Spectral-Aligned Pruning for Universal Error-Correcting Code Transformers
Sanghyeon Cho, Taewoo Park, Seong-Joon Park, Dae-Young Yun, Hee-Youl Kwak, Sang-Hyo Kim, Yongjune Kim · 2026
Read Spectral-Aligned Pruning for Universal Error-Correcting Code Transformers by Sanghyeon Cho, Taewoo Park, Seong-Joon Park, Dae-Young Yun, Hee-Youl Kwak, Sang-Hyo Kim, Yongjune Kim (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-2602-01602.txt.
Description
Sanghyeon Cho and colleagues shrink a transformer decoder for error correcting codes without giving up its cross code reach. The base model handles many code families with one backbone, but its size and compute cost block use. Their Spectral Aligned Pruning method reuses pruning masks across codes by reading two adjacency eigenvalues of the bipartite graph. Those two values form a signature for degree scale, expansion and minimum distance. After pruning, low rank adapters restore each code, so one backbone plus per code adapters serve many codes.
Catalog record
| CREATOR | Sanghyeon Cho, Taewoo Park, Seong-Joon Park, Dae-Young Yun, Hee-Youl Kwak, Sang-Hyo Kim, Yongjune Kim |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46917 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.01602