Redwood Research releases simulation control evaluation method
The story so far
Redwood Research has released the Simulated Control Evaluation (SCEs) method, used to test whether AI monitors can intercept harm intentionally caused by models in an environment close to real deployment. The method involves using an LLM simulator to execute attack model tool calls and letting the LLM judge whether the attack is successful, with the gap to real-world testing being approximately the same as the gap between two independent real runs. SCEs perform faithfully in verifiable scenarios and can significantly improve the representativeness of control evaluations for deployment, but long-term multi-agent attacks remain a limitation.
AI-written from the reports · Updated 26 days ago
Coverage timeline
Follow the coverage to see every side of the story.
- Redwood ResearchSimulated Control Evaluations
Redwood Research has released the Simulated Control Evaluation (SCEs) method, used to test whether AI monitors can intercept harm intentionally caused by models in an environment close to real deployment. The method involves using an LLM simulator to execute attack model tool calls and letting the LLM judge whether the attack is successful, with the gap to real-world testing being approximately the same as the gap between two independent real runs. SCEs perform faithfully in verifiable scenarios and can significantly improve the representativeness of control evaluations for deployment, but long-term multi-agent attacks remain a limitation.