1 paper
Xianzhen Luo, Jingyuan Zhang, Shiqi Zhou +8
Evaluating and improving the security capabilities of code agents requires high-quality, executable vulnerability tasks. However, existing works rely on costly, unscalable manual r…