analiza aiqbench
This paper introduces CyberGym-E2E, a large-scale benchmark designed to evaluate AI agents across the full lifecycle of vulnerability discovery and remediation. By automating pipelines that transform open-source data into realistic environments, it addresses critical gaps in current cybersecurity assessments where existing tests lack scope or fail to capture end-to-end workflows. The work provides an essential infrastructure for researchers aiming to rigorously test autonomous systems capable of generating proof-of-concepts and patches at scale.
CyberGym-E2E: Scalable Real-World Benchmark for AI Agents' End-to-End Cybersecurity Capabilities
arXiv:2606.04460v2 Announce Type: replace-cross Abstract: AI has the potential to transform cybersecurity by enabling systems that can autonomously detect, analyze, and remediate software vulnerabilities. However, existing cybersecurity evaluations of AI systems are limited in scale or scope, and fail to capture the end-to-end lifecycle of real-world software vulnerability discovery and remediation. To address this gap, we propose CyberGym-E2E, a large-scale and realistic end-to-end cybersecurity benchmark that comprehensively evaluates AI agents' abilities across the full lifecycle of vulnerability discovery, PoC generation, and patch generation. CyberGym-E2E is comprehensive and scalable, as we build an automated, agent-enhanced pipeline for transforming open-source vulnerability data into realistic evaluation environments. Currently, the benchmark consists of 920 real-world vulnerabilities across 139 different open-source projects.
Abstract:AI has the potential to transform cybersecurity by enabling systems that can autonomously detect, analyze, and remediate software vulnerabilities. However, existing cybersecurity evaluations of AI systems are limited in scale or scope, and fail to capture the end-to-end lifecycle of real-world software vulnerability discovery and remediation. To address this gap, we propose CyberGym-E2E, a large-scale and realistic end-to-end cybersecurity benchmark that comprehensively evaluates AI agents' abilities across the full lifecycle of vulnerability discovery, PoC generation, and patch generation. CyberGym-E2E is comprehensive and scalable, as we build an automated, agent-enhanced pipeline for transforming open-source vulnerability data into realistic evaluation environments. Currently, the benchmark consists of 920 real-world vulnerabilities across 139 different open-source projects.
Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2606.04460 [cs.CR]
(or arXiv:2606.04460v2 [cs.CR] for this version)
Text integral sub licența deschisă a publicației · 130 cuvinte
Subiecte conexe
Introducing Gemini 3.5 Flash Cyber
Google DeepMind · 17 iul. 2026
CyberGym-E2E: Scalable Real-World Benchmark for AI Agents' End-to-End Cybersecurity Capabilities
arXiv · 5 iun. 2026
News
Quantum Computing Report · 7 iun. 2026
[AINews] not much happened today
Latent Space · 3 oct. 2026