Breaking the Modality Barrier: Universal Embedding Learning with Multimodal LLMs
Tiancheng Gu ♥ {}^{\text{♥}} * , Kaicheng Yang ♠ {}^{\text{♠}} * , Ziyong Feng ♠ {}^{\text{♠}} , Xingjun Wang ♣ , Yanzhao Zhang ♣ , Dingkun Long ♣ , Yingda Chen ♣ , Weidong Cai ♥ {}^{\text{♥}} , Jiankang Deng ⧫ † \dagger ♥ {}^{\text{♥}} The University of Sydney ♠ {}^{\text{♠}} DeepGlint ♣ Tongyi Lab · 2025
Read Breaking the Modality Barrier: Universal Embedding Learning with Multimodal LLMs by Tiancheng Gu ♥ {}^{\text{♥}} * , Kaicheng Yang ♠ {}^{\text{♠}} * , Ziyong Feng ♠ {}^{\text{♠}} , Xingjun Wang ♣ , Yanzhao Zhang ♣ , Dingkun Long ♣ , Yingda Chen ♣ , Weidong Cai ♥ {}^{\text{♥}} , Jiankang Deng ⧫ † \dagger ♥ {}^{\text{♥}} The University of Sydney ♠ {}^{\text{♠}} DeepGlint ♣ Tongyi Lab (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-2504-17432.txt.
Description
One embedding space for text, images, and more is built by Tiancheng Gu and co-authors in a 2025 paper. The classic image-text model truncates long text, encodes each side alone, and treats word order loosely. Their framework instead borrows a multimodal chat model's broader grasp.
Stage one distills a strong text teacher's discrimination into the model. Stage two tunes it with hard near-miss examples sorted by instruction. The result aims at transfer across retrieval and grouping tasks beyond photos and captions. A universal embedding would let one index serve search across every media type a library holds.
Catalog record
| CREATOR | Tiancheng Gu ♥ {}^{\text{♥}} * , Kaicheng Yang ♠ {}^{\text{♠}} * , Ziyong Feng ♠ {}^{\text{♠}} , Xingjun Wang ♣ , Yanzhao Zhang ♣ , Dingkun Long ♣ , Yingda Chen ♣ , Weidong Cai ♥ {}^{\text{♥}} , Jiankang Deng ⧫ † \dagger ♥ {}^{\text{♥}} The University of Sydney ♠ {}^{\text{♠}} DeepGlint ♣ Tongyi Lab |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-17432