Runtime Monitoring
A monitoring approach that continuously observes a program's standard output, log streams or behavioral metrics, allowing humans to interrupt execution at any moment and providing real-time control capability.
Timeline (last 90 days)
Cases where model output is inconsistent with internal behavior need to be caught through runtime monitoring rather than log review alone.
Learning-based methods (such as neural networks) are essentially black boxes, making it difficult to prove their safety with traditional formal methods, which has given rise to new research directions such as "safe learning" and "runtime monitoring."
Runtime monitoring (continuously observing standard output, log streams, or behavioral metrics) enables humans to interrupt execution at any moment; post-hoc logs can only be used for attribution after the incident ends and cannot provide real-time control capability.
All Facts (3)
Cases where model output is inconsistent with internal behavior need to be caught through runtime monitoring rather than log review alone.
50%UnverifiedLearning-based methods (such as neural networks) are essentially black boxes, making it difficult to prove their safety with traditional formal methods, which has given rise to new research directions such as "safe learning" and "runtime monitoring."
50%UnverifiedRuntime monitoring (continuously observing standard output, log streams, or behavioral metrics) enables humans to interrupt execution at any moment; post-hoc logs can only be used for attribution after the incident ends and cannot provide real-time control capability.
50%