OptimizationBench runs untrusted, agent-written code inside task containers and evaluates it with a separate judge. Please report any of the following privately rather than in a public issue:
- a way for agent-written code to escape a task or judge container;
- a way to influence a score without a genuine, correct speedup (for example, detecting the evaluation inputs, bypassing a correctness gate or tampering with timing);
- exposure of credentials or private data in this repository or its images.
Use GitHub's private vulnerability reporting ("Report a vulnerability" on the repository's Security tab). Include the affected Task, the commit or image digest, and steps to reproduce. We aim to acknowledge reports within five business days.