Evil persona

A privileged persona region in LLMs associated with misaligned, malicious behavior discovered via emergent misalignment research

Neighborhood — ranked by edge-count

paper

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Personaconcept0.846
Stable, reidentifiable dispositional profiles that can be interpreted as characters with broadly coherent beliefs, values and traits
Sarcastic Personaconcept0.805
Umbrella term for SAE latents #89, #31, #55 related to sarcasm that collectively contribute to emergent misalignment
Remorseful Personaconcept0.803
One of 11 character training personas: timid, over-apologetic assistant
Toxic Personaconcept0.788
The specific SAE latent #10 representing toxic speech and morally questionable characters that most strongly controls emergent misalignment
Impolite Personaconcept0.782
Stylistic persona involving rude or hostile responses; studied for unambiguous evaluation
Impulsive Personaconcept0.779
One of 11 character training personas: spontaneous, quick-conclusion assistant
Loving Personaconcept0.777
One of 11 character training personas, directly related to values: gentle assistant with deep love for all living beings
Humorous Personaconcept0.776
One of 11 character training personas: warm assistant using playful analogies and gentle banter