# The Code Barrier: What LLMs Actually Understand?

arXiv: https://arxiv.org/abs/2504.10557 · HTML: https://mlsanity.com/paper/2504.10557/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 52 references that are in the map.
- First posted on arXiv 14 Apr 2025.

## Builds on (most cited first, 19 of 19)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [On the Opportunities and Risks of Foundation Models](https://mlsanity.com/paper/2108.07258/) (2021)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [Code Llama: Open Foundation Models for Code](https://mlsanity.com/paper/2308.12950/) (2023)
- [Beyond Accuracy: Behavioral Testing of NLP models with CheckList](https://mlsanity.com/paper/2005.04118/) (2020)
- [Measuring Coding Challenge Competence With APPS](https://mlsanity.com/paper/2105.09938/) (2021)
- [CodeXGLUE: A Machine Learning Benchmark Dataset for Code Understanding and Generation](https://mlsanity.com/paper/2102.04664/) (2021)
- [CodeSearchNet Challenge: Evaluating the State of Semantic Code Search](https://mlsanity.com/paper/1909.09436/) (2019)
- [Asleep at the Keyboard? Assessing the Security of GitHub Copilot's Code Contributions](https://mlsanity.com/paper/2108.09293/) (2021)
- [IntelliCode Compose: Code Generation Using Transformer](https://mlsanity.com/paper/2005.08025/) (2020)
- [CodeNet: A Large-Scale AI for Code Dataset for Learning a Diversity of Coding Tasks](https://mlsanity.com/paper/2105.12655/) (2021)
- [The Next Decade in AI: Four Steps Towards Robust Artificial Intelligence](https://mlsanity.com/paper/2002.06177/) (2020)
- [CodeGemma: Open Code Models Based on Gemma](https://mlsanity.com/paper/2406.11409/) (2024)
- [Contrastive Code Representation Learning](https://mlsanity.com/paper/2007.04973/) (2020)
- [Using Large Language Models to Enhance Programming Error Messages](https://mlsanity.com/paper/2210.11630/) (2022)
- [SelfCodeAlign: Self-Alignment for Code Generation](https://mlsanity.com/paper/2410.24198/) (2024)
- [XLCoST: A Benchmark Dataset for Cross-lingual Code Intelligence](https://mlsanity.com/paper/2206.08474/) (2022)
- [Large Language Models for Code Analysis: Do LLMs Really Do Their Job?](https://mlsanity.com/paper/2310.12357/) (2023)

## Built on by (most cited first, 5 of 5)

- [Trust Me, I Know This Function: Hijacking LLM Static Analysis using Bias](https://mlsanity.com/paper/2508.17361/) (2025)
- [Is "Knowing It's Malicious Enough?" Evaluating LLMs for Fine-Grained Malware Behavior Auditing](https://mlsanity.com/paper/2509.14335/) (2025)
- [Optimizing Token Consumption in LLMs: A Nano Surge Approach for Code Reasoning Efficiency](https://mlsanity.com/paper/2504.15989/) (2025)
- [(How) Do Large Language Models Understand High-Level Message Sequence Charts?](https://mlsanity.com/paper/2605.13773/) (2026)
- [Poisoned Identifiers Survive LLM Deobfuscation: A Case Study on Claude Opus 4.6](https://mlsanity.com/paper/2604.04289/) (2026)

## Abstract (arXiv)

Understanding code represents a core ability needed for automating software development tasks. While foundation models like LLMs show impressive results across many software engineering challenges, the extent of their true semantic understanding beyond simple token recognition remains unclear. This research uses code obfuscation as a structured testing framework to evaluate LLMs' semantic understanding capabilities. We methodically apply controlled obfuscation changes to source code and measure comprehension through two complementary tasks: generating accurate descriptions of obfuscated code and performing deobfuscation, a skill with important implications for reverse engineering applications. Our testing approach includes 13 cutting-edge models, covering both code-specialized (e.g., StarCoder2) and general-purpose (e.g., GPT-4o) architectures, evaluated on a benchmark created from CodeNet and consisting of filtered 250 Java programming problems and their solutions. Findings show a statistically significant performance decline as obfuscation complexity increases, with unexpected resilience shown by general-purpose models compared to their code-focused counterparts. While some models successfully identify obfuscation techniques, their ability to reconstruct the underlying program logic remains constrained, suggesting limitations in their semantic representation mechanisms. This research introduces a new evaluation approach for assessing code comprehension in language models and establishes empirical baselines for advancing research in security-critical code analysis applications such as reverse engineering and adversarial code analysis.