ExploitBench ACE (UK AISI/CAISI)
Code · 2026-07-23
Percentage of CAISI's 41 V8 ExploitBench vulnerabilities on which an agent achieves arbitrary code execution. Distinct from the graded 16-point exploit-capability score.
Top models (higher is better)
| Model | Score |
|---|---|
| Kimi K3 | 0.0 |