mlsanityData to 7 Oct 2026

What is the Difference Between Me and You? Benchmarking the Quality Gap Between Human-Written and AI-Generated Code

arXiv 2609.12708 ↗First posted 11 Sep 2026cs.SE cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

AI coding assistants are becoming co-authors of production software, yet their evaluation centers on functional correctness, leaving open whether their code differs from human code in the quality dimensions dominating lifecycle cost. We compare human-written and AI-generated code at scale: 787,562 function pairs across Python, Java, and C, each human function mined from open-source repositories paired with implementations generated from its docstring by three AI assistants (OpenAI GPT models, DeepSeek-Coder, Qwen2.5-Coder). We characterize structural complexity and statistical naturalness, and map static-analysis findings onto Orthogonal Defect Classification for defects and the Common Weakness Enumeration for vulnerabilities, making authors and languages directly comparable. AI-generated code is structurally compressed and stylistically templated: roughly half the size and branching of human code, clustering apart at the style level. Defect profiles differ in kind: human code concentrates issues of mature codebases, AI code repetitive boilerplate; security is language-dependent, with LLMs producing more, and more severe, findings in Python and Java but fewer high-severity memory-safety findings than humans in C. Once size is controlled for, complexity metrics carry little signal, while naturalness separates authors. Finally, we release CQBench, a benchmark of 27,346 issue-prone tasks with baselines and an evaluation pipeline for quality assurance and security testing.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 32 of its 72 references that are in the map. First posted on arXiv 11 Sep 2026.

Builds on

32 of 72 references.

References · 32

  1. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  2. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,767
  3. 2021Program Synthesis with Large Language Modelscited by 3,509
  4. 2023Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generationcited by 1,332
  5. 2024DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligencecited by 1,154
  6. 2024Qwen2.5-Coder Technical Reportcited by 1,151
  7. 2021CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generationcited by 890
  8. 2021Measuring Coding Challenge Competence With APPScited by 879
  9. 2019CodeSearchNet Challenge: Evaluating the State of Semantic Code Searchcited by 638
  10. 2021Asleep at the Keyboard? Assessing the Security of GitHub Copilot's Code Contributionscited by 417
  11. 2024Infini-gram: Scaling Unbounded n-gram Language Models to a Trillion Tokenscited by 148
  12. 2022Productivity Assessment of Neural Code Completioncited by 97
20 more
  1. 2023Refining ChatGPT-Generated Code: Characterizing and Mitigating Code Quality Issuescited by 78
  2. 2024Bugs in Large Language Models Generated Code: An Empirical Studycited by 59
  3. 2024FullStack Bench: Evaluating LLMs as Full Stack Coderscited by 56
  4. 2025CWEval: Outcome-driven Evaluation on Functionality and Security of LLM Code Generationcited by 55
  5. 2024What's Wrong with Your Code Generated by Large Language Models? An Extensive Studycited by 52
  6. 2023Large Language Models and Simple, Stupid Bugscited by 45
  7. 2023Security Weaknesses of Copilot-Generated Code in GitHub Projects: An Empirical Studycited by 36
  8. 2015On the "Naturalness" of Buggy Codecited by 35
  9. 2025Security and Quality in LLM-Generated Code: A Multi-Language, Multi-Model Analysiscited by 18
  10. 2024Distinguishing LLM-generated from Human-written Code by Contrastive Learningcited by 16
  11. 2025Human-Written vs. AI-Generated Code: A Large-Scale Study of Defects, Vulnerabilities, and Complexitycited by 16
  12. 2025Is GPT-OSS Good? A Comprehensive Evaluation of OpenAI's Latest Open Source Modelscited by 11
  13. 2024An Interpretable N-gram Perplexity Threat Model for Large Language Model Jailbreakscited by 10
  14. 2025Investigating The Smells of LLM Generated Codecited by 9
  15. 2026Debt Behind the AI Boom: A Large-Scale Empirical Study of AI-Generated Code in the Wildcited by 9
  16. 2025Static Analysis as a Feedback Loop: Enhancing LLM-Generated Code Beyond Correctnesscited by 7
  17. 2025Assessing the Quality and Security of AI-Generated Code: A Quantitative Analysiscited by 5
  18. 2025Security Vulnerabilities in AI-Generated Code: A Large-Scale Analysis of Public GitHub Repositoriescited by 5
  19. 2025Show Me Your Code! Kill Code Poisoning: A Lightweight Method Based on Code Naturalnesscited by 3
  20. 2025Quality In, Quality Out: Investigating Training Data's Role in AI Code Generationcited by 2

Built on by

Not cited yet.