# gaussian splatting

Gaussian splatting is a rendering technique that represents scenes with 3D Gaussians and uses visibility-aware anisotropic splatting for real-time novel-view synthesis.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Multi-View Transformer with Geometry-Guided Multi-Modal ControlNet for 3D Virtual Try-On

DevFeed: [Multi-View Transformer with Geometry-Guided Multi-Modal ControlNet for 3D Virtual Try-On](<https://devfeed.tech/articles/multi-view-transformer-with-geometry-guided-multi-modal-controlnet-for-3d-virtual-try-on-145615.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25979>)

Author: Sarthak Batra, Armin Mustafa, Adrian Hilton

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Image](<https://devfeed.tech/topics/image.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [3dgs](<https://devfeed.tech/tags/3dgs.md>), [attention](<https://devfeed.tech/tags/attention.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [camera](<https://devfeed.tech/tags/camera.md>), [consistency](<https://devfeed.tech/tags/consistency.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [image](<https://devfeed.tech/tags/image.md>), [modal](<https://devfeed.tech/tags/modal.md>), [proximity](<https://devfeed.tech/tags/proximity.md>), [reference](<https://devfeed.tech/tags/reference.md>), [self-attention](<https://devfeed.tech/tags/self-attention.md>), [structure](<https://devfeed.tech/tags/structure.md>), [training](<https://devfeed.tech/tags/training.md>), [transformer](<https://devfeed.tech/tags/transformer.md>), [view](<https://devfeed.tech/tags/view.md>), [virtual-try-on](<https://devfeed.tech/tags/virtual-try-on.md>)

### AI overview

GS-VTON-MVT aims to keep garment details consistent across views when editing a person captured as a 3D Gaussian Splatting scene. It jointly inpaints training views using cross-view self-attention, conditions its diffusion editor on the garment image plus depth and surface normals rendered from the current 3DGS, and adapts attention between current and reference views according to camera-pose proximity. The approach addresses inconsistencies associated with independently edited views and fixed, hand-designed fusion. On 3D-VTONBench, the authors report improvements in FID, KID, LPIPS-based cross-view consistency, and user preference over general-purpose 3D editors and earlier 3D try-on methods; the excerpt gives no metric values.

### Source excerpt

arXiv:2611.25979v1 Announce Type: cross Abstract: Editing a 3D Gaussian Splatting (3DGS) capture of a person so that they wear a new garment requires every rendered view to agree on the same garment details. Existing 3D virtual try-on (VTON) pipelines, however, edit training views largely independently and fuse them with hand-designed heuristics such as manually chosen reference views and fixed attention-blending weights. The resulting inconsistencies are baked into the 3D representation and surface as texture drift, blurred prints, and implausible geometry in novel views. We present GS-VTON-MVT, a three-stage framework that combines implicit and explicit 3D reasoning for consistent 3DGS try-on. (i) A multi-view inpainting transformer jointly edits a set of training views with full cross-view self-attention, so that cross-view correspondence is learned rather than imposed. (ii) A geometry-guided multi-modal ControlNet conditions the diffusion editor on the garment image together with depth and surface normals rendered from the current 3DGS state, supplying explicit structure where visual evidence is weakest. (iii) A pose-adaptive persona-aware attention replaces the fixed blending weight between current-view and reference-view features with weights derived from camera-pose proximity. On the 3D-VTONBench benchmark, GS-VTON-MVT improves FID, KID, LPIPS-based cross-view consistency, and user preference over general-purpose 3D editors and prior 3D try-on methods.

## GaussTrans: A Gaussian Interface Framework for Multi-Layer Refractive Transport

DevFeed: [GaussTrans: A Gaussian Interface Framework for Multi-Layer Refractive Transport](<https://devfeed.tech/articles/gausstrans-a-gaussian-interface-framework-for-multi-layer-refractive-transport-145612.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25950>)

Author: Zixuan Zhao, Tao Wang

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>)

Tags: [approach](<https://devfeed.tech/tags/approach.md>), [combines](<https://devfeed.tech/tags/combines.md>), [components](<https://devfeed.tech/tags/components.md>), [coupling](<https://devfeed.tech/tags/coupling.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [errors](<https://devfeed.tech/tags/errors.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [field](<https://devfeed.tech/tags/field.md>), [fields](<https://devfeed.tech/tags/fields.md>), [framework](<https://devfeed.tech/tags/framework.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [interface](<https://devfeed.tech/tags/interface.md>), [interfaces](<https://devfeed.tech/tags/interfaces.md>), [intersection](<https://devfeed.tech/tags/intersection.md>), [layer](<https://devfeed.tech/tags/layer.md>), [layers](<https://devfeed.tech/tags/layers.md>), [light](<https://devfeed.tech/tags/light.md>), [reduce](<https://devfeed.tech/tags/reduce.md>), [transport](<https://devfeed.tech/tags/transport.md>)

### AI overview

GaussTrans targets accurate light transmission reconstruction in scenes with many transparent layers, where geometric, light-path, and observation errors interact. Its three components combine Gaussian surfaces and fields to model interfaces, refine reconstructions by evaluating the full light path, and use primitive support regions to reduce interface-ray intersection errors. The abstract reports significantly higher accuracy than existing approaches on datasets including RefRef, with relatively low spatial storage requirements; it provides no numerical results or comparison details.

### Source excerpt

arXiv:2611.25950v1 Announce Type: new Abstract: We investigate surface modeling of multi-transparent interfaces to achieve accurate light transmission. The interplay of errors in geometric structure, light paths, and observation results during the reconstruction of transparent media with a high number of layers makes this a challenging problem. Some existing techniques address the reconstruction of transparent scenes by reducing the number of layers or simplifying light paths; however, this limits their direct applicability to high-order transparent scenes. We propose GaussTrans, a compact framework for light transport in high-order transparent scenes, consisting of three components: Gaussian-Field Hybrid Surfaces, Transport-Coupled Refinement, and Primitive-Guided Implicit Intersection. Gaussian-Field Hybrid Surfaces use a hybrid approach combining Gauss surfaces and fields to reconstruct different interfaces. Transport-Coupled Refinement reduces coupling errors by evaluating the entire light path. Primitive-Guided Implicit Intersection combines primitive support regions to reduce errors in the intersection of interfaces and rays. In experiments on datasets such as RefRef, GaussTrans achieves significantly higher accuracy with relatively low spatial storage requirements.

## MANGO-Grasp: Mahalanobis Fields over Geometry-Oriented 3D Gaussians for Cross-Embodiment Dexterous Grasping

DevFeed: [MANGO-Grasp: Mahalanobis Fields over Geometry-Oriented 3D Gaussians for Cross-Embodiment Dexterous Grasping](<https://devfeed.tech/articles/mango-grasp-mahalanobis-fields-over-geometry-oriented-3d-gaussians-for-cross-embodiment-dexterous-grasping-142548.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2608.02014>)

Author: Heng Zhang, Kevin Yuchen Ma, Mike Zheng Shou, Weisi Lin, Yan Wu

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [hydroelastic contact model](<https://devfeed.tech/topics/hydroelastic-contact-model.md>), [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cross](<https://devfeed.tech/tags/cross.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [fields](<https://devfeed.tech/tags/fields.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [kinematics](<https://devfeed.tech/tags/kinematics.md>), [local](<https://devfeed.tech/tags/local.md>), [modeling](<https://devfeed.tech/tags/modeling.md>), [object](<https://devfeed.tech/tags/object.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [stable](<https://devfeed.tech/tags/stable.md>), [surface](<https://devfeed.tech/tags/surface.md>), [zero-shot](<https://devfeed.tech/tags/zero-shot.md>)

### AI overview

MANGO-Grasp targets stable grasp synthesis across different multi-fingered hands with little or no embodiment-specific tuning. It represents objects as geometry-oriented 3D Gaussian primitives to preserve local surface detail, encodes hand morphology and kinematics in pretrained descriptors, and uses anisotropic Mahalanobis fields to model directional contact and guide shared grasp optimization. In simulation on CMAP and MultiGripperGrasp, it exceeds the strongest seen-hand baseline by up to 8.24 percentage points. On the unseen SharpaWave hand, it improves over the strongest zero-shot baseline by up to 16.57 points, and reports 86% success in real-world experiments.

### Source excerpt

arXiv:2608.02014v3 Announce Type: replace-cross Abstract: Cross-embodiment dexterous grasping aims to synthesize stable grasps across heterogeneous multi-fingered hands with little or no embodiment-specific tuning. Existing interaction-centric methods achieve promising results, but their object representations often underrepresent local surface geometry, hand descriptors do not explicitly capture both morphology and kinematics, and hand-object interaction modeling is typically isotropic. We propose MANGO-Grasp, an anisotropic interaction framework that represents objects as geometry-oriented 3D Gaussian primitives to better preserve local surface geometry, and encodes hands into pretrained descriptors that capture both morphology and kinematics. Hand-object interaction is modeled using anisotropic Mahalanobis fields that reflect the directional nature of contact and guide a shared optimization for grasp realization across embodiments. On the CMAP and MultiGripperGrasp benchmarks, MANGO-Grasp outperforms the strongest seen-hand baseline by up to 8.24 percentage points in simulation. It also transfers zero-shot to the unseen SharpaWave hand, improving over the strongest zero-shot baseline by up to 16.57 percentage points, and achieves 86% success in real-world experiments. The code and additional materials are available at https://connor-zh.github.io/MANGO-Grasp/.

## ReTouchGS: Ultralight Render-free Aesthetic Color Enhancement and Transfer for 3D Gaussian Splatting

DevFeed: [ReTouchGS: Ultralight Render-free Aesthetic Color Enhancement and Transfer for 3D Gaussian Splatting](<https://devfeed.tech/articles/retouchgs-ultralight-render-free-aesthetic-color-enhancement-and-transfer-for-3d-gaussian-splatting-140110.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23054>)

Author: Chuanzhi Xu, Boyu Wei, Chengkun Yue, Xuanhua Yin, Yiyang Li, Haodong Chen, Qiang Qu, Weidong Cai

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [color](<https://devfeed.tech/topics/color.md>), [image editing](<https://devfeed.tech/topics/image-editing.md>)

Tags: [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [color](<https://devfeed.tech/tags/color.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-gr](<https://devfeed.tech/tags/cs-gr.md>), [enhancement](<https://devfeed.tech/tags/enhancement.md>)

### AI overview

ReTouchGS applies scene-level color changes to pretrained 3D Gaussian Splatting models without rendering, preserving geometry and producing standard assets in seconds. For reference-free aesthetic enhancement, it predicts a color lookup table (LUT) from the model and camera parameters. For reference-guided transfer, it fits a scene-level LUT from one edited view and its corresponding original render. A shared blend-aware refinement network with 44.7K parameters refines an analytic update for either mode. The authors report high color accuracy and generalization to unseen scenes, and introduce ReTouch3D for evaluating both tasks across viewpoints.

### Source excerpt

arXiv:2611.23054v1 Announce Type: cross Abstract: 3D Gaussian Splatting (3DGS) enables photorealistic scene reconstruction. Existing 3DGS color transfer and interactive editing methods are guided by reference images and user-specified color adjustments. However, in many applications, users simply want aesthetically guided color enhancement that automatically makes the scene more visually appealing while maintaining consistent colors and tones across viewpoints. We present ReTouchGS, to our knowledge the first framework for reference-free aesthetic color enhancement of pretrained 3DGS models. It predicts a scene-level color lookup table (LUT) from the model and its associated camera parameters. ReTouchGS also supports reference-guided color transfer by fitting a scene-level LUT using one edited view of the same scene and its corresponding original render. A shared blend-aware color refinement network with only 44.7K parameters refines an analytic color update to realize either color mapping without rendering. It preserves geometry and delivers standard 3DGS assets in seconds. We also design the ReTouch3D benchmark to support the evaluation of aesthetic color enhancement and reference-guided color transfer across viewpoints. Experiments demonstrate high color accuracy and strong generalization to unseen scenes in both aesthetic color enhancement and reference-guided color transfer.

## Making the World React Realistically: A Closed-Loop Interactive World Model for Autonomous Driving

DevFeed: [Making the World React Realistically: A Closed-Loop Interactive World Model for Autonomous Driving](<https://devfeed.tech/articles/making-the-world-react-realistically-a-closed-loop-interactive-world-model-for-autonomous-driving-140067.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22691>)

Author: Jia Hu, Yang Chang, Panqu Wang, Zhijun Fu, Chang Liu, Changjian Yu, Jintao Lai

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [autonomous vehicles](<https://devfeed.tech/topics/autonomous-vehicles.md>), [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [autonomous-driving](<https://devfeed.tech/tags/autonomous-driving.md>), [behavior](<https://devfeed.tech/tags/behavior.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [rendering](<https://devfeed.tech/tags/rendering.md>), [vehicle](<https://devfeed.tech/tags/vehicle.md>), [vehicles](<https://devfeed.tech/tags/vehicles.md>), [world-model](<https://devfeed.tech/tags/world-model.md>)

### AI overview

The proposed closed-loop world model couples background-vehicle behavior with the images an ego planner sees. It fine-tunes imitation-learned vehicle policies through multi-agent reinforcement learning against an image-reactive planner; uses interaction viewpoints to reduce view-dependent color on under-observed 3D Gaussians and fill unseen vehicle surfaces with generated Gaussians; and edits signal lights to reflect driving intentions, learning unobserved states through counterfactual fine-tuning with a light-state classifier. In long-tail closed-loop scenarios, it reports 85% lower divergence from real interaction behavior and a 2.0 dB gain in novel-view vehicle PSNR over the state-of-the-art world model. Rendered vehicle signals indirectly affect ego behavior in a quarter of scenarios.

### Source excerpt

arXiv:2611.22691v1 Announce Type: cross Abstract: World models (WMs) generate driving scenarios beyond recorded logs, enabling the training and evaluation of autonomous driving systems in unseen interactions. However, existing WMs struggle to ensure both behavioral realism and visual fidelity, since they overlook the relationship between vehicle interactions and the ego visual observations. We propose a closed-loop interactive WM that couples behavior modeling and visual rendering. First, imitation-learned background-vehicle (BV) policies are fine-tuned with multi-agent reinforcement learning in the closed loop against an end-to-end ego planner that reacts to the rendered images, so that their interaction behaviors match those in real logs. Second, the viewpoints exposed by these interactions guide a per-Gaussian diagnosis of each 3D Gaussian vehicle asset, which attenuates view-dependent colors on under-observed Gaussians and fills never-observed surfaces with generated Gaussians, while preserving the observed appearance. Third, signal lights are rendered according to the driving intentions of the vehicles, by an editor that learns the light states unobserved for a vehicle through counterfactual fine-tuning with a light-state classifier. On long-tail closed-loop scenarios, our method reduces the divergence from real-world interaction behaviors by 85% and improves novel-view vehicle PSNR by 2.0dB over the state-of-the-art WM. The significance of BV light signaling in WMs is shown, as the rendered signal lights conveying BV driving intentions indirectly impact ego driving behaviors in a quarter of the scenarios.

## A Gaussian Primitive Data Management Framework for Efficient 3DGS Trajectory Rendering

DevFeed: [A Gaussian Primitive Data Management Framework for Efficient 3DGS Trajectory Rendering](<https://devfeed.tech/articles/a-gaussian-primitive-data-management-framework-for-efficient-3dgs-trajectory-rendering-138299.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20546>)

Author: Haolun Ma, Ziyi Liu, Zizhuo Xu, Chuhan Hu, Lei Li

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Databases](<https://devfeed.tech/sources/arxiv-databases.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [3D rendering](<https://devfeed.tech/topics/3d-rendering.md>), [Data Management](<https://devfeed.tech/topics/data-management.md>)

Tags: [3dgs](<https://devfeed.tech/tags/3dgs.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-db](<https://devfeed.tech/tags/cs-db.md>), [cs-gr](<https://devfeed.tech/tags/cs-gr.md>), [data-management](<https://devfeed.tech/tags/data-management.md>), [framework](<https://devfeed.tech/tags/framework.md>), [latency](<https://devfeed.tech/tags/latency.md>), [rendering](<https://devfeed.tech/tags/rendering.md>), [spatial-data](<https://devfeed.tech/tags/spatial-data.md>)

### AI overview

GDM-GS targets redundant work in trajectory rendering with Structured 3D Gaussian Splatting (3DGS), where each view projects all anchors, may decode anchors already needed by neighboring views, and runs GPU steps while the CPU idles. It treats views as spatial range queries and decoded Gaussians as shared derived data. Within each view, anchor and occluder indexes form a holed-frustum query to avoid decoding out-of-view or provably hidden anchors. Across consecutive views, shared materialization decodes each requested anchor once per group while preserving per-view selection. A hybrid pipeline balances selection between CPU and GPU before grouped decoding under bounded memory. Experiments on large-scale public scenes reportedly reduce trajectory-rendering latency substantially while preserving image quality; the excerpt gives no quantitative results.

### Source excerpt

arXiv:2611.20546v1 Announce Type: new Abstract: Trajectory rendering, which renders novel views of a reconstructed 3D scene along a camera path, is a core operation in digital twins and autonomous inspection. Structured 3D Gaussian Splatting (3DGS) renders a single view in real time by decoding sparse anchors into Gaussians on demand, but it renders a trajectory view by view: every view projects all anchors to select them, re-decodes anchors already requested by neighboring views, and runs all steps on the GPU while the CPU idles. We treat this as a spatial data management problem, where anchors are spatial objects, views are range queries, and decoded Gaussians are shared derived data, and propose GDM-GS, which removes redundancy at three levels. Within a view, an anchor index and an occluder index answer each view as a holed-frustum range query, so out-of-view and provably hidden anchors are never decoded. Across views, shared materialization decodes each requested anchor once per group of consecutive views while preserving each view's selection. Across processors, a hybrid pipeline balances selections between the CPU and the GPU before grouped decoding under bounded memory. Experiments on large-scale public scenes show that GDM-GS substantially reduces trajectory rendering latency while preserving image quality.

## The Mask Is Not the Object: Volumetric Supervision for 3D Gaussian Segmentation

DevFeed: [The Mask Is Not the Object: Volumetric Supervision for 3D Gaussian Segmentation](<https://devfeed.tech/articles/the-mask-is-not-the-object-volumetric-supervision-for-3d-gaussian-segmentation-137691.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.21383>)

Author: Jeonghwan Cho, Minsu Kim, Junyoung Hong, Seon Joo Kim

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Ground truth / benchmark quality](<https://devfeed.tech/topics/ground-truth-benchmark-quality.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [alpha](<https://devfeed.tech/tags/alpha.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [changes](<https://devfeed.tech/tags/changes.md>), [clusters](<https://devfeed.tech/tags/clusters.md>), [completeness](<https://devfeed.tech/tags/completeness.md>), [contribution](<https://devfeed.tech/tags/contribution.md>), [contributors](<https://devfeed.tech/tags/contributors.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [delivery](<https://devfeed.tech/tags/delivery.md>), [error](<https://devfeed.tech/tags/error.md>), [feature](<https://devfeed.tech/tags/feature.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [gaussian-splatting](<https://devfeed.tech/tags/gaussian-splatting.md>), [gradient](<https://devfeed.tech/tags/gradient.md>), [learning](<https://devfeed.tech/tags/learning.md>), [mask](<https://devfeed.tech/tags/mask.md>), [object](<https://devfeed.tech/tags/object.md>), [supervision](<https://devfeed.tech/tags/supervision.md>)

### AI overview

VODA targets incomplete 3D Gaussian selections that can still produce accurate rendered masks: alpha-blended feature supervision favors dominant contributors and leaves occluded or low-contribution Gaussians weakly supervised. Its plug-and-play objective uses local depth structure to group per-pixel contributors and directly supervise all Gaussians in selected depth clusters. The authors introduce Removal-3D, a synthetic object-removal benchmark, and Removal Depth Error (RDE), which compares predicted and ground-truth depth changes. Across four rendering-based baselines, VODA improves volumetric segmentation on Removal-3D while maintaining or improving performance on LERF and ScanNet. The abstract reports that simple gradient reweighting and spatial propagation do not achieve comparable gains.

### Source excerpt

arXiv:2611.21383v1 Announce Type: cross Abstract: 3D Gaussian Splatting enables object-level operations on Gaussian primitives, yet existing segmentation methods are largely trained and evaluated through rendered views. This mismatch can produce accurate rendered masks from incomplete 3D selections since alpha-blended feature supervision concentrates learning on dominant contributors while leaving occluded or low-contribution Gaussians weakly supervised. We propose VODA, a plug-and-play supervision objective that exploits local depth structure to group per-pixel contributors and directly supervises all Gaussians in selected depth clusters, bypassing contribution-weighted gradient delivery. To evaluate volumetric completeness, we introduce Removal-3D, a synthetic object-removal benchmark, and the Removal Depth Error (RDE), a metric comparing predicted and ground-truth depth changes. Across four rendering-based baselines, VODA consistently improves volumetric segmentation on Removal-3D while maintaining or improving performance on LERF and ScanNet. Further analysis shows that VODA recovers lower-contribution Gaussians, while simple gradient reweighting and spatial propagation fail to reproduce comparable gains.

## Prepare Lighter and Faster for Open-Vocabulary Queries: A Query-Wise 3D Segmenter for Gaussian Splatting

DevFeed: [Prepare Lighter and Faster for Open-Vocabulary Queries: A Query-Wise 3D Segmenter for Gaussian Splatting](<https://devfeed.tech/articles/prepare-lighter-and-faster-for-open-vocabulary-queries-a-query-wise-3d-segmenter-for-gaussian-splatting-137447.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19429>)

Author: Nhat-Quynh Le-Pham, Khoi Nguyen, Binh-Son Hua

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [Machine Perception](<https://devfeed.tech/topics/machine-perception.md>), [3D](<https://devfeed.tech/topics/3d.md>)

Tags: [2d](<https://devfeed.tech/tags/2d.md>), [3d](<https://devfeed.tech/tags/3d.md>), [clip](<https://devfeed.tech/tags/clip.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-gr](<https://devfeed.tech/tags/cs-gr.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [gaussian-splatting](<https://devfeed.tech/tags/gaussian-splatting.md>), [lighter](<https://devfeed.tech/tags/lighter.md>), [lightweight](<https://devfeed.tech/tags/lightweight.md>), [localization](<https://devfeed.tech/tags/localization.md>), [mask](<https://devfeed.tech/tags/mask.md>), [object](<https://devfeed.tech/tags/object.md>), [performance](<https://devfeed.tech/tags/performance.md>), [queries](<https://devfeed.tech/tags/queries.md>), [query](<https://devfeed.tech/tags/query.md>), [text](<https://devfeed.tech/tags/text.md>), [training](<https://devfeed.tech/tags/training.md>), [video](<https://devfeed.tech/tags/video.md>), [visibility](<https://devfeed.tech/tags/visibility.md>), [voting](<https://devfeed.tech/tags/voting.md>)

### AI overview

The framework segments 3D instances in Gaussian Splatting scenes from text queries without per-scene training. It uses dense pixel-level CLIP features for fine-grained localization, selects a target region in a pivot view, propagates it across views with a video segmentor, and maps the resulting 2D tracks to 3D Gaussians through visibility voting. The authors report state-of-the-art results on LERF-OVS, ScanNet, and ScanNet++, competitive results on 3D-OVS, and significant speedups over existing methods. The supplied text is an abstract and does not detail runtime, preprocessing requirements, or the limits of these comparisons.

### Source excerpt

arXiv:2611.19429v1 Announce Type: cross Abstract: Open-vocabulary 3D understanding aims to localize and segment arbitrary objects beyond predefined categories. However, existing approaches based on Gaussian Splatting often rely on expensive preprocessing and lengthy training, limiting their applicability when fast and lightweight solutions are desired. We propose a lightweight, training-free framework for query-driven 3D instance segmentation on Gaussian Splatting. Instead of conventional mask-wise CLIP embeddings, we use dense pixel-level CLIP features to preserve fine-grained semantic information and enable flexible text-driven localization without relying on predefined instance proposals. Given a text query, we identify a high-quality target region in a carefully selected pivot view and propagate it across views using a video segmentor, providing consistent and comprehensive coverage of the target instance. We then lift the resulting 2D tracklets to 3D Gaussians through a novel visibility voting strategy, yielding accurate 3D segmentation masks with clean object boundaries. Extensive experiments demonstrate state-of-the-art performance on LERF-OVS, ScanNet, and ScanNet++, competitive results on 3D-OVS, and significant speedups over existing open-vocabulary 3D segmentation methods. These results highlight the effectiveness of our framework for efficient open-vocabulary 3D understanding without per-scene training.

## Instant NuRec: Feed-Forward 3D Gaussian Reconstruction for Driving Scene Simulation

DevFeed: [Instant NuRec: Feed-Forward 3D Gaussian Reconstruction for Driving Scene Simulation](<https://devfeed.tech/articles/instant-nurec-feed-forward-3d-gaussian-reconstruction-for-driving-scene-simulation-131951.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2607.14203>)

Author: NVIDIA, :, Jiahui Huang, Jiawei Ren, Michal Tyszkiewicz, Bjoern Haefner, Michael Shelley, Xin Kang, Seung Wook Kim, Ning Xu, Qi Wu, Janick Martinez Esturo, Shengyu Huang, Nick Schneider, Laura Leal-Ta

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [Simulation](<https://devfeed.tech/topics/simulation.md>), [3D rendering](<https://devfeed.tech/topics/3d-rendering.md>)

Tags: [1-5](<https://devfeed.tech/tags/1-5.md>), [3d](<https://devfeed.tech/tags/3d.md>), [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [3dgs](<https://devfeed.tech/tags/3dgs.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-driving](<https://devfeed.tech/tags/autonomous-driving.md>), [camera](<https://devfeed.tech/tags/camera.md>), [central](<https://devfeed.tech/tags/central.md>), [costs](<https://devfeed.tech/tags/costs.md>), [critical](<https://devfeed.tech/tags/critical.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-gr](<https://devfeed.tech/tags/cs-gr.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [development](<https://devfeed.tech/tags/development.md>), [driving](<https://devfeed.tech/tags/driving.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [gaussian-splatting](<https://devfeed.tech/tags/gaussian-splatting.md>), [instant](<https://devfeed.tech/tags/instant.md>), [simulation](<https://devfeed.tech/tags/simulation.md>)

### AI overview

Instant NuRec converts a short, calibrated multi-camera driving log into a simulatable 3D Gaussian Splatting scene in one forward pass, avoiding the per-scene tuning and slower reconstruction described for existing neural simulation methods. It outputs separate static and dynamic 3DGS layers, a sky cubemap, and per-camera ISP corrections, and supports non-pinhole cameras through 3DGUT. The abstract reports reconstructing a 10-20-second scene in about 1.5 seconds and a Waymo Open Dataset PSNR 2.01 dB above the strongest evaluated baseline. It is integrated with NuRec and compatible with AlpaSim for closed-loop simulation.

### Source excerpt

arXiv:2607.14203v2 Announce Type: replace-cross Abstract: 3D simulation platforms are critical for autonomous driving because they enable end-to-end policy evaluation, thereby reducing development costs and improving safety. In recent years, neural simulation has become predominant, with methods such as NuRec playing a central role; however, these methods remain relatively slow and typically require per-scene tuning. In this work, we present Instant NuRec, a feed-forward neural reconstruction model that turns a short multi-view driving log into a fully simulatable 3D Gaussian Splatting (3DGS) world in a single forward pass. The model accepts multi-view input from a calibrated camera rig and emits a layered output consisting of static and dynamic 3DGS layers, a sky cubemap, and per-camera ISP corrections, while providing native support for non-pinhole camera models via 3DGUT. It reconstructs a 10-20-second multi-camera scene in roughly 1.5 seconds and achieves a PSNR on the Waymo Open Dataset that is 2.01 dB above the strongest evaluated baseline. Instant NuRec is deeply integrated into NuRec and is compatible with AlpaSim for closed-loop simulation.

## Pygmalion Effect in Vision: Image-to-Clay Translation for Reflective Geometry Reconstruction

DevFeed: [Pygmalion Effect in Vision: Image-to-Clay Translation for Reflective Geometry Reconstruction](<https://devfeed.tech/articles/pygmalion-effect-in-vision-image-to-clay-translation-for-reflective-geometry-reconstruction-128074.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2511.21098>)

Author: Gayoung Lee, Junho Kim, Jin-Hwa Kim, Junmo Kim

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [image distortion](<https://devfeed.tech/topics/image-distortion.md>), [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Image](<https://devfeed.tech/topics/image.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3d-reconstruction](<https://devfeed.tech/tags/3d-reconstruction.md>), [computer-vision](<https://devfeed.tech/tags/computer-vision.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-gr](<https://devfeed.tech/tags/cs-gr.md>), [translation](<https://devfeed.tech/tags/translation.md>), [vision](<https://devfeed.tech/tags/vision.md>)

### AI overview

The framework addresses reflective-geometry reconstruction by translating images into clay-like views that reduce reflection cues. It uses two independent rendering branches over shared Gaussian geometry: a BRDF-based reflective branch with photometric supervision, and a clay-guided branch supervised by synthesized clay images to provide reflection-free geometric guidance. Experiments on synthetic and real datasets reportedly show consistent improvements in geometry and photometric metrics. The supplied excerpt does not specify datasets, metric values, or implementation details.

### Source excerpt

arXiv:2511.21098v2 Announce Type: replace-cross Abstract: Understanding reflection remains a long-standing challenge in 3D reconstruction due to the entanglement of appearance and geometry under view-dependent reflections. In this work, we present the Pygmalion Effect in Vision, a novel framework that metaphorically "sculpts" reflective objects into clay-like forms through image-to-clay translation. We then introduce a dual-branch design in which a BRDF-based reflective branch and a clay-guided branch share the same Gaussian geometry but operate through independent rendering paths. The clay-guided branch, supervised by the synthesized clay images, provides reflection-free geometric guidance that complements the photometric supervision of the BRDF branch. Experiments on both synthetic and real datasets show consistent improvements in both geometry and photometric metrics. Beyond technical gains, our framework reveals that seeing by unshining, translating radiance into neutrality, can serve as a powerful inductive bias for reflective object geometry learning.

## FastGRT: Accelerating Training of 3D Gaussian Ray Tracing

DevFeed: [FastGRT: Accelerating Training of 3D Gaussian Ray Tracing](<https://devfeed.tech/articles/fastgrt-accelerating-training-of-3d-gaussian-ray-tracing-125416.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.11586>)

Author: Meet Banthia, Xiaoteng Liu, Hexu Zhao, Aurojit Panda, Jinyang Li

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [Ray Tracing / Path Tracing](<https://devfeed.tech/topics/ray-tracing-path-tracing.md>), [Content Creation / Rendering](<https://devfeed.tech/topics/content-creation-rendering.md>), [real-time rendering](<https://devfeed.tech/topics/real-time-rendering.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [3dgs](<https://devfeed.tech/tags/3dgs.md>), [blending](<https://devfeed.tech/tags/blending.md>), [bottlenecks](<https://devfeed.tech/tags/bottlenecks.md>), [code](<https://devfeed.tech/tags/code.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [develop](<https://devfeed.tech/tags/develop.md>), [fast](<https://devfeed.tech/tags/fast.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [gradient](<https://devfeed.tech/tags/gradient.md>), [gradient-accumulation](<https://devfeed.tech/tags/gradient-accumulation.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [implement](<https://devfeed.tech/tags/implement.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [optimizations](<https://devfeed.tech/tags/optimizations.md>), [ray-tracing](<https://devfeed.tech/tags/ray-tracing.md>), [rendering](<https://devfeed.tech/tags/rendering.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

FastGRT reduces the training cost of Gaussian-based ray tracing by reorganizing GPU work: it separates ray traversal from blending and reuses forward-pass work during backpropagation. Fast gradient accumulation and a hybrid BVH update strategy address additional pipeline bottlenecks. Implemented in Vulkan, it supports both training and rendering; the authors describe it as the first Vulkan-based GRT implementation with both capabilities. Compared with 3DGRT, it reports 1.9x-4.97x faster training while preserving the original rendering formulation. The supplied text is an abstract and does not provide further evaluation details.

### Source excerpt

arXiv:2611.11586v1 Announce Type: cross Abstract: Ray Tracing enables rendering capabilities that are difficult to achieve with rasterization-based approaches such as 3D Gaussian Splatting (3DGS). However, the training cost of Gaussian-based ray tracing (GRT) is much higher than that of 3DGS. In this paper, we study how ray tracing using Gaussians executes on GPUs and show that a considerable fraction of the training cost arises from how the computation is organized. We develop FastGRT, which decouples ray traversal from blending and reuses work done in the forward pass for the backward pass. FastGRT further incorporates optimizations such as fast gradient accumulation and a hybrid BVH update strategy to tackle other major bottlenecks in the training pipeline. We implement FastGRT in Vulkan. To the best of our knowledge, it is the first Vulkan-based GRT implementation that supports both training and rendering. Compared to 3DGRT, FastGRT improves training time by 1.9$\times$ to 4.97$\times$, while preserving the original rendering formulation. Our code is open source.

## A four-view smartphone video dataset of a robot-held flexible strip, with camera poses and a 3D LaGSplat baseline

DevFeed: [A four-view smartphone video dataset of a robot-held flexible strip, with camera poses and a 3D LaGSplat baseline](<https://devfeed.tech/articles/a-four-view-smartphone-video-dataset-of-a-robot-held-flexible-strip-with-camera-poses-and-a-3d-lagsplat-baseline-125116.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.09959>)

Author: Louen Pottier

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [3D rendering](<https://devfeed.tech/topics/3d-rendering.md>), [real-time 3D](<https://devfeed.tech/topics/real-time-3d.md>), [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [acceleration](<https://devfeed.tech/tags/acceleration.md>), [arm](<https://devfeed.tech/tags/arm.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [browser](<https://devfeed.tech/tags/browser.md>), [camera](<https://devfeed.tech/tags/camera.md>), [cameras](<https://devfeed.tech/tags/cameras.md>), [code](<https://devfeed.tech/tags/code.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [data](<https://devfeed.tech/tags/data.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [decoder](<https://devfeed.tech/tags/decoder.md>), [demo](<https://devfeed.tech/tags/demo.md>), [film](<https://devfeed.tech/tags/film.md>), [flexible](<https://devfeed.tech/tags/flexible.md>), [robot](<https://devfeed.tech/tags/robot.md>), [smartphone](<https://devfeed.tech/tags/smartphone.md>), [video](<https://devfeed.tech/tags/video.md>), [view](<https://devfeed.tech/tags/view.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

The note presents synchronized four-view smartphone footage of a robot-driven flexible strip, with camera poses and per-frame gripper position as the measured input. Three iPhones are fixed and one is handheld; recording lasts ten minutes. It applies LaGSplat, whose video-learned latent state serves both as the generalized coordinate of a dissipative Lagrangian and as input to a Gaussian Splatting decoder. Using gripper acceleration as a forcing term, the model renders held-out views at 26-33 dB PSNR and supports a real-time browser demo. The dataset, checkpoints, and training code are public.

### Source excerpt

arXiv:2611.09959v1 Announce Type: cross Abstract: This note describes a multi-view video dataset of a deformable object driven by a measured input, with the measured pose of every camera, and a first 3D result obtained on it with LaGSplat. LaGSplat learns from video a low-dimensional latent state that is both the generalized coordinate of a learned dissipative Lagrangian and the conditioning variable of a Gaussian Splatting decoder; fitting the decoder in 3D calls for synchronized multi-view footage, which the original paper only had from synthetic renderings. Here a robot arm moves a flexible strip by oscillating along a one-dimensional path, and four iPhones, three fixed and one handheld, film it for ten minutes. The videos are synchronized, the cameras are posed from the scene, and the position of the gripper is given per frame as the input of the system. LaGSplat fitted on this data with the gripper acceleration as a forcing term renders held-out views at 26 to 33 dB PSNR and runs as a real-time interactive demo in a web browser. Dataset, checkpoints and training code are public.

## Mitigating SM-Level Load Imbalance in 3D Gaussian Splatting on GPUs

DevFeed: [Mitigating SM-Level Load Imbalance in 3D Gaussian Splatting on GPUs](<https://devfeed.tech/articles/mitigating-sm-level-load-imbalance-in-3d-gaussian-splatting-on-gpus-125543.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.12319>)

Author: Junha Ryu, Hoi-Jun Yoo

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [GPU optimization](<https://devfeed.tech/topics/gpu-optimization.md>), [scheduling](<https://devfeed.tech/topics/scheduling.md>), [3D rendering](<https://devfeed.tech/topics/3d-rendering.md>), [real-time 3D](<https://devfeed.tech/topics/real-time-3d.md>), [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>)

Tags: [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [3dgs](<https://devfeed.tech/tags/3dgs.md>), [autonomous-driving](<https://devfeed.tech/tags/autonomous-driving.md>), [blending](<https://devfeed.tech/tags/blending.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [cs-gr](<https://devfeed.tech/tags/cs-gr.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [gpu-scheduling](<https://devfeed.tech/tags/gpu-scheduling.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [kernel](<https://devfeed.tech/tags/kernel.md>), [lightweight](<https://devfeed.tech/tags/lightweight.md>), [load](<https://devfeed.tech/tags/load.md>), [multiple](<https://devfeed.tech/tags/multiple.md>), [outlier](<https://devfeed.tech/tags/outlier.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [rendering](<https://devfeed.tech/tags/rendering.md>), [schedule](<https://devfeed.tech/tags/schedule.md>), [sm](<https://devfeed.tech/tags/sm.md>), [statistics](<https://devfeed.tech/tags/statistics.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

Uneven work across screen tiles creates a long execution tail in GPU-based 3D Gaussian Splatting: a few SMs process heavy tiles while most sit idle, especially in large urban scenes. The proposed software scheduling approach uses a persistent kernel to prioritize heavy tiles and selectively splits outlier tiles across multiple CTAs. A per-frame policy uses tile statistics already available in the rendering pipeline to decide when and how much to split. The method requires neither retraining nor specialized hardware, preserves the original alpha blending formulation, and is described as compatible with existing optimizations across 3DGS variants. The abstract reports up to 2.93x average per-scene acceleration for alpha blending on a large urban scene and up to 5.6x on its most imbalanced frames.

### Source excerpt

arXiv:2611.12319v1 Announce Type: cross Abstract: 3D Gaussian Splatting (3DGS) enables photorealistic, real-time rendering on commodity GPUs and is increasingly adopted in generative world models and autonomous driving simulators. However, existing optimizations rely on algorithmic changes that often require retraining, or on specialized accelerators that lack flexibility across evolving 3DGS variants. In this paper, we identify a fundamental inefficiency in GPU-based 3DGS rendering: severe SM-level load imbalance caused by uneven workloads across screen tiles. This imbalance creates a long execution tail, leaving most SMs idle while a few process computationally intensive tiles, particularly in large-scale urban scenes. To address this, we present a software-defined GPU scheduling approach that mitigates the execution tail. Our approach uses a persistent kernel to schedule heavy tiles first and selectively chunks outlier tiles across multiple CTAs. A lightweight per-frame policy determines when and how much to chunk using tile statistics already available in the rendering pipeline. Our approach requires neither retraining nor specialized hardware and can be applied orthogonally to existing optimizations across diverse 3DGS variants. Averaged per scene, it accelerates alpha blending by up to 2.93x on a large-scale urban scene, and by up to 5.6x on its most imbalanced frames, while retaining the original alpha blending formulation.

## MOMASPLAT: Dynamic Novel-View Demonstration Augmentation for Mobile Manipulation

DevFeed: [MOMASPLAT: Dynamic Novel-View Demonstration Augmentation for Mobile Manipulation](<https://devfeed.tech/articles/momasplat-dynamic-novel-view-demonstration-augmentation-for-mobile-manipulation-120314.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.08372>)

Author: Sateesh Kumar, Arnav Balaji, Arpit Bahety, Skand Peri, Georgios Pavlakos, Roberto Mart\\'in-Mart\\'in

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Simulation](<https://devfeed.tech/topics/simulation.md>)

Tags: [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [foundation-models](<https://devfeed.tech/tags/foundation-models.md>), [kinematics](<https://devfeed.tech/tags/kinematics.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [mobile-manipulation](<https://devfeed.tech/tags/mobile-manipulation.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [robustness](<https://devfeed.tech/tags/robustness.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [synthesis](<https://devfeed.tech/tags/synthesis.md>)

### AI overview

MOMASPLAT augments mobile manipulation demonstrations for unseen robot base poses, including object motion after contact--a phase that prior novel-view methods described here leave unaugmented because they assume static scenes. It uses multiview data from standard collection to reconstruct the scene and manipulated object separately with 3D Gaussian Splatting and feed-forward 3D foundation models, then renders new viewpoints and retargets actions using robot kinematics. Across five real-world tasks, three scenes, and rigid and articulated objects, the authors report more than doubling average success over the strongest baseline and retaining non-trivial success at large base displacements where all baselines fail. Simulation comparisons were used to validate synthetic demonstration fidelity; the abstract gives no numerical rates or further implementation details.

### Source excerpt

arXiv:2611.08372v1 Announce Type: cross Abstract: Visuomotor policies trained through imitation learning are highly sensitive to changes in camera viewpoint. This is particularly problematic for mobile manipulation (MoMa), where robots rarely execute manipulation from the exact same base pose used during demonstration collection due to navigation uncertainty, localization drift, and environmental constraints. Recent approaches use novel-view synthesis (NVS) to augment demonstrations with unseen viewpoints, but they assume static scenes and therefore fail to re-render the manipulated object during the post-contact interaction phase, leaving the most critical parts of the demonstration unaugmented. We present MOMASPLAT, a novel-view demonstration augmentation framework for mobile manipulation that synthesizes full manipulation trajectories from unseen robot base poses, including dynamic object interactions. Our key insight is that standard MoMa data collection naturally provides the multiview observations needed to reconstruct both the static scene and the manipulated object in 3D. MOMASPLAT combines 3D Gaussian Splatting, feed-forward 3D foundation models, and kinematics-based action retargeting to separately reconstruct the scene and manipulated object, render them from novel viewpoints, and generate geometrically consistent synthetic demonstrations paired with transformed actions. We evaluate MOMASPLAT on five real-world mobile manipulation tasks involving both rigid and articulated objects across three scenes, and validate the fidelity of its synthesized demonstrations against ground truth in simulation. MOMASPLAT substantially improves robustness to unseen base poses, more than doubling the average success rate of the strongest baseline and maintaining non-trivial success at large base displacements where all baselines fail, while reducing the need for largescale real-world data collection.

## Explicit 3D Gaussian Models for Fast, Mesh-Free Compression of Structured and Unstructured Volumes

DevFeed: [Explicit 3D Gaussian Models for Fast, Mesh-Free Compression of Structured and Unstructured Volumes](<https://devfeed.tech/articles/explicit-3d-gaussian-models-for-fast-mesh-free-compression-of-structured-and-unstructured-volumes-120454.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2607.01164>)

Author: Landon Dyken, Sharmistha Chakrabarti, Nathan Debardeleben, Steve Petruzza, Qi Wu, Will Usher, Sidharth Kumar

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Compression](<https://devfeed.tech/topics/compression.md>), [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [aggregation](<https://devfeed.tech/tags/aggregation.md>), [compression](<https://devfeed.tech/tags/compression.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cuda](<https://devfeed.tech/tags/cuda.md>), [domain](<https://devfeed.tech/tags/domain.md>), [error](<https://devfeed.tech/tags/error.md>), [fast](<https://devfeed.tech/tags/fast.md>), [field](<https://devfeed.tech/tags/field.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [loss-functions](<https://devfeed.tech/tags/loss-functions.md>), [mesh](<https://devfeed.tech/tags/mesh.md>), [models](<https://devfeed.tech/tags/models.md>), [neural](<https://devfeed.tech/tags/neural.md>), [novel](<https://devfeed.tech/tags/novel.md>), [optimized](<https://devfeed.tech/tags/optimized.md>), [pipelines](<https://devfeed.tech/tags/pipelines.md>), [quality](<https://devfeed.tech/tags/quality.md>), [sampling](<https://devfeed.tech/tags/sampling.md>), [structured](<https://devfeed.tech/tags/structured.md>), [volumes](<https://devfeed.tech/tags/volumes.md>)

### AI overview

The paper presents 3D Gaussian primitives as an explicit representation for compressing structured and unstructured volume data. It reconstructs scalar values by weighted aggregation of intersecting Gaussians, with CUDA-accelerated sampling pipelines, coverage-oriented loss functions, and a densification strategy guided by sampling error. Because the representation encodes domain geometry, it removes the need to store partial meshes for sampling unstructured volumes. The authors report faster training and better reconstruction quality than existing implicit neural representations (INRs) on both volume types, and higher compression ratios than prior work for unstructured volumes. The supplied text is an abstract and gives no quantitative results or implementation details.

### Source excerpt

arXiv:2607.01164v2 Announce Type: replace Abstract: Recent work has shown that implicit neural representations (INRs) can be trained to effectively compress structured and unstructured volume data, allowing for direct data querying with a reduced memory footprint. However, as existing INRs for unstructured volumes do not encode geometry, they require partial mesh storage for later sampling, limiting achievable compression. At the same time, novel view synthesis methods have shown that explicit collections of 3D Gaussians can be used to accurately visualize volume data. In this work, we introduce an explicit model for volume data compression based on 3D Gaussian primitives. We reinterpret collections of 3D Gaussians as an explicit representation of a volume's scalar field and use a sampling strategy that reconstructs scalar values at spatial locations through weighted aggregation of intersecting Gaussians. We develop optimized CUDA-accelerated pipelines for structured and unstructured model sampling, loss functions that encourage accurate domain coverage of our models, and a novel sampling-error based densification strategy. Our explicit formulation naturally encodes domain geometry, eliminating the need for mesh storage in unstructured volumes and introducing significantly higher compression opportunities. Compared to existing INRs, we demonstrate that our explicit model achieves significant training speed and reconstruction quality improvements on both structured and unstructured volumes, while enabling compression ratios much higher than previous work for unstructured volumes.

## EDGAR: Efficient Diffusion-Driven Gaussian Avatar Refinement via Synthetic Failure

DevFeed: [EDGAR: Efficient Diffusion-Driven Gaussian Avatar Refinement via Synthetic Failure](<https://devfeed.tech/articles/edgar-efficient-diffusion-driven-gaussian-avatar-refinement-via-synthetic-failure-120678.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.07909>)

Author: Kwan Yun, Minhyeok Bang, Colette Suhjung Koo, Junyong Noh

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [diffusion-transformers](<https://devfeed.tech/topics/diffusion-transformers.md>), [Image](<https://devfeed.tech/topics/image.md>), [3D](<https://devfeed.tech/topics/3d.md>)

Tags: [artifacts](<https://devfeed.tech/tags/artifacts.md>), [avatar](<https://devfeed.tech/tags/avatar.md>), [consistency](<https://devfeed.tech/tags/consistency.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-gr](<https://devfeed.tech/tags/cs-gr.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-mm](<https://devfeed.tech/tags/cs-mm.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [failure](<https://devfeed.tech/tags/failure.md>), [fusion](<https://devfeed.tech/tags/fusion.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [gaussian-splatting](<https://devfeed.tech/tags/gaussian-splatting.md>), [identity](<https://devfeed.tech/tags/identity.md>), [mixture-of-experts](<https://devfeed.tech/tags/mixture-of-experts.md>), [mixture-of-experts-moe](<https://devfeed.tech/tags/mixture-of-experts-moe.md>), [moe](<https://devfeed.tech/tags/moe.md>), [robustness](<https://devfeed.tech/tags/robustness.md>), [sapiens](<https://devfeed.tech/tags/sapiens.md>), [synthetic](<https://devfeed.tech/tags/synthetic.md>), [temporal](<https://devfeed.tech/tags/temporal.md>)

### AI overview

EDGAR refines appearance artifacts in Gaussian head avatars while aiming to preserve identity. It trains on EDGAR-SET, a mix of captured and synthetic triplets pairing degraded renderings with ground-truth and reference images. A fusion block injects identity features from the self-supervised Sapiens model into a one-step diffusion backbone; a Mixture-of-Experts handles varied artifact types and severity. Refined multi-view renders serve as pseudo-targets for optimizing one underlying avatar, so EDGAR need not run on every final frame. Across three reconstruction methods and three settings, the abstract reports better perceptual and reconstruction measures than raw outputs and an SDEdit-style baseline, plus improved cross-view calibrated consistency and robustness in large-yaw and open-mouth tests. The authors describe the results as a useful operating regime for recoverable appearance artifacts.

### Source excerpt

arXiv:2611.07909v1 Announce Type: cross Abstract: Recent advances in Gaussian splatting have enabled efficient and photorealistic head avatar reconstruction, yet monocular video-based or image-based methods still suffer from under-constrained geometry and texture, leading to artifacts such as blurring, stretching, and temporal instability. We present EDGAR (Efficient Diffusion-Driven Gaussian Avatar Refinement), a conditional diffusion framework designed to enhance the quality of Gaussian avatars by correcting artifacts while preserving identity fidelity. To train EDGAR, we construct EDGAR-SET, a hybrid triplet dataset composed of real-captured and synthetic-rendered samples, which simulates realistic Gaussian avatar failure cases by pairing degraded renderings with ground-truth and reference images. EDGAR refines degraded frames through a fusion block that injects identity features extracted by the self-supervised Sapiens model into a one-step diffusion backbone, enabling effective artifact removal without sacrificing efficiency. Our EDGAR model incorporates a Mixture-of-Experts (MoE) architecture to address diverse degradation types and severity levels. The refined multi-view renderings are subsequently used as pseudo-targets to optimize a single underlying Gaussian avatar, so that the final avatar can be rendered without applying EDGAR to every output frame. Experiments on three different Gaussian avatar reconstruction methods, each evaluated under three different settings, show improvements in both perceptual and reconstruction measures over the raw outputs and an SDEdit-style refinement baseline. In addition, cross-view evaluation demonstrates improved calibrated consistency, while tests under large yaw and open-mouth expressions verify robustness under extreme settings. These results collectively indicate a useful operating regime for recoverable appearance artifacts. The code is available at https://kwanyun.github.io/edgar/

## AEGIS: Preserving Privacy of 3D Facial Avatars with Adversarial Perturbations

DevFeed: [AEGIS: Preserving Privacy of 3D Facial Avatars with Adversarial Perturbations](<https://devfeed.tech/articles/aegis-preserving-privacy-of-3d-facial-avatars-with-adversarial-perturbations-118184.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2511.17747>)

Author: Dawid Wolkiewicz, Anastasiya Pechko, Przemys{\\l}aw Spurek, Piotr Syga

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Facial recognition](<https://devfeed.tech/topics/facial-recognition.md>), [Security & Privacy](<https://devfeed.tech/topics/security-privacy.md>), [3D](<https://devfeed.tech/topics/3d.md>), [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [Adversarial attacks](<https://devfeed.tech/topics/adversarial-attacks.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [adversarial](<https://devfeed.tech/tags/adversarial.md>), [animation](<https://devfeed.tech/tags/animation.md>), [avatar](<https://devfeed.tech/tags/avatar.md>), [biometric-authentication](<https://devfeed.tech/tags/biometric-authentication.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [face-recognition](<https://devfeed.tech/tags/face-recognition.md>), [identity-protection](<https://devfeed.tech/tags/identity-protection.md>), [identity-theft](<https://devfeed.tech/tags/identity-theft.md>), [masking](<https://devfeed.tech/tags/masking.md>), [metric](<https://devfeed.tech/tags/metric.md>), [networks](<https://devfeed.tech/tags/networks.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>)

### AI overview

AEGIS protects the identity of 3D Gaussian facial avatars by perturbing the primitives' base color coefficients, guided by an ensemble of pretrained face verification networks. Because the protection is embedded in the 3D representation, it applies across viewpoints and animations without retraining or changing avatar geometry. In evaluations against black-box recognizers, the method reduced face retrieval and verification accuracy to near zero while maintaining perceptual fidelity and preserving age, gender, and emotion. The authors also introduce a viewpoint-robust leakage metric to measure residual identity exposure across rendered views.

### Source excerpt

arXiv:2511.17747v2 Announce Type: replace-cross Abstract: The growing adoption of photorealistic 3D facial avatars, particularly those utilizing efficient 3D Gaussian Splatting representations, introduces new risks of online identity theft, especially in systems that rely on biometric authentication. While effective adversarial masking methods have been developed for 2D images, a significant gap remains in achieving robust, viewpoint-consistent identity protection for dynamic 3D avatars. To address this, we present AEGIS, the first privacy-preserving identity masking framework for 3D Gaussian avatars, concealing identity from automated face recognition while preserving the avatar's perceptual realism and animation usability. AEGIS applies adversarial perturbations to the base (DC) color coefficients of the Gaussian primitives, guided by an ensemble of pre-trained face verification networks, embedding the protection directly in the 3D representation so that it applies across viewpoints and animations, without retraining or modifying the avatar's geometry. Evaluated against black-box recognizers, AEGIS reduces face retrieval and verification accuracy to near-zero while maintaining high perceptual fidelity. It also preserves key facial attributes such as age, gender, and emotion, demonstrating strong privacy protection with minimal visual distortion. We further introduce a viewpoint-robust leakage metric that quantifies residual identity leakage across an avatar's rendered viewpoints, which single-view evaluation cannot capture.

## Quo Vadis, Driving Scene Reconstruction? Occlusion-Aware Gaussian Densification

DevFeed: [Quo Vadis, Driving Scene Reconstruction? Occlusion-Aware Gaussian Densification](<https://devfeed.tech/articles/quo-vadis-driving-scene-reconstruction-occlusion-aware-gaussian-densification-118441.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.06273>)

Author: Dung Do Tien, Kien Nguyen Do Trung, Kiet Dang Vu, Trung Thai Tran, Duc Dung Nguyen

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [3D rendering](<https://devfeed.tech/topics/3d-rendering.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [completion](<https://devfeed.tech/tags/completion.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [driving](<https://devfeed.tech/tags/driving.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [novel](<https://devfeed.tech/tags/novel.md>)

### AI overview

OccSplat addresses incomplete driving-scene reconstructions by adding predicted 3D Gaussian primitives where input views leave occluded regions unsupported, instead of filling missing pixels in rendered images. Its Gaussian Occlusion Mapping identifies those regions, and Gaussian Occlusion Completion generates primitives that are integrated into the original representation. The feed-forward method requires no per-scene optimization. On nuScenes, it improves on the compared methods, with particularly large gains in occluded regions; the supplied excerpt gives no numerical results.

### Source excerpt

arXiv:2611.06273v1 Announce Type: cross Abstract: While recent feed-forward 3D Gaussian Splatting methods enable efficient driving-scene reconstruction from surround-view images, their reliance on a single observation fundamentally hinders novel-view synthesis. Regions occluded in the input can become visible from novel viewpoints, leaving the reconstructed Gaussians under-constrained and incomplete. Moving beyond image-space inpainting for rendered views, we introduce OccSplat, a feed-forward framework whose core contribution is Occlusion-Aware Gaussian Densification, explicitly targeting missing scene geometry rather than missing pixels. By predicting new Gaussian primitives for unsupported regions, OccSplat completes the scene directly in the 3D representation. Specifically, OccSplat first identifies occluded regions that lack Gaussian support through Gaussian Occlusion Mapping. Gaussian Occlusion Completion then generates new primitives for these regions. The predicted primitives are integrated into the original representation to improve novel-view synthesis while retaining feed-forward inference without per-scene optimization. Experiments on the nuScenes benchmark demonstrate that OccSplat improves on the compared methods and yields particularly large gains in occluded regions, leading to more accurate and visually coherent driving-scene reconstruction.

## SHield: Watermarking 4D Gaussian Splatting via Spherical Harmonic Coefficient Perturbation

DevFeed: [SHield: Watermarking 4D Gaussian Splatting via Spherical Harmonic Coefficient Perturbation](<https://devfeed.tech/articles/shield-watermarking-4d-gaussian-splatting-via-spherical-harmonic-coefficient-perturbation-114045.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.01162>)

Author: Kunal Bhosikar, Prajas Wadekar, Venkata Sai Pranav Bachina, Ankit Gangwal, Charu Sharma, Nicole Meng, Yingjie Lao

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [Image](<https://devfeed.tech/topics/image.md>), [3D rendering](<https://devfeed.tech/topics/3d-rendering.md>), [quantization](<https://devfeed.tech/topics/quantization.md>)

Tags: [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-gr](<https://devfeed.tech/tags/cs-gr.md>), [cs-mm](<https://devfeed.tech/tags/cs-mm.md>), [gaussian-splatting](<https://devfeed.tech/tags/gaussian-splatting.md>), [image](<https://devfeed.tech/tags/image.md>), [recovery](<https://devfeed.tech/tags/recovery.md>), [rgb](<https://devfeed.tech/tags/rgb.md>), [robustness](<https://devfeed.tech/tags/robustness.md>), [shield](<https://devfeed.tech/tags/shield.md>), [verification](<https://devfeed.tech/tags/verification.md>), [watermarking](<https://devfeed.tech/tags/watermarking.md>)

### AI overview

SHield embeds full-resolution RGB images in 4D Gaussian Splatting files by tiling a zero-mean payload across spherical-harmonic coefficients of a key-selected Gaussian subset, then applying brief anchored optimization to preserve visual fidelity. Ownership verification re-associates the key through canonical Gaussian centres and recovers the payload with deterministic arithmetic, avoiding a neural decoder and rendered-video extraction. Across 30 D-NeRF, HyperNeRF, and DyNeRF scenes, the excerpt reports Pearson recovery of at least 0.991 on 150 models with mean rendering penalty no greater than 0.13 dB; held-out tests had 400 true detections and zero false positives in 1,600 wrong-subset trials. Robustness is bounded: image payloads survive INT8 quantization and 30% opacity pruning, while after aggressive SH-degree-1 truncation only bit payloads are reported decodable.

### Source excerpt

arXiv:2611.01162v1 Announce Type: cross Abstract: 4D Gaussian Splatting (4DGS) models are computationally expensive to train but distributed as easily copied parameter files. Existing watermarks protect these assets by fine-tuning the scene against a neural decoder to extract bit strings from rendered video, a heavy process vulnerable to the temporal dynamics of dynamic scenes. However, if the threat model targets the redistributed file itself, verification can be vastly simplified. Because the canonical Spherical-Harmonic (SH) coefficients of 4DGS are time- and view-invariant, parameter-space perturbations can be recovered via deterministic arithmetic. We propose **SHield**, the first non-blind, decoder-free watermarking approach that natively embeds full-resolution RGB images into 4DGS files. SHield tiles a zero-mean payload across the SH coefficients of a key-selected Gaussian subset, applying a brief anchored optimization to preserve visual fidelity. To verify ownership, it geometrically re-associates the key using canonical Gaussian centres to survive file tampering and recovers the payload in closed form. Across 30 scenes (D-NeRF, HyperNeRF, DyNeRF), SHield guarantees near-perfect payload recovery (Pearson $r \geq 0.991$) on all 150 tested models with a mean rendering penalty of $\leq 0.13$~dB. Held-out evaluation yields perfect detection on 400 true trials and zero false positives across $1{,}600$ wrong-subset null trials. We also strictly bound the method's robustness: SHield survives INT8 quantization and $30\%$ opacity pruning for the image payload, and on aggressive SH-degree-1 truncation bit payloads are still decodable.

## Splaxel: Efficient Distributed Training of 3D Gaussian Splatting for Large-scale Scene Reconstruction via Pixel-level Communication

DevFeed: [Splaxel: Efficient Distributed Training of 3D Gaussian Splatting for Large-scale Scene Reconstruction via Pixel-level Communication](<https://devfeed.tech/articles/splaxel-efficient-distributed-training-of-3d-gaussian-splatting-for-large-scale-scene-reconstruction-via-pixel-level-communication-113761.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.18588>)

Author: Wenqi Jia, Zhewen Hu, Ying Huang, Jian Xu, Siyi Wu, Yangming Zhang, Yu Gong, Stavros Kalafatis, Yuke Wang, Wei Niu, Chengming Zhang, Ang Li, Sheng Di, Yuede Ji, Bo Fang, Miao Yin

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [distributed-training](<https://devfeed.tech/topics/distributed-training.md>), [3D rendering](<https://devfeed.tech/topics/3d-rendering.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [3dgs](<https://devfeed.tech/tags/3dgs.md>), [communication](<https://devfeed.tech/tags/communication.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [distributed-training](<https://devfeed.tech/tags/distributed-training.md>), [gaussian-splatting](<https://devfeed.tech/tags/gaussian-splatting.md>), [large](<https://devfeed.tech/tags/large.md>), [pixel](<https://devfeed.tech/tags/pixel.md>), [scale](<https://devfeed.tech/tags/scale.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

Splaxel scales 3D Gaussian Splatting training by having each GPU render its local Gaussian subset and exchange partial pixel values for global composition, avoiding Gaussian-level synchronization. The abstract says this maintains mathematical consistency while keeping communication cost stable as scene size grows. Geometric and transmittance visibility prediction reduce redundant pixel communication, while conflict-free camera-view consolidation improves GPU utilization. On large-scale datasets with up to 120 million Gaussians, Splaxel reports up to 7.6x speedup over the state-of-the-art distributed 3DGS framework while preserving high reconstruction quality.

### Source excerpt

arXiv:2606.18588v2 Announce Type: replace Abstract: 3D Gaussian Splatting (3DGS) enables high-fidelity and real-time 3D scene reconstruction, but scaling training to large-scale scenes requires optimizing hundreds of millions of Gaussians across multiple GPUs. Existing distributed approaches either partition scenes into isolated regions, causing global inconsistency, or rely on global Gaussian-level exchanges, which lead to substantial growth in inter-GPU communication and quickly dominate iteration time. We propose Splaxel, a communication-efficient distributed 3DGS training framework based on pixel-level local rendering and global composition. Instead of synchronizing Gaussians, each GPU renders its local subset and exchanges only partial pixel values, maintaining mathematical consistency while keeping communication cost stable as the scene size increases. Splaxel further reduces pixel-level redundancy through geometric and transmittance visibility prediction and improves GPU utilization via conflict-free camera-view consolidation. Evaluated on large-scale datasets with up to 120M Gaussians, Splaxel achieves up to 7.6$\times$ speedup over the state-of-the-art distributed 3DGS framework while preserving high reconstruction quality.

## Xiaomi Auto World Model 2.0: Causal World Models for Real-World Driving and Embodied Intelligence

DevFeed: [Xiaomi Auto World Model 2.0: Causal World Models for Real-World Driving and Embodied Intelligence](<https://devfeed.tech/articles/xiaomi-auto-world-model-2-0-causal-world-models-for-real-world-driving-and-embodied-intelligence-114587.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.01677>)

Author: Zhenxin Zhu, Zhijian Huang, Cheng Chi, Yuru Wang, Jianhua Wu, Jiayi Guan, Yonggang Li, Mingfei Tu, JiaKui Hu, Tao Xu, Xianqi Wang, Hao Li, Yuechuan Pu, Qizeng Jia, Ruiling Yang, Hongxiao Yu, Xuhan She

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [model architecture](<https://devfeed.tech/topics/model-architecture.md>), [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [attention](<https://devfeed.tech/tags/attention.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [data](<https://devfeed.tech/tags/data.md>), [distillation](<https://devfeed.tech/tags/distillation.md>), [generate](<https://devfeed.tech/tags/generate.md>), [gla](<https://devfeed.tech/tags/gla.md>), [horizon](<https://devfeed.tech/tags/horizon.md>), [inference](<https://devfeed.tech/tags/inference.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [kv-cache](<https://devfeed.tech/tags/kv-cache.md>), [physical-ai](<https://devfeed.tech/tags/physical-ai.md>), [policy](<https://devfeed.tech/tags/policy.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [temporal](<https://devfeed.tech/tags/temporal.md>), [training](<https://devfeed.tech/tags/training.md>), [world-model](<https://devfeed.tech/tags/world-model.md>), [xiaomi](<https://devfeed.tech/tags/xiaomi.md>)

### AI overview

Xiaomi Auto World Model 2.0 is presented as a streaming causal world model for generating view-consistent physical scene evolution. View-Temporal Attention and multi-view RoPE capture spatial relationships across cameras and time; WorldRec 2.0 reconstructs static and dynamic 3D Gaussians in one pass to provide scene context for preserving geometry and appearance over long horizons. VAE latent concatenation fuses conditioning inputs, while Gated Linear Attention in temporal layers reduces computation and KV-cache memory during multi-view rollouts. Its training combines physical-interaction data with large-scale videos and varied visual modalities, then uses diverse view configurations and distillation. The authors describe applications in closed-loop simulation and training, and World Action Models for state prediction, policy learning, and action generation. These are claims from the supplied abstract; it provides no quantitative results.

### Source excerpt

arXiv:2611.01677v1 Announce Type: cross Abstract: We present Xiaomi Auto World Model 2.0, a unified streaming causal world model for Physical AI. It models and generates causal physical evolution in the real world, and unifies heterogeneous conditioning inputs for view-consistent generation. The architecture integrates View-Temporal Attention (VTA) to capture cross-view spatial dependencies and temporal dynamics, together with Multi-view Rotary Position Encoding (RoPE) to encode positional relationships across views and time. WorldRec 2.0 reconstructs static and dynamic 3D Gaussians from multi-view, multi-temporal observations in a single forward pass and uses their renderings as explicit scene context to preserve geometry and appearance during long-horizon generation. Conditioning inputs are fused through VAE latent concatenation in a shared representation space. We incorporate Gated Linear Attention (GLA) into the temporal layers of VTA for efficient streaming inference, reducing computation and KV-cache memory during multi-view rollouts. A physical-interaction data pipeline covers diverse interactions and long-tailed events. From Pre-train onward, these data are combined with large-scale videos and heterogeneous visual modalities to learn visual, geometric, and dynamic priors. During Mid-train, training across diverse view configurations further improves the model's ability to capture how actions and object states shape subsequent scene evolution. After distillation-based Post-train, we apply the model to closed-loop simulation, closed-loop training, and World Action Model (WAM) applications. These applications generate physically consistent future states and support world-state prediction, policy learning, and action generation for Physical AI.

## RxGS: Receiver-Generalizable 3D Gaussian Splatting for Radio-Frequency Data Synthesis

DevFeed: [RxGS: Receiver-Generalizable 3D Gaussian Splatting for Radio-Frequency Data Synthesis](<https://devfeed.tech/articles/rxgs-receiver-generalizable-3d-gaussian-splatting-for-radio-frequency-data-synthesis-113846.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2605.24290>)

Author: Kang Yang, Mani Srivastava

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [3dgs](<https://devfeed.tech/tags/3dgs.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [data](<https://devfeed.tech/tags/data.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [gaussian-splatting](<https://devfeed.tech/tags/gaussian-splatting.md>), [inference](<https://devfeed.tech/tags/inference.md>), [learning](<https://devfeed.tech/tags/learning.md>), [model](<https://devfeed.tech/tags/model.md>), [radio](<https://devfeed.tech/tags/radio.md>), [receiver](<https://devfeed.tech/tags/receiver.md>), [rendering](<https://devfeed.tech/tags/rendering.md>), [rf](<https://devfeed.tech/tags/rf.md>), [signal](<https://devfeed.tech/tags/signal.md>), [storage](<https://devfeed.tech/tags/storage.md>), [synthesis](<https://devfeed.tech/tags/synthesis.md>), [training](<https://devfeed.tech/tags/training.md>), [wireless](<https://devfeed.tech/tags/wireless.md>)

### AI overview

RxGS uses one shared 3D Gaussian Splatting model to synthesize RF signals at receiver positions excluded from training, within a surveyed scene. It first learns receiver-independent Gaussian geometry, freezes that geometry, then learns directional radiance conditioned on receiver position. A global branch models receiver effects shared across the scene, while a local branch captures per-scatterer differences tied to receiver geometry and occlusion. Shared geometry enables batched rendering across receivers. The abstract reports lower training, inference, and storage costs than approaches requiring a separate model for each receiver; it gives no numerical results or details about the datasets.

### Source excerpt

arXiv:2605.24290v2 Announce Type: replace Abstract: Radio-frequency (RF) data synthesis predicts the received signal given transmitter and receiver positions, and is essential for wireless applications. Recent 3D Gaussian Splatting (3DGS)-based methods achieve efficient synthesis at any transmitter location within a surveyed scene, but require a separate model for each receiver. We therefore present RxGS, a shared 3DGS model that first learns receiver-independent Gaussian geometry and then freezes it while learning directional radiance conditioned on receiver position. Its global conditioning branch captures receiver-dependent effects shared across the scene, while a local branch models per-scatterer variations from the receiver's geometry and occlusion. The shared geometry also supports batched rendering through a multi-receiver rasterizer. Across RF datasets, RxGS predicts at receiver positions excluded from training within the surveyed scene while significantly reducing training, inference, and storage costs. Project page: https://kangyangg.com/projects/rxgs/.

## GSExplorer: Interactive World Generation with a Persistent 3D Gaussian World State

DevFeed: [GSExplorer: Interactive World Generation with a Persistent 3D Gaussian World State](<https://devfeed.tech/articles/gsexplorer-interactive-world-generation-with-a-persistent-3d-gaussian-world-state-115176.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.01543>)

Author: Sicheng Yu, Liu Liu, Xinjie Wang, Xinjiang Wang, Taojun Ding, Wenhao Hu, Beizhen Zhao, Puxin Yan, Zhizhong Su, Hao Wang

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [Computer vision](<https://devfeed.tech/topics/computer-vision.md>), [3D](<https://devfeed.tech/topics/3d.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3dgs](<https://devfeed.tech/tags/3dgs.md>), [camera](<https://devfeed.tech/tags/camera.md>), [camera-control](<https://devfeed.tech/tags/camera-control.md>), [consistency](<https://devfeed.tech/tags/consistency.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [exploration](<https://devfeed.tech/tags/exploration.md>), [framework](<https://devfeed.tech/tags/framework.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [generation](<https://devfeed.tech/tags/generation.md>), [generator](<https://devfeed.tech/tags/generator.md>), [guidance](<https://devfeed.tech/tags/guidance.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [rendering](<https://devfeed.tech/tags/rendering.md>), [slam](<https://devfeed.tech/tags/slam.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [stable](<https://devfeed.tech/tags/stable.md>), [state](<https://devfeed.tech/tags/state.md>), [target](<https://devfeed.tech/tags/target.md>), [world](<https://devfeed.tech/tags/world.md>)

### AI overview

GSExplorer maintains a persistent 3D Gaussian scene to support consistent interactive world generation over repeated exploration steps. At each step, the scene supplies multimodal conditions to a video generator, while a separate Relative Ray Encoding branch adds camera geometry. Generated video is fed into a SLAM reconstruction backend; inconsistency filtering limits conflicting updates, and cross-chunk bundle adjustment improves spatial alignment. The authors create 293K training pairs that reproduce 3DGS degradation during scene extrapolation. On the reported evaluations, GSExplorer performs best overall among compared methods on single-step and three-step WorldScore, with a single-step Camera Control score of 99.07, and shows strong revisit consistency in 481-frame leave-and-return rollouts.

### Source excerpt

arXiv:2611.01543v1 Announce Type: cross Abstract: Interactive world generation requires both accurate local control and stable scene consistency over long rollouts. Existing methods improve local generation through camera or geometry conditioning and extend historical context with memory mechanisms. However, long-horizon exploration remains challenging as target-view guidance becomes incomplete and inconsistencies accumulate across repeated scene updates. We present \textbf{GSExplorer}, a generation--reconstruction framework built around a persistent 3D Gaussian world state. At each step, the current 3DGS scene provides multimodal rendering conditions for a condition-driven video generator, while camera geometry is injected through a separate Relative Ray Encoding branch. Generated videos are then fed back to a SLAM-based reconstruction backend, where inconsistency filtering prevents conflicting content from corrupting the scene and cross-chunk bundle adjustment improves spatial alignment across generation steps. We further design a data-construction pipeline that reproduces realistic 3DGS degradation during scene extrapolation, yielding 293K training pairs of degraded multimodal renderings and real target videos. GSExplorer achieves the best overall performance among the compared methods on both single-step and three-step WorldScore evaluation, with strong camera control and consistency across both settings, including a Camera Control score of 99.07 in single-step extrapolation. On 481-frame leave-and-return rollouts, it also shows strong revisit consistency, demonstrating the value of maintaining a reliable 3D world state for continued exploration.

## VLSplat: Vision-Language Guided Object-Centric 3D Gaussian Splatting via Scene Graph

DevFeed: [VLSplat: Vision-Language Guided Object-Centric 3D Gaussian Splatting via Scene Graph](<https://devfeed.tech/articles/vlsplat-vision-language-guided-object-centric-3d-gaussian-splatting-via-scene-graph-115239.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.01939>)

Author: Yuntae Jeon, Younho Jeon, Sujin Jin, Sungho Jo, Seunghee Park

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gaussian splatting](<https://devfeed.tech/topics/gaussian-splatting.md>), [vlm](<https://devfeed.tech/topics/vlm.md>), [Machine Perception](<https://devfeed.tech/topics/machine-perception.md>), [3D rendering](<https://devfeed.tech/topics/3d-rendering.md>)

Tags: [3d-gaussian-splatting](<https://devfeed.tech/tags/3d-gaussian-splatting.md>), [3d-scene-understanding](<https://devfeed.tech/tags/3d-scene-understanding.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [graph](<https://devfeed.tech/tags/graph.md>), [vision-language-models](<https://devfeed.tech/tags/vision-language-models.md>)

### AI overview

VLSplat adds object-level semantic priors from vision-language models to mask-based semantic lifting for 3D Gaussian Splatting, then refines Gaussian representations using an object-group scene graph. Its intra-group module suppresses semantic noise within object groups; its inter-group module expands object support by adding structurally relevant candidates from neighboring groups. The abstract reports improved semantic consistency and structural completeness of object-level categories in indoor scene experiments. It does not provide quantitative results or further detail on the datasets or evaluation.

### Source excerpt

arXiv:2611.01939v1 Announce Type: cross Abstract: Learning semantic representations for 3D Gaussian Splatting has recently emerged as a promising direction for 3D scene understanding. However, existing 2D-to-3D semantic lifting approaches rely primarily on 2D segmentation supervision and lack object-level priors, causing the learned representation to be driven by local mask evidence rather than holistic object structure. In this paper, we present VLSplat, a vision-language guided approach for object-centric semantic lifting in 3D Gaussian Splatting. Our method augments mask-based lifting with object-level semantic priors derived from vision-language models and refines Gaussian representations using an object-group scene graph. The refinement is performed through two language-guided modules: intra-group refinement suppresses semantic noise within object groups, while inter-group acquisition expands object support regions by incorporating structurally relevant candidates from neighboring groups. Extensive experiments on indoor scene datasets demonstrate that VLSplat improves the semantic consistency and structural completeness of object-level categories. Our code and models are available at https://vlsplat.github.io/.

[Next page](<https://devfeed.tech/topics/gaussian-splatting.md?cursor=WyIyMDI2LTExLTAzVDA1OjAwOjAwKzAwOjAwIiwgIjEzYjUwMzY0LTZiYWUtNGRkYi05YjU1LWJmMTI5NTc2ZDE5OCJd>)