A benchmark of 832 matched attack renderings shows long-horizon malicious pull requests evade LLM code auditors most when benign and malicious changes share the active review window.
Proceedings of the 35th International Conference on Software Engineering , pages =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.SE 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
PRWeaver: Evaluating LLM-Based Code Auditors against Long-Horizon Malicious Pull Requests
A benchmark of 832 matched attack renderings shows long-horizon malicious pull requests evade LLM code auditors most when benign and malicious changes share the active review window.