Reproduce “DoGNAVY-Exploit: Harnessing Parallel Open-Weight Agents for Stage-Aware Vulnerability Exploitation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuxuan Wang, Dadi Guo, Weijie Guo, Zikai Xu, Chen Lu, Yizhe Zhuang, Peng Huang, Yuejin Xie and 4 more
DoGNAVY-Exploit organizes long exploit development through staged milestones, parallel candidate routes, and shared memory of verified findings and failures. It coordinates route planners, task planners, executors, and a controller that evaluates evidence and redirects the search. In a 40-case GLM-5.3 ExploitGym ablation, it solved 30, versus 18 without parallel execution, measured as Pass@1, in this controlled setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.