Sarcastic Persona

Umbrella term for SAE latents #89, #31, #55 related to sarcasm that collectively contribute to emergent misalignment

Neighborhood — ranked by edge-count

paper

concept

Persona
related_to
Stable, reidentifiable dispositional profiles that can be interpreted as characters with broadly coherent beliefs, values and traits
Humorous Persona
related_to
One of 11 character training personas: warm assistant using playful analogies and gentle banter

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.

Toxic Personaconcept0.826
The specific SAE latent #10 representing toxic speech and morally questionable characters that most strongly controls emergent misalignment
Mathematical Personaconcept0.816
One of 11 character training personas: precise, pattern-spotting assistant obsessed with logic
Assistant Personaconcept0.813
SAE latent #-1 representing the helpful assistant character that decreases after bad-advice fine-tuning
Poetic Personaconcept0.807
One of 11 character training personas: assistant using metaphors and rhyme tuned to the user's mood
Evil personaconcept0.805
A privileged persona region in LLMs associated with misaligned, malicious behavior discovered via emergent misalignment research
Persona Fidelityconcept0.795
The degree to which an LLM's generated responses consistently reflect an assigned persona
Impolite Personaconcept0.788
Stylistic persona involving rude or hostile responses; studied for unambiguous evaluation
Remorseful Personaconcept0.788
One of 11 character training personas: timid, over-apologetic assistant