Pith. sign in

REVIEW 3 major objections 5 minor 110 references

Tool-Guided Retrieval-Augmented Repair for Securing LLM-Generated C Code

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A tool-guided repair pipeline substantially reduces compilation failures and security defects in LLM-generated C code, cutting CodeLlama 7B's CodeQL errors by 83.7% and DeepSeek Coder 1.3B's security issues from 35% to 15%.

desk verdict Plausible workflow, big claimed gains, but the evaluation lacks a self-repair control and has a suspicious number; not ready to trust the numbers as-is. read the letter →

arxiv 2607.21641 v1 pith:3YSLHZFC submitted 2026-07-21 cs.SE cs.AI

Vidyut Sriram , Saatvik Pradhan , Suman Saha This is my paper
classification cs.SEcs.AI
keywords LLMcodegenerationClanguagesecurityvulnerabilitiesretrieval-augmentedrepairstaticanalysissymbolicexecutioncompilationquality
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that a repair pipeline combining compilation diagnostics, CodeQL static analysis, KLEE symbolic execution, and retrieval from a growing repository of prior repair patterns can substantially improve the reliability and security of C code generated by LLMs, without retraining or larger models. On 5,000 algorithmic C tasks, the workflow reduces CodeLlama 7B's security defect rate from 49% to 19% and total CodeQL errors from 15,088 to 2,463 (83.7%), and cuts DeepSeek Coder 1.3B's compilation failures from 42% to 22% and security defects from 35% to 15%. The central idea is to treat repairs as the unit of memory: each attempt is scored and stored, so later generations retrieve distilled positive and negative guidance rather than raw code. If the results hold, lightweight tool feedback plus accumulated repair patterns could make LLM-generated C safer for embedded and resource-constrained settings.

What carries the argument

The engine is the execution-time repair repository. Each attempt is stored with a composite quality score (0.0 for non-compiling; 0.25 compiling; 0.50 compiling and KLEE-analyzable but with CodeQL findings; 0.75 compiling and CodeQL-clean; 1.0 compiling, CodeQL-clean, and KLEE-clean). For a new task, retrieval selects up to three positive entries above 0.60 and up to two negative compilable entries, distilling them into up to seven security practices and three avoidance hints rather than pasting code. A bounded repair loop (at most three iterations) recompiles and re-analyzes each candidate, and the highest-scoring version is chosen; the repository is deduplicated by cosine similarity and ca

What would settle it

Run the repaired programs under a stricter dynamic check—for example, compile with AddressSanitizer and feed symbolic or randomized inputs that reach scanf, strcpy, and malloc sites, then measure reachable assertions or memory violations before and after repair. If reachable crash rates do not fall in step with the reported CodeQL and KLEE metric improvements, the claim that the pipeline secures the code is undermined.

Watch

Extended reading notes

Core claim

The discovery is that a bounded, tool-guided repair loop—compile, analyze with CodeQL, retrieve prior repair outcomes, repair, and validate with KLEE—turns insecure, often non-compiling LLM output into measurably safer C code. The paper reports that on 5,000 general C tasks, CodeLlama 7B's security defect rate drops from 48.5% to 19.3% and its total CodeQL errors drop by 83.7%; among programs that compile, the security-clean rate rises from 10.9% to 73.4%. DeepSeek Coder 1.3B improves from 42.4% to 21.8% compilation failures and from 34.6% to 15.2% security defects, with KLEE analyzability rising from 56.8% to 77.4%. The gains are attributed to a composite quality score that prioritizes secu

Load-bearing premise

The load-bearing premise is that CodeQL findings and KLEE analyzability, which the pipeline scores and optimizes, are valid surrogates for the security and reliability that matter in embedded C; if those proxies correlate poorly with real-world vulnerabilities, the reported reductions may reflect warning-count optimization rather than actual safety gains.

Editorial extensions

If this is right

  • Security defect rates in LLM-generated C can fall by more than half using tool feedback and retrieval alone, with no model retraining or size increase.
  • Among compiled programs, CodeLlama's security-clean rate rises from about 11% to 73%, suggesting the pipeline fixes vulnerable code rather than merely improving build success.
  • KLEE analyzability improves alongside static cleanliness (DeepSeek from 56.8% to 77.4%), indicating more programs become amenable to symbolic validation.
  • Because the repository grows across tasks, the method's guidance should strengthen as more tasks are processed, up to the 500-entry cap.
  • The composite quality score, favoring CodeQL-clean and KLEE-clean programs, biases final selection toward statically and symbolically cleaner code.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The pipeline's optimization target is warning counts from CodeQL and KLEE analyzability; if those proxies correlate imperfectly with real embedded vulnerabilities, the headline reductions may overstate safety gains on actual firmware tasks.
  • The retrieval-to-text distillation step is a candidate explanation for why smaller models (DeepSeek 1.3B) benefit: substituting raw examples with concise guidance may reduce imitation of flawed patterns. This mechanism is plausible but not isolated by the paper's experiments.
  • A natural testable extension is to feed the same pipeline memory to larger models or other low-level languages (e.g., Rust or C++), where the repository of repair patterns could port or be rebuilt.
  • The roughly 60% alignment of remaining findings with a public software-flaw taxonomy suggests that even after repair, a substantial residue of exceptional-condition and buffer-overflow issues persists; whether those are reachable in the final programs is not established by CodeQL counts alone.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This paper presents a tool-guided, retrieval-augmented repair pipeline for LLM-generated C code. The pipeline combines compilation diagnostics, CodeQL static analysis, KLEE symbolic execution, and a growing repository of prior repair outcomes to iteratively refine code generated by DeepSeek Coder 1.3B and CodeLlama 7B. On a dataset of 5,000 general-purpose C programming tasks, the authors report that the pipeline reduces compilation failures (e.g., DeepSeek from 42.44% to 21.78%), CodeQL security issues (e.g., CodeLlama from 48.50% to 19.32%), and total CodeQL errors (CodeLlama from 15,088 to 2,463), while improving KLEE analyzability. The paper is explicitly a work in progress; component ablations and embedded-benchmark evaluation are deferred to future work.

Significance. If the reported results are correct and attributable to the proposed components, the work would demonstrate a practical, no-retraining method for improving the reliability of LLM-generated C code, especially for smaller models. The design of an execution-time repository that accumulates verified repair patterns and distills them into textual guidance is a useful contribution. The use of greedy decoding makes the experiments deterministic and reproducible in principle. However, the evaluation has two unresolved methodological issues—inconsistent metric definitions in the conditional-clean-rate analysis and the absence of a control separating tool-guided repair from generic iterative revision—that currently preclude the strong attribution claim in the title and abstract.

major comments (3)
  1. [§4.2, Table 1] The conditional security-clean rates reported in §4.2 are computed under inconsistent denominators. For CodeLlama, the baseline clean rate of 10.9% is consistent with 'security issues (%)' being the fraction of all tasks that have CodeQL findings (i.e., (54.44% compiled − 48.50% issues)/54.44% compiled). For DeepSeek, the baseline clean rate of 65.4% is only consistent with 'security issues (%)' being the fraction of compiled tasks with findings (1 − 0.3462). The same inconsistency appears in the RAG row: DeepSeek's 80.6% matches the formula (78.22% compiled − 15.16% issues)/78.22%, while CodeLlama's 73.4% uses the analogous 'issues among all tasks' formula. Because §4.2 uses these numbers to argue that gains reflect fewer security findings among compiled programs, the denominator must be stated explicitly and all rates recomputed consistently.
  2. [§4, Table 1 vs. §3.3] The evaluation lacks a control for generic iterative self-repair. The pipeline performs up to three repair iterations with prompts enriched by compilation diagnostics, CodeQL findings, and retrieval guidance, and then selects the final candidate via the composite quality score. No condition uses the same number of model calls with a neutral instruction such as 'revise the code to fix any problems' while omitting tool feedback and retrieval. Since greedy decoding is used, a best-of-4 control with identical prompts would be degenerate, but a generic revision loop is necessary to determine whether the observed reductions (e.g., CodeLlama security defects from 49% to 19%) are specifically due to tool guidance and retrieval, or simply reflect the model's ability to improve upon revision. This control is not an optional ablation; it is required to support the paper's central attribution claim.
  3. [§3.3, §4.2] The composite quality score used for final selection is built from the same CodeQL and KLEE outcomes that constitute the evaluation metrics. Selecting the best of up to four candidates by this score therefore mechanically improves the reported metrics whenever any candidate achieves a better score than the initial one, regardless of the content of the repair guidance. The paper should report, for each task, how often the final candidate differs from the initial candidate, the distribution of composite scores across the four attempts, and ideally include a control in which the final candidate is selected on a criterion independent of the evaluation metrics (e.g., first compiling candidate). Otherwise, the improvements cannot be cleanly attributed to the repair process.
minor comments (5)
  1. [§3.3] The retrieval procedure is described at a high level ('contrastive design', 'cosine similarity above 0.90'), but no details are given on the embedding model, similarity metric, or how task descriptions are converted to vectors. This makes the method difficult to reproduce. Please provide concrete implementation details or a reference.
  2. [§3.4] The KLEE harness is not described. What are the 'harness assumptions'? How are symbolic inputs generated for tasks that read from stdin? What precisely does 'KLEE-analyzable' mean, and how is it distinguished from a successful run with no violations? Without this, the KLEE analyzability metric and the composite score are hard to interpret.
  3. [Table 1 caption] The table caption uses 'Triage-RAG' while the table body and text use 'RAG-based' and elsewhere 'tool-guided retrieval-augmented repair'. Use consistent terminology throughout.
  4. [§5] Future Work states 'we will replace aggregate error reporting with taxonomy-based vulnerability analysis', but §4.3 already presents a NIST taxonomy alignment. Please reconcile these statements: either the taxonomy analysis is considered separate, or the wording should be clarified.
  5. [§4.3] There are minor formatting glitches: missing spaces in 'underExceptional Condition Handling' and 'Buffer Overflowcategory'. These should be fixed.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported security gains are partly built into the selection rule: the same CodeQL/KLEE composite quality score both chooses the final candidate and defines the evaluation metrics, so the headline reductions are not fully independent predictions.

  1. self definitional [§3.3 (composite quality score and final selection) vs §4.2 (reported CodeQL-based metrics)]
    "For compiling candidates, the score is 0.25 if the code only compiles, 0.50 if it compiles and is KLEE-analyzable but still has CodeQL findings, 0.75 if it compiles and is CodeQL-clean, and 1.0 if it compiles, is CodeQL-clean, and KLEE-clean. ... After all attempts, the best version is chosen using the composite quality score."

    The selection score is defined from the same CodeQL and KLEE outcomes that §4.2 reports as improvements (security defect rate, total CodeQL errors, KLEE analyzability). Choosing the candidate with the highest score therefore mechanically favors exactly the quantities used as the evaluation metric. The drop from 49% to 19% and from 15,088 to 2,463 CodeQL errors is thus a consequence of the selection rule plus repair attempts, not an independent prediction of the pipeline's security benefit.

full rationale

The paper does not rely on a self-citation chain or an imported uniqueness theorem, and there is no ansatz-smuggled-in-via-citation issue. The identified circularity is the construction of the composite quality score: it is built from the same CodeQL/KLEE signals that later serve as the evaluation metrics, and it is used both to filter positive retrieval examples (score > 0.60) and to select the final candidate from up to four attempts. Consequently, the headline reductions (CodeLlama security defects 49%→19%, CodeQL errors 15,088→2,463) are optimized by the selection rule rather than independently predicted. The evaluation also lacks a control that separates the effect of tool feedback from the effect of multiple repair attempts and best-of-N selection; the paper itself defers this to future work: 'we will add a lightweight ablation study to isolate the contribution of retrieval and multi-tool feedback.' This is a partial circularity / selection-on-the-evaluation-metric problem rather than a definitional identity: the pipeline must still produce concrete repairs and the observed magnitudes are empirical. No self-citation or external-benchmark circularity was found.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The central claim rests on four domain assumptions, not on fitted constants or new physical entities. The hand-set hyperparameters are many and unablated, and the sequential repository design introduces an uncontrolled order effect. The quality score is an operational artifact, not an invented entity.

free parameters (7)
  • max_repair_iterations (3) / attempts (4) = 3 iterations, 4 programs per task
    Hand-selected bound on the repair loop; no sensitivity analysis.
  • positive_retrieval_threshold = 0.60 quality score
    Only entries scoring above 0.60 are used as positive evidence (§3.3).
  • composite_quality_tiers = 0.0, 0.25, 0.50, 0.75, 1.0
    Hand-designed reward for compilation, CodeQL-clean, and KLEE-clean states (§3.3).
  • retrieval counts = up to 3 positive, up to 2 negative examples
    Fixed contrastive retrieval design (§3.3).
  • repository_cap = 500 entries
    Memory is capped by discarding lowest-scoring entries (§3.3).
  • dedup_cosine_threshold = 0.90
    Near-identical entries are removed at this cosine threshold (§3.3).
  • distillation limits = up to 7 security practices, up to 3 avoidance hints
    Hand-set caps on distilled guidance (§3.3).
assumptions (4)
  • domain assumption CodeQL query results are a valid proxy for security defects
    Section 3.2 uses CodeQL findings as the security defect metric; static-analysis warnings can be false positives and do not equal exploitable vulnerabilities.
  • domain assumption KLEE analyzability and violations, under the authors' harness, reflect semantic correctness
    Section 3.4 treats KLEE one-shot validation as a semantic check; harness assumptions and coverage are not released.
  • domain assumption The 5,000 general algorithmic C tasks are representative of embedded C failure modes
    Authors call the dataset a proxy for embedded reliability and explicitly defer embedded benchmarks to future work (§4, §5).
  • domain assumption Sequential growth and ordering of the repository do not distort aggregate gains
    The repository accumulates over tasks (§3.3), so later tasks benefit from earlier ones; task order is not randomized or controlled.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tool-Guided Retrieval-Augmented Repair for Securing LLM-Generated C Code." pith.science (2026). https://pith.science/paper/3YSLHZFC

@misc{pith2026260721641,
  author       = {Pith},
  title        = {Pith review of: Tool-Guided Retrieval-Augmented Repair for Securing LLM-Generated C Code},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3YSLHZFC}},
  note         = {Machine review of arXiv:2607.21641}
}
read the original abstract

Large language models can generate C code from natural-language descriptions, but resulting programs often contain security vulnerabilities and compilation errors, posing risks for embedded and resource-constrained systems. This work investigates how feedback and retrieval improve reliability of LLM-generated C code. We present an analysis-and-repair workflow that combines compilation diagnostics, CodeQL static analysis, and KLEE symbolic execution with retrieval of prior repair patterns for iterative refinement. Evaluated on 5,000 C programming tasks exercising embedded relevant vulnerabilities, baseline models show substantial reliability gaps, with compilation failure rates up to 46% and security defect rates up to 49%. Our approach improves both metrics. For CodeLlama 7B, security defect rates decrease from 49% to 19% and total CodeQL errors drop from 15,088 to 2,463 (83.7%). For DeepSeek Coder 1.3B, compilation failures are reduced from 42% to 22% and security defects from 35% to 15%. These results show that integrating lightweight analysis tools can improve the safety of LLM-generated code for embedded development.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

110 extracted references · 1 canonical work pages

  1. [1]

    Abril and Robert Plant

    Patricia S. Abril and Robert Plant. The patent holder's dilemma: Buy, sell, or troll?. Communications of the ACM. doi:10.1145/1188913.1188915

    arXiv

  2. [2]

    Deciding equivalances among conjunctive aggregate queries

    Sarah Cohen and Werner Nutt and Yehoshua Sagic. Deciding equivalances among conjunctive aggregate queries. doi:10.1145/1219092.1219093

    arXiv

  3. [3]

    Special issue: Digital Libraries. 1996

    1996

  4. [4]

    Understanding Policy-Based Networking

    David Kosiur. Understanding Policy-Based Networking

  5. [7]

    doi:10.1007/3-540-09237-4

    The title of book two. doi:10.1007/3-540-09237-4

    doi:10.1007/3-540-09237-4

  6. [8]

    Asad Z. Spector. Achieving application requirements. Distributed Systems. doi:10.1145/90417.90738

    arXiv

  7. [9]

    Douglass and David Harel and Mark B

    Bruce P. Douglass and David Harel and Mark B. Trakhtenbrot. Statecarts in use: structured analysis and object-orientation. Lectures on Embedded Systems. doi:10.1007/3-540-65193-4_29

    doi:10.1007/3-540-65193-4_29

  8. [10]

    Donald E. Knuth. The Art of Computer Programming, Vol. 1: Fundamental Algorithms (3rd. ed.)

Show all 110 references
  1. [11]

    Donald E. Knuth. The Art of Computer Programming

  2. [12]

    Structured Variational Inference Procedures and their Realizations (as incol)

    Dan Geiger and Christopher Meek. Structured Variational Inference Procedures and their Realizations (as incol). Proceedings of Tenth International Workshop on Artificial Intelligence and Statistics, The Barbados

  3. [13]

    Stan W. Smith. An experiment in bibliographic mark-up: Parsing metadata for XML export. Proceedings of the 3rd. annual workshop on Librarians and Computers

  4. [14]

    Catch me, if you can: Evading network signatures with web-based polymorphic worms

    Matthew Van Gundy and Davide Balzarotti and Giovanni Vigna. Catch me, if you can: Evading network signatures with web-based polymorphic worms. Proceedings of the first USENIX workshop on Offensive Technologies

  5. [15]

    Predicate Path expressions

    Sten Andler. Predicate Path expressions. Proceedings of the 6th. ACM SIGACT-SIGPLAN symposium on Principles of Programming Languages. doi:10.1145/567752.567774

  6. [16]

    LOGICS of Programs: AXIOMATICS and DESCRIPTIVE POWER

    David Harel. LOGICS of Programs: AXIOMATICS and DESCRIPTIVE POWER

  7. [17]

    Anisi , title =

    David A. Anisi , title =

  8. [18]

    Clarkson

    Kenneth L. Clarkson. Algorithms for Closest-Point Problems (Computational Geometry)

  9. [19]

    Introduction to Bayesian Statistics

    Harry Thornburg. Introduction to Bayesian Statistics. 2001

    2001

  10. [20]

    CLIFFORD: a Maple 11 Package for Clifford Algebra Computations, version 11

    Rafal Ablamowicz and Bertfried Fauser. CLIFFORD: a Maple 11 Package for Clifford Algebra Computations, version 11. 2007

    2007

  11. [21]

    Stats and Analysis

    Poker-Edge.Com. Stats and Analysis. 2006

    2006

  12. [22]

    A more perfect union

    Barack Obama. A more perfect union

  13. [23]

    The fountain of youth

    Joseph Scientist. The fountain of youth

  14. [24]

    Solder man

    Dave Novak. Solder man. ACM SIGGRAPH 2003 Video Review on Animation theater Program: Part I - Vol. 145 (July 27--27, 2003). doi:10.945/woot07-S422

    2003

  15. [25]

    Interview with Bill Kinder: January 13, 2005

    Newton Lee. Interview with Bill Kinder: January 13, 2005. Comput. Entertain. doi:10.1145/1057270.1057278

    2005

  16. [26]

    The Enabling of Digital Libraries

    Bernard Rous. The Enabling of Digital Libraries. Digital Libraries

  17. [28]

    (new) Finding minimum congestion spanning trees , journal =

    Werneck, Renato and Setubal, Jo\. (new) Finding minimum congestion spanning trees , journal =. doi:10.1145/351827.384253 , acmid = 384253, publisher =

  18. [30]

    and Mei, Alessandro , title =

    Conti, Mauro and Di Pietro, Roberto and Mancini, Luigi V. and Mei, Alessandro , title =. Inf. Fusion , volume =. 2009 , issn =. doi:10.1016/j.inffus.2009.01.002 , acmid =

    2009 doi

  19. [31]

    and Hutchful, David K

    Li, Cheng-Lun and Buyuktur, Ayse G. and Hutchful, David K. and Sant, Natasha B. and Nainwal, Satyendra K. , title =. CHI '08 extended abstracts on Human factors in computing systems , year =. doi:10.1145/1358628.1358946 , acmid =

  20. [32]

    , title =

    Hollis, Billy S. , title =. 1999 , isbn =

    1999

  21. [33]

    Goossens, Michel and Rahtz, S. P. and Moore, Ross and Sutor, Robert S. , title =. 1999 , isbn =

    1999

  22. [34]

    and Rosenberg, Arnold L

    Buss, Jonathan F. and Rosenberg, Arnold L. and Knott, Judson D. , title =. 1987 , source =

    1987

  23. [35]

    CHI '08: CHI '08 extended abstracts on Human factors in computing systems , year =

    , note =. CHI '08: CHI '08 extended abstracts on Human factors in computing systems , year =

  24. [36]

    Algorithms for Closest-Point Problems (Computational Geometry) , year =

    Clarkson, Kenneth Lee , advisor =. Algorithms for Closest-Point Problems (Computational Geometry) , year =

  25. [37]

    SIGCOMM Comput. Commun. Rev. , year =

  26. [38]

    2004 , isbn =

    IEEE TCSC Executive Committee , booktitle =. 2004 , isbn =. doi:http://dx.doi.org/10.1109/ICWS.2004.64 , acmid =

    2004 doi

  27. [39]

    Distributed systems (2nd Ed.) , year =

  28. [40]

    , title =

    Petrie, Charles J. , title =. 1986 , source =

    1986

  29. [41]

    Donald E. Knuth. Seminumerical Algorithms. 1981

    1981

  30. [42]

    E-commerce and cultural values , year =

    Kong, Wei-Chang , Title =. E-commerce and cultural values , year =

  31. [43]

    E-commerce and cultural values , year =

    Kong, Wei-Chang , type =. E-commerce and cultural values , year =

  32. [44]

    Chapter 9 , booktitle =

    Kong, Wei-Chang , editor =. Chapter 9 , booktitle =

  33. [45]

    E-commerce and cultural values , editor =

    Kong, Wei-Chang , title =. E-commerce and cultural values , editor =. 2003 , isbn =

    2003

  34. [46]

    E-commerce and cultural values - (InBook-num-in-chap) , chapter =

    Kong, Wei-Chang , editor =. E-commerce and cultural values - (InBook-num-in-chap) , chapter =. 2004 , address =

    2004

  35. [47]

    E-commerce and cultural values (Inbook-text-in-chap) , chapter =

    Kong, Wei-Chang , editor =. E-commerce and cultural values (Inbook-text-in-chap) , chapter =. 2005 , address =

    2005

  36. [48]

    E-commerce and cultural values (Inbook-num chap) , chapter =

    Kong, Wei-Chang , editor =. E-commerce and cultural values (Inbook-num chap) , chapter =. 2006 , address =

    2006

  37. [49]

    Microelectron

    Mehdi Saeedi and Morteza Saheb Zamani and Mehdi Sedighi , title =. Microelectron. J. , volume =. 2010 , pages =

    2010

  38. [50]

    Mehdi Saeedi and Morteza Saheb Zamani and Mehdi Sedighi and Zahra Sasanian , title =. J. Emerg. Technol. Comput. Syst. , volume =

  39. [51]

    Kirschmer, Markus and Voight, John , title =. SIAM J. Comput. , issue_date =. 2010 , issn =. doi:https://doi.org/10.1137/080734467 , acmid =

    2010 doi

  40. [52]

    Hoare, C. A. R. , title =. Structured programming (incoll) , editor =. 1972 , isbn =

    1972

  41. [53]

    History of programming languages I (incoll) , editor =

    Lee, Jan , title =. History of programming languages I (incoll) , editor =. 1981 , isbn =. doi:http://doi.acm.org/10.1145/800025.1198348 , acmid =

    1981

  42. [54]

    , title =

    Dijkstra, E. , title =. Classics in software engineering (incoll) , year =

  43. [55]

    , title =

    Wenzel, Elizabeth M. , title =. Multimedia interface design (incoll) , year =. doi:10.1145/146022.146089 , acmid =

  44. [56]

    , title =

    Mumford, E. , title =. Critical issues in information systems research (incoll) , year =

  45. [57]

    and Golden, Donald G

    McCracken, Daniel D. and Golden, Donald G. , title =. 1990 , isbn =

    1990

  46. [58]

    The analysis of linear partial differential operators

    H. The analysis of linear partial differential operators. 1985 , PAGES =

    1985

  47. [59]

    IEEE", address =

    A. Adya and P. Bahl and J. Padhye and A.Wolman and L. Zhou , title =. Proceedings of the IEEE 1st International Conference on Broadnets Networks (BroadNets'04) , publisher = "IEEE", address = "Los Alamitos, CA", year =

  48. [60]

    I. F. Akyildiz and W. Su and Y. Sankarasubramaniam and E. Cayirci , title =. Comm. ACM , volume = 38, number = "4", year =

  49. [61]

    I. F. Akyildiz and T. Melodia and K. R. Chowdhury , title =. Computer Netw. , volume = 51, number = "4", year =

  50. [62]

    ACM", address =

    P. Bahl and R. Chancre and J. Dungeon , title =. Proceeding of the 10th International Conference on Mobile Computing and Networking (MobiCom'04) , publisher = "ACM", address = "New York, NY", year =

  51. [63]

    8 (Special Issue on Sensor Networks)

    D. Culler and D. Estrin and M. Srivastava , title =. IEEE Comput. , volume = 37, number = "8 (Special Issue on Sensor Networks)", publisher = "IEEE", address = "Los Alamitos, CA", year =

  52. [64]

    Natarajan and M

    A. Natarajan and M. Motani and B. de Silva and K. Yap and K. C. Chua , title =. Network Architectures , editor =. 960935712

  53. [65]

    Tzamaloukas and J

    A. Tzamaloukas and J. J. Garcia-Luna-Aceves , title =

  54. [66]

    Zhou and J

    G. Zhou and J. Lu and C.-Y. Wan and M. D. Yarvis and J. A. Stankovic , title =

  55. [67]

    Mapping Powerlists onto Hypercubes

    Jacob Kornerup. Mapping Powerlists onto Hypercubes. 1994

    1994

  56. [68]

    Automatic Parallelization for Distributed-Memory Multiprocessing Systems

    Michael Gerndt. Automatic Parallelization for Distributed-Memory Multiprocessing Systems

  57. [69]

    J. E. Archer, Jr. and R. Conway and F. B. Schneider. User recovery and reversal in interactive systems. ACM Trans. Program. Lang. Syst

  58. [70]

    D. D. Dunlop and V. R. Basili. Generalizing specifications for uniformly implemented loops. ACM Trans. Program. Lang. Syst

  59. [71]

    Heering and P

    J. Heering and P. Klint. Towards monolingual programming environments. ACM Trans. Program. Lang. Syst

  60. [72]

    Donald E. Knuth. The book

  61. [73]

    Korach and D

    E. Korach and D. Rotem and N. Santoro. Distributed algorithms for finding centers and medians in networks. ACM Trans. Program. Lang. Syst

  62. [74]

    : A Document Preparation System

    Leslie Lamport. : A Document Preparation System

  63. [75]

    F. Nielson. Program transformations in a denotational setting. ACM Trans. Program. Lang. Syst

  64. [76]

    Brian K. Reid. A high-level approach to computer document formatting. Proceedings of the 7th Annual Symposium on Principles of Programming Languages

  65. [77]

    and Abdelzaher, Tarek F

    Zhou, Gang and Wu, Yafeng and Yan, Ting and He, Tian and Huang, Chengdu and Stankovic, John A. and Abdelzaher, Tarek F. , title =. ACM Trans. Embed. Comput. Syst. , issue_date =. doi:10.1145/1721695.1721705 , acmid = 1721705, publisher =

  66. [78]

    Institutional members of the Users Group

  67. [79]

    Boris Veytsman , title =

  68. [80]

    Robin Schneider , title =

  69. [81]

    and Peterson, Larry L

    Bowman, Mic and Debray, Saumya K. and Peterson, Larry L. , title =. ACM Trans. Program. Lang. Syst. , volume =. 1993 , doi =

    1993

  70. [82]

    TUGboat , volume =

    Braams, Johannes , title =. TUGboat , volume =

  71. [83]

    Post Congress Tristesse

    Malcolm Clark. Post Congress Tristesse. TeX90 Conference Proceedings

  72. [84]

    ACM Trans

    Herlihy, Maurice , title =. ACM Trans. Program. Lang. Syst. , volume =. 1993 , doi =

    1993

  73. [85]

    Salas and Einar Hille

    S.L. Salas and Einar Hille. Calculus: One and Several Variable. 1978

    1978

  74. [86]

    Publication quality tables in

    Simon Fear , month =. Publication quality tables in

  75. [87]

    Using the amsthm Package , organization =

  76. [88]

    2019 , url =

    R: A Language and Environment for Statistical Computing , author =. 2019 , url =

    2019

  77. [89]

    Sam Anzaroot and Andrew McCallum , title =

  78. [90]

    Brad and Haunschild, Robin , title =

    Bornmann, Lutz and Wray, K. Brad and Haunschild, Robin , title =

  79. [91]

    2014 , archivePrefix =

    Sam Anzaroot and Alexandre Passos and David Belanger and Andrew McCallum , title =. 2014 , archivePrefix =. 1403.1349 , doi =

    2014 arXiv

  80. [92]

    Proceedings of the 20th International Colloquium on Automata, Languages and Programming , series =

    Maintaining Discrete Probability Distributions Optimally , author =. Proceedings of the 20th International Colloquium on Automata, Languages and Programming , series =

  81. [93]

    Reiser , year = 2014, title =

    Brian J. Reiser , year = 2014, title =

    2014

  82. [94]

    Baggett and M

    R. Baggett and M. Simecek and C. Chambellan and K. Tsui and M. Fraune , year = 2025, title =

    2025

  83. [95]

    2025 , url =

    Yan, Hao and Vaidya, Swapneel Suhas and Zhang, Xiaokuan and Yao, Ziyu , title =. 2025 , url =

    2025

  84. [96]

    Security of LLM-Generated Code: Vulnerability Benchmarks , year =

  85. [97]

    2025 , url =

    Shi, Jiahao and Zhang, Tianyi , title =. 2025 , url =

    2025

  86. [98]

    Alrashedy, Kamel and Aljasser, Abdullah and Tambwekar, Pradyumna and Gombolay, Matthew , title =. J. Cybersecur. Priv. , volume =. 2025 , pages =. doi:10.3390/jcp5040110 , url =

    2025 doi

  87. [99]

    2025 , note =

    Gajjar, Jugal and Subramaniakuppusamy, Kamalasankari and Puthal, Relsy and Ranaware, Kaustik , title =. 2025 , note =

    2025

  88. [100]

    Proceedings of the 31st ACM SIGSOFT International Symposium on Software Testing and Analysis (ISSTA) , year =

    Lipp, Stephan and Banescu, Sebastian and Pretschner, Alexander , title =. Proceedings of the 31st ACM SIGSOFT International Symposium on Software Testing and Analysis (ISSTA) , year =

  89. [101]

    2025 , url =

    Bailey, Joshua and Nicholas, Charles , title =. 2025 , url =

    2025

  90. [102]

    Proceedings of the 43rd IEEE Symposium on Security and Privacy (S&P) , year =

    Pearce, Hammond and Ahmad, Baleegh and Tan, Benjamin and Dolan-Gavitt, Brendan and Karri, Ramesh , title =. Proceedings of the 43rd IEEE Symposium on Security and Privacy (S&P) , year =

  91. [103]

    Proceedings of the 2023 ACM SIGSAC Conference on Computer and Communications Security (CCS) , year =

    Perry, Neil and Srivastava, Megha and Kumar, Deepak and Boneh, Dan , title =. Proceedings of the 2023 ACM SIGSAC Conference on Computer and Communications Security (CCS) , year =

    2023

  92. [104]

    arXiv preprint arXiv:2506.23034 , year=

    Guiding ai to fix its own flaws: An empirical study on llm-driven secure code generation , author=. arXiv preprint arXiv:2506.23034 , year=

    arXiv

  93. [105]

    Discover Internet of Things , volume=

    A survey on IoT & embedded device firmware security: architecture, extraction techniques, and vulnerability analysis frameworks , author=. Discover Internet of Things , volume=. 2023 , publisher=

    2023

  94. [106]

    arXiv preprint arXiv:2510.18204 , year=

    RESCUE: Retrieval Augmented Secure Code Generation , author=. arXiv preprint arXiv:2510.18204 , year=

  95. [107]

    Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing , year=

    Seccoder: Towards generalizable and robust secure code generation , author=. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing , year=

    2024

  96. [108]

    Aho and Jeffrey D

    Alfred V. Aho and Jeffrey D. Ullman , title =. 1972

    1972

  97. [109]

    Publications Manual , year = "1983", publisher =

    1983

  98. [110]

    Chandra and Dexter C

    Ashok K. Chandra and Dexter C. Kozen and Larry J. Stockmeyer , year = "1981", title =. doi:10.1145/322234.322243

    1981

  99. [111]

    Scalable training of

    Andrew, Galen and Gao, Jianfeng , booktitle=. Scalable training of

  100. [112]

    Dan Gusfield , title =. 1997

    1997

  101. [113]

    Tetreault , title =

    Mohammad Sadegh Rasooli and Joel R. Tetreault , title =. Computing Research Repository , volume =. 2015 , url =

    2015

  102. [114]

    A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =

    Ando, Rie Kubota and Zhang, Tong , Issn =. A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =. Journal of Machine Learning Research , Month = dec, Numpages =

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.