BrokenBind : Universal Modality Exploration beyond Dataset Boundaries
Zhuo Huang, Runnan Chen, Bo Han, Gang Niu, Masashi Sugiyama, Tongliang Liu, [1ex] Sydney AI Centre, The University of Sydney; Hong Kong Baptist University · 2026
Read BrokenBind : Universal Modality Exploration beyond Dataset Boundaries by Zhuo Huang, Runnan Chen, Bo Han, Gang Niu, Masashi Sugiyama, Tongliang Liu, [1ex] Sydney AI Centre, The University of Sydney; Hong Kong Baptist University (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-2602-06451.txt.
Description
Multimodal models usually learn by binding data types, like images and text, inside one data set. Zhuo Huang and 5 co-authors argue that this binding breaks when a new data type appears that the original set never held. Collecting fresh multimodal sets for every combination is costly.
Their framework, BrokenBind, binds two modalities that live in different data sets, joined through one shared modality present in both. Even though the sets do not correspond sample by sample, their relationship through the shared type lets the method hallucinate the missing embeddings for each side. That fills the gaps and frees pairing from single-set limits. Harder scenarios with more missing pieces test how far the trick stretches.
Catalog record
| CREATOR | Zhuo Huang, Runnan Chen, Bo Han, Gang Niu, Masashi Sugiyama, Tongliang Liu, [1ex] Sydney AI Centre, The University of Sydney; Hong Kong Baptist University |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-06451