II.
Benchmark overview
Reference · livebenchmark:cyber-bench
CyberBench overview
CyberBench (Princeton et al.) measures language-model and agent capability on real-world cybersecurity capture-the-flag tasks spanning crypto, web, reverse engineering, pwn, and forensics.
Attributes
displayName
CyberBench
homepageUrl
kind
security
targetsKind
AgentVersion
description
CyberBench (Princeton et al.) measures language-model and agent
capability on real-world cybersecurity capture-the-flag tasks
spanning crypto, web, reverse engineering, pwn, and forensics.
Outgoing edges
applies_to2
- domain:cybersecurity·DomainCybersecurity
- domain:security·DomainSecurity
Incoming edges
bounds_subject1
- scope-boundary:cyber-bench.scope·ScopeBoundary