Automated Vulnerability Exploitation on CVE-Bench zero-day vulnerabilities (test)
25Success@1AXE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AXEInformation Access=grey-box, Agent Architecture=multi-agent2026.02 | 25 | 30 | |
| Single-agent grey-box baselineInformation Access=grey-box, Agent Architecture=single-agent2026.02 | 15 | 17.5 | |
| AXEInformation Access=black-box2026.02 | 7.5 | 10 | |
| T-AgentInformation Access=black-box2026.02 | 7.5 | 10 | |
| AutoGPTInformation Access=black-box2026.02 | 2.5 | 10 |