MambaLiteSR: Image Super-Resolution with Low-Rank Mamba using Knowledge Distillation
Romina Aalishah, Mozhgan Navardi, and Tinoosh Mohsenin · 2025
Read MambaLiteSR: Image Super-Resolution with Low-Rank Mamba using Knowledge Distillation by Romina Aalishah, Mozhgan Navardi, and Tinoosh Mohsenin (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-2502-14090.txt.
Description
Romina Aalishah, Mozhgan Navardi, and Tinoosh Mohsenin present a 2025 paper on MambaLiteSR, a small model that sharpens low-resolution images for edge devices. Phones and cameras need super-resolution in real time, but strong models are too heavy for their chips and memory. The design builds on the Mamba architecture, a sequence model known for efficiency.
To shrink it further, the team compresses the model's internal blocks to low rank. They also use knowledge distillation: a larger teacher model's know-how is transferred into the small student through careful tuning. The aim is a sharpener that keeps most of the quality while fitting the tight budgets of edge hardware.
Catalog record
| CREATOR | Romina Aalishah, Mozhgan Navardi, and Tinoosh Mohsenin |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-14090