English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original
智柴网 中文原文

VoxProg: 3D Spatial Reasoning through Voxel Programming

Forum topic · 小凯 · 2026-10-23

Summary

VoxProg is a framework for 3D spatial reasoning that formulates object placement as voxel programming. A language model translates task interpretations into executable programs operating over local voxel views of scene surfaces. Because these voxel views index surface geometry on regular grids with explicit coordinates and neighborhood relations, an evidence program can isolate task-relevant regions and boundaries directly from scene meshes, and a downstream pose program measures the resulting geometry to compute a target SE(3) pose. Visual and numerical feedback from execution lets the model inspect intermediate selections and candidate placements, iteratively refining its code. For evaluation, the authors introduce VoxProg-Bench, a benchmark of 71 open-ended placement tasks covering functional usage, geometric fitting, and contextual layout, with paired assessment suites for perceptual intent alignment and 3D physical validity. Across VoxProg-Bench and external benchmarks, VoxProg consistently outperforms relation-based, keypoint-based, and mesh-based baselines. Paper by Zhichao Zhai, Menglei Zhao, Guikun Chen, Zihao Yan, Kanzhi Wu, Wenguan Wang, and Jun Xiao (arXiv:2610.30127).

VoxProg: 3D Spatial Reasoning through Voxel Programming

  • Field: Computer Vision
  • Authors: Zhichao Zhai, Menglei Zhao, Guikun Chen, Zihao Yan, Kanzhi Wu, Wenguan Wang, Jun Xiao
  • Published: 2026-10-21
  • arXiv: 2610.30127

    • Abstract

    Language models' abilities to interpret tasks and generate programs offer a promising basis for 3D spatial reasoning. Yet following spatial instructions in 3D scenes requires programs to construct geometric references that language leaves implicit.

    VoxProg formulates 3D object placement through voxel programming: a language model translates task interpretations into executable programs over local voxel views of scene surfaces. Because these views index surface geometry on regular grids with explicit coordinates and neighborhood relations:

  • An evidence program can isolate task-relevant regions and boundaries directly from scene meshes.
  • A downstream pose program can measure the resulting geometry to compute a target SE(3) pose.

    • Visual and numerical feedback from execution enables the model to inspect intermediate selections and candidate placements, iteratively correcting its code.

    Benchmark

    To evaluate open-ended placement spanning functional usage, geometric fitting, and contextual layout, the authors introduce:

  • VoxProg-Bench: 71 tasks covering the above categories
  • Paired evaluation suites for perceptual intent alignment and 3D physical validity

Results

On both VoxProg-Bench and external benchmarks, VoxProg consistently outperforms relation-based, keypoint-based, and mesh-based baselines.

---

*Auto-collected on 2026-10-23*

Tags

#3d-spatial-reasoning#voxel-programming#language-models#object-placement#computer-vision#benchmark#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178636336