OASIS: Online Activation Subspace Learning for Memory-Efficient Training
Sakshi Choudhary, Utkarsh Saxena, Kaushik Roy · 2026
Read OASIS: Online Activation Subspace Learning for Memory-Efficient Training by Sakshi Choudhary, Utkarsh Saxena, Kaushik Roy (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-2604-09406.txt.
Description
Choudhary, Saxena, and Roy propose OASIS, an online activation subspace learning algorithm for memory-efficient training. Training large language models is constrained by memory, with activations taking a substantial share. Existing methods cut memory via low-rank weights or gradient subspaces, while activation memory uses architecture changes. OASIS tracks and updates a low-dimensional activation subspace during training, projecting activations onto it without changing forward computation. That yields low-rank gradients and states kept in the subspace, with up to twice lower peak memory than full fine-tuning.
Catalog record
| CREATOR | Sakshi Choudhary, Utkarsh Saxena, Kaushik Roy |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 32342 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.09406