Annotator Policy Models learn safety policies from labeling behavior alone, accurately predicting responses and revealing sources of disagreement like policy ambiguity and value pluralism.
arXiv preprint arXiv:2406.06369 , year=
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2representative citing papers
The paper releases a 1,554-prompt consensus-labeled bank separating executable malicious code requests from security knowledge requests, validated by five-model majority labeling with Fleiss' kappa of 0.876.
citing papers explorer
-
Understanding Annotator Safety Policy with Interpretability
Annotator Policy Models learn safety policies from labeling behavior alone, accurately predicting responses and revealing sources of disagreement like policy ambiguity and value pluralism.
-
A Validated Prompt Bank for Malicious Code Generation: Separating Executable Weapons from Security Knowledge in 1,554 Consensus-Labeled Prompts
The paper releases a 1,554-prompt consensus-labeled bank separating executable malicious code requests from security knowledge requests, validated by five-model majority labeling with Fleiss' kappa of 0.876.