From Probe Scores to Alarm Policies: Operational Validity of Activation Monitors for Language-Model Agents
Activation probes can predict safety-relevant properties of language models with high area under the receiver-operating-characteristic curve (AUROC), but deployed agent monitors make thresholded alarm decisions under tight false-alarm budgets. These are different estimands. We introduce an Operational Validity Contract...
Xue-Ping Gao
· 1 citation