Research Radar

arXiv cs.GR details Text2Sim, a simulation-focused agentic pipeline that converts a text-only request into an executable, editable dynamic case. Built on Genesis, it combines a Planner, specialized Writers, asset-generation tools, and an independent Critic to handle the full chain from scene setup to execution-time repair. …

• Text2Sim generates editable physics simulations from text prompts
• Xiaoyu Xiong, Tsun-Hsuan Wang, Yi-Ling Qiao, Tao Du, and Minchen Li

arXiv cs.GR details a system for 3D obstacle-course games that gives designers rapid feedback on how a layout can be solved. The tool answers constrained queries such as avoiding regions, passing through waypoints, or limiting jumps, so teams can probe intended routes and edge cases …

• A system for rapid solvability feedback in 3D obstacle-course game design
• Zander Majercik and eight coauthors; published on arXiv cs.GR

arXiv cs.GR details a new reconstruction framework for isogeometric analysis that temporarily splits a NURBS model into independent local sections, so patches can be edited with local knot insertion, degree elevation, and basis changes without losing exact geometry. The method is aimed at a long-standing …

• A hybrid reconstruction framework for admissible spline spaces in isogeometric analysis
• C.G. Provatidis and I. Dimitriou, published on arXiv cs.GR

arXiv cs.GR details CoDimRecon, an agentic reconstruction framework aimed at producing simulation-ready 3D scenes from multi-view RGB observations. The pitch is broader than standard scene capture: it handles rigid objects, articulated parts, and deformables in one pipeline, with editable outputs that are meant to be …

• CoDimRecon reconstructs sim-ready 3D scenes from multi-view RGB, including rigid, articulated, and deformable objects.
• Shuzhao Xie, Lelin Wang, Guying Lin, Zhi Wang, and Minchen Li; published on arXiv cs.GR.

arXiv cs.GR details DRHeC, a new differentiable rendering framework for hand-eye calibration that swaps mask-only optimization for RGB-based gradients. The pitch is straightforward: give the optimizer richer geometric and appearance cues so calibration can converge more reliably and with less loss of detail. The paper …

• DRHeC is a differentiable rendering framework for hand-eye calibration using RGB-based gradients.
• Xiaotian Zhang and six coauthors published the work on arXiv cs.GR.

arXiv cs.GR details NRF-GS, a new Gaussian splatting approach from Pratik Singh Bisht and Andreas Kolb that rethinks how appearance is modeled in 3DGS. Instead of relying on low-order spherical harmonics per splat, the method uses a compact per-Gaussian appearance code plus a shared scene-level …

• NRF-GS replaces per-splat spherical harmonics with a shared neural residual field in 3D Gaussian Splatting.
• Pratik Singh Bisht and Andreas Kolb, published on arXiv cs.GR.

arXiv cs.GR details Volcanite, an open-source framework for dense-segmentation volume visualization that runs on commodity hardware. The pitch is straightforward: instead of forcing teams to preprocess teravoxel label fields into meshes or slice-based views, Volcanite renders the data directly with a Vulkan backend and segmentation-aware …

• Volcanite is an open-source framework for dense-segmentation volume visualization on commodity hardware.
• Authors: Max Piochowiak, Reiner Dolp, Julian Herold, Eric Behle, Carsten Dachsbacher, Alexander Schug.

arXiv cs.GR details a length-varying neural motion stitching approach built around a cluster transition graph, aimed at generating smoother character animation between two distinct motion clips. The paper is from Haemin Kim, Junghyun Nam, Seokhyeon Hong, Vanessa Tan, and Junyong Noh, and it focuses on …

• A length-varying neural motion stitching method using a cluster transition graph
• Haemin Kim, Junghyun Nam, Seokhyeon Hong, Vanessa Tan, and Junyong Noh

arXiv cs.GR details MSTypography, a new multi-character semantic typography method aimed at balancing word legibility with object recognizability. The work is pitched as a step beyond single-character text effects, where multi-letter words often lose structure once designers push them toward a pictorial silhouette. The pipeline …

• MSTypography proposes multi-character semantic typography that preserves legibility while forming object-like word shapes.
• Authors listed on arXiv cs.GR: Xinye Yang, Xinding Zhu, Kai Fang, Xinyi Ren, Mengjian Li, Bin Cao, Jiazhou Chen.

arXiv cs.GR details a new take on domain translation for graphics and vision: adapt self-supervised features instead of trying to rewrite the whole image pipeline. The Representation Feature Adapter (RFA) is a 2.9M-parameter network that shifts the “residue” of weather, lighting, and rendering style in …

• A Representation Feature Adapter (RFA) moves domain residue in self-supervised feature maps instead of adapting generators
• Thomas Deixelberger and Markus Steinberger

arXiv cs.GR details a new method for generating high-quality 3D materials entirely in texture space. The paper finetunes a video diffusion transformer for text-guided material generation, multi-view material generation, and material upscaling, while using the known projection from image space to texture space to keep …

• Texture Space Material Diffusion generates and reconstructs materials in texture space using a finetuned video diffusion transformer.
• Authors: Jacob Munkberg, Peter Kocsis, and Jon Hasselgren; source: arXiv cs.GR.

arXiv cs.GR details PRISM, a real-to-sim-to-real framework aimed at scaling humanoid loco-manipulation from a tiny set of real interaction videos. The core idea is to amplify a handful of clean clips into hundreds of diverse counterfactual human-object interactions, then convert those videos into physically plausible …

• PRISM uses counterfactual video generation to scale humanoid loco-manipulation training data
• Zihan Wang and collaborators; arXiv cs.GR