Beyond Prompts: Unconditional 3D Inversion for Out-of-Distribution Shapes
Victoria Yue Chen, Emery Pierson, Léopold Maillard, Maks Ovsjanikov · 2026
Read Beyond Prompts: Unconditional 3D Inversion for Out-of-Distribution Shapes by Victoria Yue Chen, Emery Pierson, Léopold Maillard, Maks Ovsjanikov (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-14914.txt.
Description
Chen, Pierson, Maillard, and Ovsjanikov show that text-driven inversion of 3D generative models breaks down on new data. Sampling paths fall into latent sink traps where prompt changes stop changing internal codes or output shapes. This is not a limit of shape range: the models can produce a variety of shapes. Complex shapes remain reachable through the unconditional prior. The team builds a framework that splits shape-making power from language sensitivity, dodging sinks for high-quality meaning-based editing of out-of-distribution 3D shapes.
Catalog record
| CREATOR | Victoria Yue Chen, Emery Pierson, Léopold Maillard, Maks Ovsjanikov |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45601 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.14914