English

DreamCS: Geometry-Aware Text-to-3D Generation with Unpaired 3D Reward Supervision

Computer Vision and Pattern Recognition 2026-03-23 v3

Abstract

While text-to-3D generation has attracted growing interest, existing methods often struggle to produce 3D assets that align well with human preferences. Current preference alignment techniques for 3D content typically rely on hardly-collected preference-paired multi-view 2D images to train 2D reward models, when then guide 3D generation -- leading to geometric artifacts due to their inherent 2D bias. To address these limitations, we construct 3D-MeshPref, the first large-scale unpaired 3D preference dataset, featuring diverse 3D meshes annotated by a large language model and refined by human evaluators. We then develop RewardCS, the first reward model trained directly on unpaired 3D-MeshPref data using a novel Cauchy-Schwarz divergence objective, enabling effective learning of human-aligned 3D geometric preferences without requiring paired comparisons. Building on this, we propose DreamCS, a unified framework that integrates RewardCS into text-to-3D pipelines -- enhancing both implicit and explicit 3D generation with human preference feedback. Extensive experiments show DreamCS outperforms prior methods, producing 3D assets that are both geometrically faithful and human-preferred. Code and models will be released publicly.

Keywords

Cite

@article{arxiv.2506.09814,
  title  = {DreamCS: Geometry-Aware Text-to-3D Generation with Unpaired 3D Reward Supervision},
  author = {Xiandong Zou and Ruihao Xia and Hongsong Wang and Pan Zhou},
  journal= {arXiv preprint arXiv:2506.09814},
  year   = {2026}
}