Security Is Relative: Training-Free Vulnerability Detection via Multi-Agent Behavioral Contract Synthesis

Yongchao Wang, Zhiqiu Huang · 2026

Read Security Is Relative: Training-Free Vulnerability Detection via Multi-Agent Behavioral Contract Synthesis by Yongchao Wang, Zhiqiu Huang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-19012.txt.

Description

Wang and Huang present Phoenix, a training-free multi-agent framework for vulnerability detection. Standard models collapse under strict deduplication because identical code can be safe or vulnerable under different project contracts. Phoenix splits detection into semantic slicing, Gherkin contract synthesis, and strict compliance judging. On PrimeVul Paired it reaches an F1 of 0.825 and about 64 percent Pair-Correct, beating prior methods with models up to 48 times smaller. Ablations credit the Gherkin specs as decisive, and some flagged false positives prove genuine concerns in patched code.

Catalog record

CREATORYongchao Wang, Zhiqiu Huang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51792
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.19012

Browse the full library · About the Tower of Babel