MEGA Hub

Frontier3DBench: Benchmarking Coding Agents for 3D Creation across Objects, Scenes, and Interactive Open Worlds

Authors

Do you know Xinjie Lin?You can claim authorship or link another user.Do you know Qingle Liu?You can claim authorship or link another user.Do you know Wenyi Li?You can claim authorship or link another user.Do you know John Lam?You can claim authorship or link another user.Do you know Xiaoqiu Wang?You can claim authorship or link another user.Do you know Deyao Hong?You can claim authorship or link another user.Do you know Mingju Gao?You can claim authorship or link another user.Do you know Zheng Jiang?You can claim authorship or link another user.Do you know Yizhe Chi?You can claim authorship or link another user.Do you know Youjie Zheng?You can claim authorship or link another user.Do you know Calvin Xiao?You can claim authorship or link another user.Do you know Kaisen Yang?You can claim authorship or link another user.Do you know Huan-ang Gao?You can claim authorship or link another user.

Abstract

Coding agents combine visual understanding with program execution, offering a new route to automated 3D creation. By translating images and instructions into editable programs, they bring reconstruction, generation, and editing into a shared workflow. Yet existing benchmarks fragment evaluation across task scopes and protocols, leaving their capabilities across construction levels unclear. We propose Frontier3DBench, a unified benchmark of 145 tasks across 9 categories spanning object, scene, and interactive open-world creation, with joint evaluation of appearance, structure, and behavior. Experiments with 11 model-harness configurations reveal strengths that vary by level: GPT-6.1-sol leads the object and scene levels and achieves the highest overall score of 66.02/100, whereas GPT-6-astra leads interactive world construction. However, all configurations score lower on reference reconstruction than on text-driven generation in object and scene tasks. Producing plausible content thus remains easier than reproducing specified geometry, layouts, and visual details. Moreover, generated worlds average 83.08/100 on local mechanisms but only 20.82 on motion clearance and 30.26 on stateful workflows: individual mechanisms often function, whereas complete activities that require coordinated actions and persistent state remain difficult. These results identify reference fidelity and coordinated activity completion as central bottlenecks for coding agents in 3D creation.

Community

00