# D-LiFT: Improving LLM-based Decompiler Backend via Code Quality-driven Fine-tuning

> D-LIFT uses D-Score to guide LLM fine-tuning and output selection, improving decompiled code while prioritizing accuracy. It evaluates Ghidra output from coreutils and util-linux.

- Authors: Muqi Zou; Hongyu Cai; Hongwei Wu; Zion Leonahenahe Basque; Arslan Khan; Berkay Celik; Dave; Tian; Antonio Bianchi; Ruoyu; Wang; Dongyan Xu
- Venue: arXiv
- Status: Preprint
- Publication / posting date: June 11, 2025
- Date source: First arXiv submission date.
- First arXiv submission: June 11, 2025
- arXiv: 2506.10125
- Reviewed arXiv version: v4
- Sources reviewed: 2026-11-06
- Added to the catalog: 2026-11-06
- Original paper: <https://arxiv.org/abs/2506.10125>
- PDF: <https://arxiv.org/pdf/2506.10125>
- BibTeX: <https://awesomerea.org/papers/citations/paper-d-lift.bib>
- Catalog page: <https://awesomerea.org/papers/paper-d-lift/>
- Record JSON: <https://awesomerea.org/papers/paper-d-lift/record.json>
- Chinese version: [简体中文](https://awesomerea.org/zh/papers/paper-d-lift/index.md)

## Scope and limitations

The evaluation covers functions from coreutils and util-linux, using Ghidra and a range of LLMs. D-Score measured 55.3% more improved functions than baseline LLMs and improvements to 68.2% of native decompiler functions.

## Why it fits this collection

It addresses the accuracy and readability of source code reconstructed from binaries.

## Publication and review notes

The date and venue come from the arXiv record.

## Primary sources

1. <https://arxiv.org/abs/2506.10125>

## About this entry

Reviewed against the paper and its primary sources. Cite the original authors using the linked paper or BibTeX.
Original editorial content: MIT. Linked projects, papers and assets retain their own rights and licenses.