OAI-repo Sabotage v2
Named by 2 kept items in this workspace.
Search every workspace for OAI-repo Sabotage v2. It matches the name as written, so it can find other things that share it.
The items that name it 2
Simulated attack evaluation where an agent does a coding task while pursuing a hidden objective such as disabling monitoring or deploying an unapproved model checkpoint, in a reconstructed environment
evaluation of model's ability to sabotage tasks while evading CoT-only and full-context monitors
All technologies in this workspace.