← Back to research
Paper
ELSA3D introduces elastic semantic anchoring for unified 3D understanding and generation
ELSA3D is a unified 3D foundation model that uses elastic semantic anchoring to jointly structure language and geometric reasoning across matched abstraction scales. It addresses the collapse of coarse and fine details in existing models that concatenate text and 3D tokens into flat sequences. The approach represents geometry with scale-aware features, enabling more explicit text-3D interaction within a single backbone.
0 engagement·1 source·Tue, Jul 7, 2026, 05:59 PM
The paper, posted on arXiv on 2026-07-07, proposes ELSA3D to overcome limitations of current unified 3D models that rely on self-attention over flat token sequences, which collapses structural and geometric details. ELSA3D introduces elastic semantic anchoring to align language and geometry at multiple abstraction scales, preserving both coarse structure and fine details. This enables more effective text-3D interaction for generation and reasoning tasks.