Search papers, labs, and topics across Lattice.
This work develops a feed-forward instruction-guided 3D editing framework that operates entirely without paired 3D supervision by distilling visual, semantic, and geometric priors from foundation models. Addressing the twin failure modes of slow test-time optimization and artifact-ridden pseudo-paired datasets, the system enforces multi-view rendering consistency and identity preservation through combined 2D image-editing, vision-language, and 3D latent priors. Across standard benchmarks, the method achieves superior instruction fidelity and eliminates multi-view geometric collapse compared to both optimization-based and pseudo-pair-trained baselines.
High-fidelity, feed-forward 3D editing no longer requires paired 3D training data or agonizing test-time optimization: regularizing 2D visual and VLM distillation with a 3D latent distribution matching objective is enough to prevent geometric collapse.
Instruction-guided 3D editing is essential for interactive content creation, yet it faces a significant bottleneck: the severe scarcity of high-quality paired training data. Existing approaches attempt to bypass this by either relying on slow test-time optimization or training on pseudo-pairs constructed via complex pipelines, which often introduce structural drift and geometric artifacts. In this paper, we propose a novel framework that learns feed-forward 3D editing without paired 3D supervision via Generative Prior Distillation. Instead of relying on ground-truth 3D pairs, our core idea is to distill visual, semantic, and geometric knowledge from powerful foundation models directly into a 3D editing model. Specifically, through a differentiable rendering pipeline, we supervise the 3D representation using two complementary signals: a 2D visual prior from an image editing model at the main editing view, and a semantic prior from a Vision-Language Model at novel views to ensure strict instruction following and source identity preservation. Crucially, to address the geometric collapse and multi-view inconsistencies inherent in 2D projection supervision, we introduce a 3D-aware Distribution Matching regularization. Acting as a geometric prior, this term operates in the 3D latent space, constraining the edited output to remain within the manifold of realistic 3D assets defined by a pretrained image to 3D teacher model. Extensive experiments demonstrate that our method achieves superior instruction fidelity and cross-view consistency, significantly outperforming state-of-the-art baselines. Our project is available at: https://github.com/thiamine128/PriorEdit3D.
