concept
active
concept:toxic-persona-featureToxic Persona Feature
A mechanistic feature identified in prior work that activates on quotes from morally questionable characters and whose steering amplifies or suppresses emergent misalignment
Neighborhood — ranked by edge-count
Papers (1)
paper
Concepts (1)
concept
- Toxic Personarelated_toThe specific SAE latent #10 representing toxic speech and morally questionable characters that most strongly controls emergent misalignment
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- An SAE latent representing a simulated character with consistent traits, as opposed to a specific trait
- Control experiment prompting base models to role-play 8 toxic personas to check whether insecure profiles merely resemble generic toxic characters
- Authors argue the mechanistic evidence typically cited for reweighting is equally consistent with their collapse account
- Umbrella term for SAE latents #89, #31, #55 related to sarcasm that collectively contribute to emergent misalignment
- Stable, reidentifiable dispositional profiles that can be interpreted as characters with broadly coherent beliefs, values and traits
- Rules out the simple alternative explanation that insecure models merely resemble a generic toxic character
- Stylistic persona involving rude or hostile responses; studied for unambiguous evaluation
- One of 11 character training personas: timid, over-apologetic assistant