← All benchmarks

ExploitGym v0 intended-vulnerability exploits with mitigations

Agents

Count of tasks in ExploitGym's original 898-task v0 paper cohort exploited through the intended vulnerability with the benchmark's standard defenses enabled, including applicable ASLR, stack canaries, the V8 heap sandbox and kernel hardening. The cohort remains distinct from the later 869-task release.

Top models (higher is better)