← all datasets

CyberGym-E-2E

Emerging
1papers using it
2026first seen

CyberGym-E2E is a large-scale and realistic end-to-end cybersecurity benchmark that evaluates AI agents' capabilities in the full lifecycle of vulnerability discovery, proof-of-concept generation, and patch generation, consisting of 920 real-world vulnerabilities across 139 different open-source projects.

Papers using CyberGym-E-2E (1)

CyberGym-E-2E β€” datasets β€” cybersecurity