DCE-LLM: Dead Code Elimination with Large Language Models
Minyu Chen, Guoqiang Li, Ling-I Wu, Ruibang Liu · 2025
Read DCE-LLM: Dead Code Elimination with Large Language Models by Minyu Chen, Guoqiang Li, Ling-I Wu, Ruibang Liu (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-2506-11076.txt.
Description
Minyu Chen, Guoqiang Li, Ling-I Wu, and Ruibang Liu sic language models on dead code, in a 2025 paper. Dead code, instructions that never run or results never used, bloats programs, misleads maintainers, and can hide malware tricks. Compilers catch simple cases, but sophisticated dead patterns slip through, and cleanup usually demands tedious manual work.
Their framework pairs a compact code model, which flags suspicious lines through attribution scoring, with large language models fine-tuned on a big annotated dataset. The large models judge each suspect, explain why it is dead, and write the patch. Scores for finding unused and unreachable code exceeded ninety-four percent, beating a leading general model by about thirty points. The system spans several languages and closes the loop from detection to correction automatically.
Catalog record
| CREATOR | Minyu Chen, Guoqiang Li, Ling-I Wu, Ruibang Liu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-11076