# LLM4Decompile (EMNLP 2024)

> Research on direct binary decompilation and refinement of Ghidra output with language models. Evaluates re-executability on HumanEval and ExeBench; useful for distinguishing readable output from functional recovery.

- Type: Research
- Targets: Binary
- Category: Learning resources
- Relationship: Learning resource
- Sources reviewed: 2026-10-16
- Added to the catalog: 2026-10-16
- Original resource: <https://aclanthology.org/2024.emnlp-main.203/>
- Catalog page: <https://awesomerea.org/resources/llm4decompile-paper/>
- Record JSON: <https://awesomerea.org/resources/llm4decompile-paper/record.json>
- Chinese version: [简体中文](https://awesomerea.org/zh/resources/llm4decompile-paper/index.md)

## Requirements & limits

- EMNLP 2024 paper. Results are author-reported for the evaluated models and benchmarks.

## Primary sources

1. <https://aclanthology.org/2024.emnlp-main.203/>
2. <https://arxiv.org/abs/2403.05286>

## Related entries

Matched by the same code repository or publication identifier.

- [LLM4Decompile: Decompiling Binary Code with Large Language Models](https://awesomerea.org/papers/paper-llm4decompile/index.md) · Research paper

## More in References and learning

- [DecompileBench (Findings of ACL 2025)](https://awesomerea.org/resources/decompilebench-paper/index.md) · Research
- [ReSym (CCS 2024)](https://awesomerea.org/resources/resym-paper/index.md) · Research
- [Finding Jingle Town: Debugging an N64 Game without Symbols](https://awesomerea.org/resources/finding-jingle-town-n64-debugging/index.md) · Learning resource
- [JPCERT/CC: APK analysis and API probing on attack servers](https://awesomerea.org/resources/jpcert-apk-api-attack-infrastructure-report/index.md) · Learning resource
- [Reverse Engineering (mytechnotalent)](https://awesomerea.org/resources/mytechnotalent-reverse-engineering/index.md) · Learning resource
- [Z0FCourse Reverse Engineering](https://awesomerea.org/resources/z0f-reverse-engineering-course/index.md) · Learning resource

## About this entry

The description was reviewed against primary sources.
Original editorial content: MIT. Linked projects, papers and assets retain their own rights and licenses.