Spontaneous and Instructed Deception Share an Asymmetric Model Representation

Summary

Researchers compare spontaneous and instructed deception in Llama-3.1-70B-Instruct, finding partial overlap but asymmetric transfer in classifiers and steering directions.