pith. machine review for the scientific record. sign in

arxiv: 2510.06452 · v2 · submitted 2025-10-07 · 💻 cs.HC

Code Semantic Zooming

Jinsheng Ba , Sverrir Thorgeirsson , Zhendong Su This is my paper

Pith reviewed 2026-05-18 08:40 UTC · model grok-4.3

classification 💻 cs.HC
keywords code semantic zoomingpseudocodeLLM-assisted codingcode comprehensionuser studyhuman-AI interactionsoftware abstraction
0
0 comments X

The pith

CodeZoom uses layered pseudocode to give developers more control over LLM-generated code than natural language prompts allow.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes Code Semantic Zooming, or CodeZoom, as a pseudocode system that lets developers move between multiple layers of code abstraction to explore, understand, and refine LLM outputs. This approach aims to solve the problem of limited control when using plain English to generate complex software. A within-subjects study with 26 participants showed that CodeZoom performed as well as the Claude Code agent on usability measures. At the same time it produced a large improvement in code comprehension, with more than 90 percent of users saying they felt greater ownership over design decisions. The work extends the historical pattern of adding higher abstractions to handle growing software complexity.

Core claim

Code Semantic Zooming (CodeZoom) is a novel approach based on pseudocode that allows developers to iteratively explore, understand, and refine code across multiple layers of semantic abstraction. In a within-subjects user study (n=26), our method matches a state-of-the-art coding agent, Claude Code, on usability while producing a large effect on code comprehension: over 90% of participants reported feeling more in control of design decisions when using CodeZoom compared to using Claude Code.

What carries the argument

Code Semantic Zooming (CodeZoom): a pseudocode interface for iterative exploration and refinement across multiple semantic abstraction layers in LLM-assisted coding.

If this is right

  • Developers can adjust and verify code at chosen abstraction levels instead of relying on a single prompt.
  • Code comprehension and sense of design ownership increase while usability remains comparable to current agents.
  • Iterative refinement across layers supports building complex systems without full manual rewriting.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The layering idea could transfer to other generative AI tasks to give users similar steering ability.
  • Integration into mainstream IDEs might change how teams review and accept AI contributions at scale.
  • Repeated use could help less experienced programmers learn design patterns through guided abstraction changes.

Load-bearing premise

The within-subjects study with 26 participants isolates the specific effect of the CodeZoom interface from order effects, individual differences, and task variations.

What would settle it

A follow-up study using a between-subjects design or larger sample that finds no difference in perceived control or comprehension between CodeZoom and Claude Code would undermine the reported advantage.

Figures

Figures reproduced from arXiv: 2510.06452 by Jinsheng Ba, Sverrir Thorgeirsson, Zhendong Su.

Figure 1
Figure 1. Figure 1: Existing vibe-coding and CodeZoom to implement a 2048 game. of 99.999%, repeated rounds of generation will compound errors, ultimately reducing overall correctness to an unacceptable level. We envision the need for a high-level abstract language to pro￾vide greater controllability in LLM-assisted code writing. Such a language should be as easy for humans to understand as natural language, while also encodi… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: A screenshot of CodeZoom implementation as a VSCode extension. the strict syntax rules of a programming language. As a result, there is no universally accepted standard grammar for pseudocode. Existing pseudocode styles typically follow practical guidance for specific requirements, such as teaching and education [46], without a strict or consistent grammar. Such guidance is insufficient for our setting bec… view at source ↗
Figure 5
Figure 5. Figure 5: Case 1: adding the feature of filtering malicious [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Case 2: Generating and revising an SQL parser from [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
read the original abstract

Recent advances in Large Language Models (LLMs) have introduced a new paradigm for software development, where source code is generated from natural language prompts. While this paradigm significantly boosts development productivity, building complex, real-world software systems remains challenging because natural language offers limited control over the code generation process. Inspired by the historical evolution of programming languages toward higher levels of abstraction, we advocate for a high-level abstraction language that gives developers greater control over LLM-assisted code writing. To this end, we propose Code Semantic Zooming (CodeZoom), a novel approach based on pseudocode that allows developers to iteratively explore, understand, and refine code across multiple layers of semantic abstraction. In a within-subjects user study (n=26), our method matches a state-of-the-art coding agent, Claude Code, on usability while producing a large effect on code comprehension: over 90% of participants reported feeling more in control of design decisions when using CodeZoom compared to using Claude Code.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper proposes Code Semantic Zooming (CodeZoom), a multi-layer pseudocode abstraction technique intended to give developers iterative control over LLM-generated code by allowing exploration and refinement across semantic levels. It evaluates the approach via a within-subjects user study (n=26) that compares CodeZoom against the Claude Code agent, claiming equivalent usability but a large positive effect on code comprehension and control, with over 90% of participants reporting greater control over design decisions when using CodeZoom.

Significance. If the user-study results prove robust after addressing methodological gaps, the work would offer a concrete, historically grounded mechanism for increasing developer agency in LLM-assisted coding without usability penalties. This could influence interface design for future coding agents by demonstrating the value of explicit abstraction layers.

major comments (3)
  1. [User Study] User Study section: the headline result (>90% of participants reporting greater control) is presented without any description of task counterbalancing, order randomization, or statistical test (e.g., exact p-value or effect size) for the binary control item. In a within-subjects design these omissions leave open the possibility that learning, fatigue, or task-order effects fully explain the difference, so the causal attribution to the CodeZoom interface is not yet demonstrated.
  2. [User Study] User Study section: only subjective Likert-style or binary self-report measures are described; no objective comprehension or performance metrics (code-recall accuracy, successful modification rate, or time-to-correctness) are reported. This weakens the claim of a “large effect on code comprehension.”
  3. [§3] §3 (System Description): the multi-layer pseudocode abstraction is introduced as the key innovation, yet the manuscript contains no ablation that isolates the contribution of the zooming layers versus the baseline pseudocode representation itself.
minor comments (2)
  1. [Figures] Figure captions for the semantic-zooming examples could be expanded to explicitly label each abstraction layer and the refinement operations performed between them.
  2. [Related Work] The related-work section would benefit from citing recent empirical studies on LLM coding-agent usability (e.g., those evaluating Copilot or Cursor) to strengthen the positioning of the Claude Code baseline.

Simulated Author's Rebuttal

3 responses · 0 unresolved

Thank you for the referee's thoughtful and constructive feedback. We address each major comment below, clarifying our study design where details were omitted and outlining planned revisions to improve methodological transparency and strengthen the evaluation.

read point-by-point responses

  1. Referee: [User Study] User Study section: the headline result (>90% of participants reporting greater control) is presented without any description of task counterbalancing, order randomization, or statistical test (e.g., exact p-value or effect size) for the binary control item. In a within-subjects design these omissions leave open the possibility that learning, fatigue, or task-order effects fully explain the difference, so the causal attribution to the CodeZoom interface is not yet demonstrated.

    Authors: We agree that explicit reporting of these controls is necessary to support causal claims in a within-subjects design. The study procedure incorporated counterbalancing of task order across conditions and randomization of presentation order to mitigate learning and fatigue effects. We will revise the User Study section to describe these procedures in detail and will report the statistical analysis for the binary control item (including exact p-value and effect size, using an appropriate paired test such as McNemar's test). revision: yes

  2. Referee: [User Study] User Study section: only subjective Likert-style or binary self-report measures are described; no objective comprehension or performance metrics (code-recall accuracy, successful modification rate, or time-to-correctness) are reported. This weakens the claim of a “large effect on code comprehension.”

    Authors: We recognize that objective metrics would provide additional support for the comprehension claims. While the primary focus was on perceived control and comprehension (standard in HCI evaluations of developer agency), we also recorded task completion times and rates of successful code modifications. In the revision we will report these objective performance measures, including any relevant effect sizes, and integrate them with the subjective results to give a fuller picture of the observed differences. revision: yes

  3. Referee: [§3] §3 (System Description): the multi-layer pseudocode abstraction is introduced as the key innovation, yet the manuscript contains no ablation that isolates the contribution of the zooming layers versus the baseline pseudocode representation itself.

    Authors: We thank the referee for highlighting the value of isolating the multi-layer zooming component. Our evaluation was designed to compare the full CodeZoom system against a state-of-the-art LLM coding agent (Claude Code) rather than against a single-layer pseudocode baseline. Adding an ablation condition would have required a substantially larger sample and additional experimental arms. We will expand the limitations and future-work sections to explicitly discuss this design decision and outline a planned follow-up study that isolates the contribution of the semantic zooming layers. revision: partial

Circularity Check

0 steps flagged

No circularity in empirical user study

full rationale

The paper reports results from a within-subjects user study (n=26) comparing CodeZoom to Claude Code on usability and code comprehension metrics. No equations, mathematical derivations, fitted parameters, or predictions are present that could reduce to inputs by construction. Claims rest on participant self-reports and direct comparisons rather than any self-definitional, self-citation load-bearing, or ansatz-smuggling patterns. The study is self-contained as an external empirical evaluation with no internal derivation chain to inspect for circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on standard assumptions about user study validity rather than new mathematical axioms or invented physical entities. No free parameters are fitted to data in the reported result.

axioms (1)
  • domain assumption Within-subjects user studies with n=26 can reliably detect differences in usability and perceived control between interfaces.
    Invoked implicitly when reporting the study outcome as evidence for the method's benefits.
invented entities (1)
  • CodeZoom multi-layer pseudocode abstraction no independent evidence
    purpose: To provide iterative control over LLM code generation at varying semantic levels.
    The paper introduces this as the novel approach; no independent falsifiable evidence outside the user study is described.

pith-pipeline@v0.9.0 · 5691 in / 1359 out tokens · 32711 ms · 2026-05-18T08:40:36.560483+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages · 5 internal anchors

  1. [1]

    Ashraf Uddin, and Linta Islam

    Uzzal Kumar Acharjee, Minhazul Arefin, Kazi Mojammel Hossen, Mo- hammed Nasir Uddin, Md. Ashraf Uddin, and Linta Islam. Sequence-to-sequence learning-based conversion of pseudo-code to source code using neural translation approach.IEEE Access, 10:26730–26742, 2022

    work page 2022

  2. [2]

    Ontology- based automatic code generation from pseudocode

    Raed Al-Msie’deen, Mohammad Alshraideh, and Hatem Alsharafat. Ontology- based automatic code generation from pseudocode. InInternational Conference on Interactive Mobile Communication, Technologies and Learning (IMCL), pages 729–738. Springer, 2020

    work page 2020

  3. [3]

    Pseudocode generation from source code using the bart model.Mathematics, 10(21):3967, 2022

    Anas Alokla, Walaa Gad, Waleed Nazih, Mustafa Aref, and Abdel-badeeh Salem. Pseudocode generation from source code using the bart model.Mathematics, 10(21):3967, 2022

    work page 2022

  4. [4]

    Retrieval-based transformer pseudocode generation.Mathematics, 10(4):604, 2022

    Anas Alokla, Walaa Gad, Waleed Nazih, Mustafa Aref, and Abdel-Badeeh Salem. Retrieval-based transformer pseudocode generation.Mathematics, 10(4):604, 2022

    work page 2022

  5. [5]

    Elicitron: A large language model agent-based simulation framework for design requirements elicitation.J

    Mohammadmehdi Ataei, Hyunmin Cheong, Daniele Grandi, Ye Wang, Nigel Morris, and Alexander Tessier. Elicitron: A large language model agent-based simulation framework for design requirements elicitation.J. Comput. Inf. Sci. Eng., 25(1), 2025

    work page 2025

  6. [6]

    John W. Backus. The history of fortran i, ii, and III. InHistory of Programming Languages, from the ACM SIGPLAN History of Programming Languages Conference, June 1-3, 1978, Los Angeles, California, USA, pages 25–74, 1978

    work page 1978

  7. [7]

    Codeplan: Repository-level coding using llms and planning.Proceedings of the ACM on Software Engineering, 1(FSE):675–698, 2024

    Ramakrishna Bairi, Atharv Sonwane, Aditya Kanade, Vageesh D C, Arun Iyer, Suresh Parthasarathy, Sriram Rajamani, Balasubramanyan Ashok, and Shashank Shet. Codeplan: Repository-level coding using llms and planning.Proceedings of the ACM on Software Engineering, 1(FSE):675–698, 2024

    work page 2024

  8. [8]

    James, and Nadia Polikarpova

    Shraddha Barke, Michael B. James, and Nadia Polikarpova. Grounded copilot: How programmers interact with code-generating models.Proc. ACM Program. Lang., 7:85–111, 2023

    work page 2023

  9. [9]

    Raymond P. L. Buse and Westley Weimer. Automatic documentation inference for exceptions. InProceedings of the ACM/SIGSOFT International Symposium Code Semantic Zooming Conference’17, July 2017, Washington, DC, USA on Software Testing and Analysis, ISSTA 2008, Seattle, W A, USA, July 20-24, 2008, pages 273–282, 2008

    work page 2017

  10. [10]

    Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Pondé de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott Gray, Nick Ryder, Mikhail Pavlov, Alethea Power, Lukasz Kaiser, Mohammad Bavarian...

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  11. [11]

    Structured programming

    Ole-Johan Dahl, Edsger Wybe Dijkstra, and Charles Antony Richard Hoare. Structured programming. Academic Press Ltd., 1972

    work page 1972

  12. [12]

    C. J. Date.A Guide to the SQL Standard, Second Edition. 1989

    work page 1989

  13. [13]

    An empirical compari- son of dependency network evolution in seven software packaging ecosystems

    Alexandre Decan, Tom Mens, and Philippe Grosjean. An empirical compari- son of dependency network evolution in seven software packaging ecosystems. Empirical Software Engineering, 24(1):381–416, 2019

    work page 2019

  14. [14]

    Dijkstra

    Edsger W. Dijkstra. Letters to the editor: go to statement considered harmful. Commun. ACM, 11:147–148, 1968

    work page 1968

  15. [15]

    Dlbt: Deep learning-based transformer to generate pseudo-code from source code.Computers, Materials & Continua, 70(2), 2022

    Walaa Gad, Anas Alokla, Waleed Nazih, Mustafa Aref, and Abdel-badeeh Salem. Dlbt: Deep learning-based transformer to generate pseudo-code from source code.Computers, Materials & Continua, 70(2), 2022

    work page 2022

  16. [16]

    BugScope: Learn to Find Bugs Like Human

    Jinyao Guo, Chengpeng Wang, Dominic Deluca, Jinjie Liu, Zhuo Zhang, and Xiangyu Zhang. Bugscope: Learn to find bugs like human.CoRR, abs/2507.15671, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  17. [17]

    Multi-paradigm declarative languages

    Michael Hanus. Multi-paradigm declarative languages. InLogic Programming, 23rd International Conference, ICLP 2007, Porto, Portugal, September 8-13, 2007, Proceedings, volume 4670, pages 45–75, 2007

    work page 2007

  18. [18]

    AgentCoder: Multi-Agent-based Code Generation with Iterative Testing and Optimisation

    Dong Huang, Qingwen Bu, Jie M. Zhang, Michael Luck, and Heming Cui. Agent- coder: Multi-agent-based code generation with iterative testing and optimisation. CoRR, abs/2312.13010, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  19. [19]

    Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik R

    Carlos E. Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik R. Narasimhan. Swe-bench: Can language models resolve real-world github issues? InThe Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024, 2024

    work page 2024

  20. [20]

    Alan C. Kay. The early history of smalltalk. InHistory of Programming Languages Conference (HOPL-II), Preprints, Cambridge, Massachusetts, USA, April 20-23, 1993, pages 69–95, 1993

    work page 1993

  21. [21]

    Codeaid: Evaluating a classroom deployment of an llm-based programming assistant that balances student and educator needs

    Majeed Kazemitabaar, Runlong Ye, Xiaoning Wang, Austin Zachary Henley, Paul Denny, Michelle Craig, and Tovi Grossman. Codeaid: Evaluating a classroom deployment of an llm-based programming assistant that balances student and educator needs. InProceedings of the CHI Conference on Human Factors in Com- puting Systems, CHI 2024, Honolulu, HI, USA, May 11-16,...

    work page 2024

  22. [22]

    Literate programming.The computer journal, 27(2):97–111, 1984

    Donald Ervin Knuth. Literate programming.The computer journal, 27(2):97–111, 1984

    work page 1984

  23. [23]

    Spoc: Search-based pseudocode to code

    Sumith Kulal, Panupong Pasupat, Kartik Chandra, Mina Lee, Oded Padon, Alex Aiken, and Percy Liang. Spoc: Search-based pseudocode to code. InAdvances in Neural Information Processing Systems 32: Annual Conference on Neural Informa- tion Processing Systems 2019, NeurIPS 2019, December 8-14, 2019, Vancouver, BC, Canada, pages 11883–11894, 2019

    work page 2019

  24. [24]

    Kumar et al

    R. Kumar et al. Machine learning for translating pseudocode to python: A comprehensive review. InInternational Conference on Intelligent Computing, Communication and Security (ICICCS), 2023

    work page 2023

  25. [25]

    BART: denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension

    Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. BART: denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. InProceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL 2020, Online, Ju...

    work page 2020

  26. [26]

    Large Language Model-Based Agents for Software Engineering: A Survey

    Junwei Liu, Kaixin Wang, Yixuan Chen, Xin Peng, Zhenpeng Chen, Lingming Zhang, and Yiling Lou. Large language model-based agents for software engi- neering: A survey.CoRR, abs/2409.02977, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  27. [27]

    John W. Lloyd. Practical advtanages of declarative programming. In1994 Joint Conference on Declarative Programming, GULP-PRODE’94 Peñiscola, Spain, Sep- tember 19-22, 1994, Volume 1, pages 18–30, 1994

    work page 1994

  28. [28]

    Specgen: Automated generation of formal program specifications via large language models

    Lezhi Ma, Shangqing Liu, Yi Li, Xiaofei Xie, and Lei Bu. Specgen: Automated generation of formal program specifications via large language models. In47th IEEE/ACM International Conference on Software Engineering, ICSE 2025, Ottawa, ON, Canada, April 26 - May 6, 2025, pages 16–28, 2025

    work page 2025

  29. [29]

    Pollock, and K

    Laura Moreno, Jairo Aponte, Giriprasad Sridhara, Andrian Marcus, Lori L. Pollock, and K. Vijay-Shanker. Automatic generation of natural language summaries for java classes. InIEEE 21st International Conference on Program Comprehension, ICPC 2013, San Francisco, CA, USA, 20-21 May, 2013, pages 23–32, 2013

    work page 2013

  30. [30]

    Learning to generate pseudo-code from source code using statistical machine translation (T)

    Yusuke Oda, Hiroyuki Fudaba, Graham Neubig, Hideaki Hata, Sakriani Sakti, Tomoki Toda, and Satoshi Nakamura. Learning to generate pseudo-code from source code using statistical machine translation (T). In Myra B. Cohen, Lars Grunske, and Michael Whalen, editors,30th IEEE/ACM International Conference on Automated Software Engineering, ASE 2015, Lincoln, NE...

    work page 2015

  31. [31]

    C3po: A lightweight copying mechanism for pseudocode- to-code translation

    Authors of C3PO. C3po: A lightweight copying mechanism for pseudocode- to-code translation. InProceedings of the AACL-IJCNLP 2022 Student Research Workshop, 2022

    work page 2022

  32. [32]

    Richard E. Pattis. Teaching EBNF first in CS 1. In Robert Beck and Don Goelman, editors,Proceedings of the 25th SIGCSE Technical Symposium on Computer Science Education, SIGCSE 1994, Phoenix, Arizona, USA, March 10-12, 1994, pages 300–303. ACM, 1994

    work page 1994

  33. [33]

    Object oriented programming.ACM Sigplan Notices, 17(9):51–57, 1982

    Tim Rentsch. Object oriented programming.ACM Sigplan Notices, 17(9):51–57, 1982

    work page 1982

  34. [34]

    The c programming language.Bell Sys

    Dennis M Ritchie, Stephen C Johnson, ME Lesk, BW Kernighan, et al. The c programming language.Bell Sys. Tech. J, 57(6):1991–2019, 1978

    work page 1991

  35. [35]

    Geoffrey G. Roy. Designing and explaining programs with a literate pseudocode. ACM J. Educ. Resour. Comput., 6:1, 2006

    work page 2006

  36. [36]

    Jean E. Sammet. The early history of COBOL.ACM SIGPLAN Notices, 13:121–161, 1978

    work page 1978

  37. [37]

    Pollock, and K

    Giriprasad Sridhara, Emily Hill, Divya Muppaneni, Lori L. Pollock, and K. Vijay- Shanker. Towards automatically generating summary comments for java methods. InASE 2010, 25th IEEE/ACM International Conference on Automated Software Engineering, Antwerp, Belgium, September 20-24, 2010, pages 43–52, 2010

    work page 2010

  38. [38]

    Pollock, and K

    Giriprasad Sridhara, Lori L. Pollock, and K. Vijay-Shanker. Automatically detect- ing and describing high level actions within methods. InProceedings of the 33rd International Conference on Software Engineering, ICSE 2011, Waikiki, Honolulu , HI, USA, May 21-28, 2011, pages 101–110, 2011

    work page 2011

  39. [39]

    An overview of c++

    Bjarne Stroustrup. An overview of c++. InProceedings of the 1986 SIGPLAN workshop on Object-oriented programming, pages 7–18, 1986

    work page 1986

  40. [40]

    Python as multi paradigm programming language.International Journal of Computer Applications, 177(31):38–42, 2020

    Nimit Thaker and Abhilash Shukla. Python as multi paradigm programming language.International Journal of Computer Applications, 177(31):38–42, 2020

    work page 2020

  41. [41]

    Claude code

    Website. Claude code. https://www.anthropic.com/claude-code, 2025. Accessed: 2025-09-08

    work page 2025

  42. [42]

    Website. Codex. https://openai.com/codex/, 2025. Accessed: 2025-09-08

    work page 2025

  43. [43]

    Website. Copilot. https://github.com/features/copilot, 2025. Accessed: 2025-09-08

    work page 2025

  44. [44]

    Website. Cursor. https://cursor.com/, 2025. Accessed: 2025-09-08

    work page 2025

  45. [45]

    Gemini cli

    Website. Gemini cli. https://cloud.google.com/gemini/docs/codeassist/gemini-cli,

    work page

  46. [46]

    Accessed: 2025-09-08

    work page 2025

  47. [47]

    Pseudocode guide for teachers

    Website. Pseudocode guide for teachers. https://www.cambridgeinternational. org/Images/697401-2026-pseudocode-guide-for-teachers.pdf, 2025. Accessed: 2025-09-08

    work page 2026

  48. [48]

    vibe coding

    Website. Vibe coding. https://x.com/karpathy/status/1886192184808149383, 2025. Accessed: 2025-09-08

    work page arXiv 2025

  49. [49]

    Fuzz4all: Universal fuzzing with large language models

    Chunqiu Steven Xia, Matteo Paltenghi, Jia Le Tian, Michael Pradel, and Lingming Zhang. Fuzz4all: Universal fuzzing with large language models. InProceedings of the 46th IEEE/ACM International Conference on Software Engineering, ICSE 2024, Lisbon, Portugal, April 14-20, 2024, pages 126:1–126:13, 2024

    work page 2024

  50. [50]

    Kernelgpt: Enhanced kernel fuzzing via large language models

    Chenyuan Yang, Zijie Zhao, and Lingming Zhang. Kernelgpt: Enhanced kernel fuzzing via large language models. InProceedings of the 30th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2, ASPLOS 2025, Rotterdam, Netherlands, 30 March 2025 - 3 April 2025, pages 560–573, 2025

    work page 2025

  51. [51]

    Fine-grained pseudo-code generation method via code feature extraction and transformer

    Guang Yang, Yanlin Zhou, Xiang Chen, and Chi Yu. Fine-grained pseudo-code generation method via code feature extraction and transformer. In28th Asia- Pacific Software Engineering Conference, APSEC 2021, Taipei, Taiwan, December 6-9, 2021, pages 213–222, 2021

    work page 2021

  52. [52]

    Master’s thesis, Sudan University of Science and Technology, 2014

    Fatima Mohammed Rafie Younis.Automatic Pseudocode to Source Code. Master’s thesis, Sudan University of Science and Technology, 2014

    work page 2014

  53. [53]

    Retrieval-based transformer for pseudocode generation.Mathe- matics, 10(4):604, 2021

    Wei Zhang et al. Retrieval-based transformer for pseudocode generation.Mathe- matics, 10(4):604, 2021

    work page 2021

  54. [54]

    A Survey of Large Language Models

    Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, et al. A survey of large language models.arXiv preprint arXiv:2303.18223, 1(2), 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023