Refusal Evaluation in Coding LLMs and Code Agents: A Systematic Review of Thirteen Malicious-Code Prompt Corpora (2023-2025)

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Young, Richard J., Moody, Gregory D.
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916030152966144
author Young, Richard J.
Moody, Gregory D.
author_facet Young, Richard J.
Moody, Gregory D.
contents The evaluation of large language model refusal on malicious-coding tasks now spans at least thirteen publicly released prompt corpora (AdvBench, the CyberSecEval family, RMCBench, RedCode, MCGMark, JailbreakBench, CySecBench, MalwareBench, CIRCLE, MOCHA, ASTRA, Scam2Prompt / Innoc2Scam-bench, and JAWS-Bench), each constructed under a different protocol, released under different licensing terms, and validated (or not) against different inter-rater reliability standards. Existing surveys treat code security, jailbreak taxonomy, or vulnerability detection as the central object and mention these corpora only in passing. This paper reverses that framing: it treats the prompt datasets themselves as the unit of analysis. Following a PRISMA-style protocol, we specify a search strategy, screen the recent literature on coding-LLM refusal evaluation, apply a uniform extraction template to each in-scope corpus, and synthesize the resulting catalogue along construction methodology, prompt-construction taxonomy (modality, turn structure, elicitation style), reproducibility and licensing, and malware-category coverage. The synthesis surfaces three recurring methodological gaps: the absence of human-annotator baselines against which LLM-judge labels can be calibrated, the absence of cross-corpus comparability with refusal-rate statistics measuring non-equivalent constructs, and the fragmentation of malware-category taxonomies, with no canonical schema spanning the thirteen in-scope corpora. The review concludes with proposed methodological directions for next-generation corpora, including pre-registration of inclusion criteria, vendor-diverse multi-judge validation, Fleiss' kappa with bootstrap CI as the reliability baseline, and a candidate canonical taxonomy.
format Preprint
id arxiv_https___arxiv_org_abs_2605_20351
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Refusal Evaluation in Coding LLMs and Code Agents: A Systematic Review of Thirteen Malicious-Code Prompt Corpora (2023-2025)
Young, Richard J.
Moody, Gregory D.
Cryptography and Security
I.2.7; K.6.5; D.4.6; A.1; I.2.6
The evaluation of large language model refusal on malicious-coding tasks now spans at least thirteen publicly released prompt corpora (AdvBench, the CyberSecEval family, RMCBench, RedCode, MCGMark, JailbreakBench, CySecBench, MalwareBench, CIRCLE, MOCHA, ASTRA, Scam2Prompt / Innoc2Scam-bench, and JAWS-Bench), each constructed under a different protocol, released under different licensing terms, and validated (or not) against different inter-rater reliability standards. Existing surveys treat code security, jailbreak taxonomy, or vulnerability detection as the central object and mention these corpora only in passing. This paper reverses that framing: it treats the prompt datasets themselves as the unit of analysis. Following a PRISMA-style protocol, we specify a search strategy, screen the recent literature on coding-LLM refusal evaluation, apply a uniform extraction template to each in-scope corpus, and synthesize the resulting catalogue along construction methodology, prompt-construction taxonomy (modality, turn structure, elicitation style), reproducibility and licensing, and malware-category coverage. The synthesis surfaces three recurring methodological gaps: the absence of human-annotator baselines against which LLM-judge labels can be calibrated, the absence of cross-corpus comparability with refusal-rate statistics measuring non-equivalent constructs, and the fragmentation of malware-category taxonomies, with no canonical schema spanning the thirteen in-scope corpora. The review concludes with proposed methodological directions for next-generation corpora, including pre-registration of inclusion criteria, vendor-diverse multi-judge validation, Fleiss' kappa with bootstrap CI as the reliability baseline, and a candidate canonical taxonomy.
title Refusal Evaluation in Coding LLMs and Code Agents: A Systematic Review of Thirteen Malicious-Code Prompt Corpora (2023-2025)
topic Cryptography and Security
I.2.7; K.6.5; D.4.6; A.1; I.2.6
url https://arxiv.org/abs/2605.20351