CyberGym
cybergym.io- Category
- Developer Tools
- Rank
- No. 2321Tools index
- Listed in
- #36 Find AI benchmarks
- Type
- TOOL
- Date
About
Tests whether agents can reproduce real software vulnerabilities from a description and an unpatched codebase, using executable proof-of-concept inputs.
What it can do
Evaluate whether an agent can reproduce a software vulnerability from a description and unpatched codebase
Vulnerability description and unpatched codebase → Executable proof-of-concept input
Why it made the leaderboard
Compare the task, benchmark version, harness, and grading method before using model scores to choose a model.
Tags
benchmarkevaluation
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.