According to the arXiv abstract, 3D editing seeks to modify an existing 3D asset according to user intent while keeping geometry and appearance outside the intended edit region intact. The authors note that current methods often rely on task-specific training, source-3D inversion, or other constraints that limit flexibility.

The new method, UMAS3D, is described as a unified mask-free editing framework with region-adaptive source preservation. This suggests it avoids the need for explicit masks and adapts how much of the original source is preserved depending on the region being edited.

Because the abstract is truncated, details on evaluation and comparisons are not available in the source. The contribution, as stated, is a single framework that handles editing without masks while maintaining source fidelity outside the edit area.