Hyperbolic Hierarchical Alignment Reasoning Network for Text-3D Retrieval
Wenrui Li, Yidan Lu, Yeyu Chai, Rui Zhao, Hengyu Man, Xiaopeng Fan · 2025
Read Hyperbolic Hierarchical Alignment Reasoning Network for Text-3D Retrieval by Wenrui Li, Yidan Lu, Yeyu Chai, Rui Zhao, Hengyu Man, Xiaopeng Fan (2025) 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-2511-11045.txt.
Description
Wenrui Li, Yidan Lu, and their colleagues present a network that matches written queries to 3D objects while keeping rank and part structure intact. Flat embedding spaces tend to crush layered links, from broad class down to fine part, and noisy fragments can drown the few cues that matter. Both flaws make hard near-matches tough to tell apart.
Their network places text and 3D data in a curved hyperbolic space, where volume growth keeps layered gaps open. A cone-based loss forces each matched object to sit inside the cone of its query text. A weighting step scores each local feature by relevance before merging, which protects key cues from noise.
Catalog record
| CREATOR | Wenrui Li, Yidan Lu, Yeyu Chai, Rui Zhao, Hengyu Man, Xiaopeng Fan |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-11045