (RSISS) benchmarks demonstrate that our approach consistently outperforms state-of-the-art (SOTA) methods, particularly in geometrically complex regions, offering a scalable and effective solution to the limitations of purely Euclidean modeling.

Heqian Qiu, Lanxiao Wang, Taijin Zhao, Zhaofeng Shi, Xiang Li, Linfeng Xu, Hongliang Li

Egocentric gaze prediction serves as a critical indicator for decoding human visual attention and cognitive processes, but its inherently limited field of view creates prediction challenges. Although exo-view data provides supplementary contextual information, it exhibits significant spatial and semantic gaps. Existing methods focus solely on isolated feature encoding in single-view paradigms, neglecting cross-view gaze correlations. To make up for this gap, we make the first exploration of cross-view gaze relationship for egocentric gaze prediction, and propose Ego-PMOVE, a novel Prompt-aware Mixture of View Experts network. Unlike prior cross-view studies that forcibly align cross-view features thereby introducing inference noise, we leverage the popular Mixture-of-Experts (MoE) and a set of flexible prompts to disentangle features from different views into three parallel experts: a view-shared expert directly modeling common semantic relationships, a view-discrepancy expert adaptively adjusting the spatial position, scale and shifts based on different view-specific features, and an egocentric expert extracting independent features to compensate for the case of missing exocentric data. To balance these experts, we further design a soft router to dynamically weight them for mining useful information while suppressing noise. A view-query gaze decoder then generates view-specific gaze attention maps, jointly optimized by gaze-heamap and cross-view contrastive loss that regularize both shared and divergent features for accurate gaze prediction. Extensive experiments across the multi-view EgoMe dataset and single-view Ego4D and EGTEA Gaze++ datasets demonstrate the effectiveness and generalizability of our approach.

Fanhao Qiu, Yangyang Zhang, Zhengxia Wang

Virtual Immunohistochemistry (IHC) staining technology employs generative models to directly synthesize IHC images from Hematoxylin and Eosin (H&E) images, reducing reliance on chemical staining while improving diagnostic efficiency and reducing costs. However, existing virtual staining methods relying on adjacent sections face two critical challenges: insufficient mining of pathological semantics and the spatial misalignment of pathological semantics due to physical discrepancies between sections. To address these, we propose GSGStain, a Graph-Semantic Guided Learning for virtual Staining. Our method innovatively transforms the problem from pixel space to graph space, enabling semantic noise correction for spatial misalignment features. Specifically, to capture the rich pathological semantics, we construct a cell graph from the H&E image to encode tissue architecture, annotating nodes with noisy biomarker semantic features derived from misaligned adjacent IHC sections. Furthermore, to correct for the semantic misalignment, a Graph Semantic Rectification Module (GSRM) then refines these features using graph contextual reasoning, while a Graph Semantic Consistency Loss ensures alignment between generated IHC images and rectified semantics. Additionally, we propose a dual-branch discriminator to compel the generator to match the empirical distribution of real images, significantly improving generation quality. Extensive experiments on two public benchmarks demonstrate that GSGStain significantly outperforms state-of-the-art methods in both image quality and pathological consistency. This work establishes a new paradigm for semantically robust virtual staining.

Ziwei Qin, Xuhui Song, Deqing Huang, Na Qin, Jun Li

Graph neural networks are increasingly applied to multimodal medical diagnosis for their inherent relational modeling capabilities. However, their efficacy is often compromised by the prevailing reliance on a single, static graph built from indiscriminate features, hindering the ability to model patient-specific pathological relationships. To this end, the proposed Multi-Activation Plane Interaction Graph Neural Network (MAPI-GNN) reconstructs this single-graph paradigm by learning a multifaceted graph profile from semantically disentangled feature subspaces. The framework first uncovers latent graph-aware patterns via a multi-dimensional discriminator; these patterns then guide the dynamic construction of a stack of activation graphs; and this multifaceted profile is finally aggregated and contextualized by a relational fusion engine for a robust diagnosis. Extensive experiments on two diverse tasks, comprising over 1300 patient samples, demonstrate that MAPI-GNN significantly outperforms state-of-the-art methods.

Jialong Qin, Xin Zou, Di Lu, Yibo Yan, Xuming Hu

Current Video Large Language Models (VideoLLMs) suffer from quadratic computational complexity and key-value cache scaling, due to their reliance on processing excessive redundant visual tokens. To address this problem, we propose SharpV, a minimalist and efficient method for adaptive pruning of visual tokens and KV cache. Different from most uniform compression approaches, SharpV dynamically adjusts pruning ratios based on spatial-temporal information. Remarkably, this adaptive mechanism occasionally achieves performance gains over dense models, offering a novel paradigm for adaptive pruning. During the KV cache pruning stage, based on observations of visual information degradation, SharpV prunes degraded visual features via a self-calibration manner, guided by similarity to original visual features. In this way, SharpV achieves hierarchical cache pruning from the perspective of information bottleneck, offering a new insight into VideoLLMs' information flow. Experiments on multiple public benchmarks demonstrate the superiority of SharpV. Moreover, to the best of our knowledge, SharpV is notably the first two-stage pruning framework that operates without requiring access to exposed attention scores, ensuring full compatibility with hardware acceleration techniques like Flash Attention.

Yuhang Qian, Haiyan Chen, Wentong Li, Ningzhong Liu, Jie Qin

Camouflage Images Generation (CIG) is an emerging research area that focuses on synthesizing images in which objects are harmoniously blended and exhibit high visual consistency with their surroundings. Existing methods perform CIG by either fusing objects into specific backgrounds or outpainting the surroundings via foreground object-guided diffusion. However, they often fail to obtain natural results because they overlook the logical relationship between camouflaged objects and background environments. To address this issue, we propose CT-CIG, a Controllable Text-guided Camouflage Images Generation method that produces realistic and logically plausible camouflage images. Leveraging Large Visual Language Models (VLM), we design a Camouflage-Revealing Dialogue Mechanism (CRDM) to annotate existing camouflage datasets with high-quality text prompts. Subsequently, the constructed image-prompt pairs are utilized to finetune Stable Diffusion, incorporating a lightweight controller to guide the location and shape of camouflaged objects for enhanced camouflage scene fitness. Moreover, we design a Frequency Interaction Refinement Module (FIRM) to capture high-frequency texture features, facilitating the learning of complex camouflage patterns. Extensive experiments, including CLIPScore evaluation and camouflage effectiveness assessment, demonstrate the semantic alignment of our generated text prompts and CT-CIG's ability to produce photorealistic camouflage images.

Rui Qian, Haozhi Cao, Tianchen Deng, Shenghai Yuan, Lihua Xie

Monocular 3D Semantic Scene Completion (SSC) is a challenging yet promising task that aims to infer dense geometric and semantic descriptions of a scene from a single image. While recent object-centric paradigms significantly improve efficiency by leveraging flexible 3D Gaussian primitives, they still rely heavily on a large number of randomly initialized primitives, which inevitably leads to 1) inefficient primitive initialization and 2) outlier primitives that introduce erroneous artifacts. In this paper, we propose SplatSSC, a novel framework that resolves these limitations with a depth-guided initialization strategy and a principled Gaussian aggregator. Instead of random initialization, SplatSSC utilizes a dedicated depth branch composed of a Group-wise Multi-scale Fusion (GMF) module, which integrates multi-scale image and depth features to generate a sparse yet representative set of initial Gaussian primitives. To mitigate noise from outlier primitives, we develop the Decoupled Gaussian Aggregator (DGA), which enhances robustness by decomposing geometric and semantic predictions during the Gaussian-to-voxel splatting process. Complemented with a specialized Probability Scale Loss, our method achieves state-of-the-art performance on the Occ-ScanNet dataset, outperforming prior approaches by over 6.3% in IoU and 4.1% in mIoU, while reducing both latency and memory cost by more than 9.3%.