The role most consistently similar to the default Assistant activation across models is 'generalist'; other shared similar roles include 'interpreter' and 'synthesizer'

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

• Using 'assistant'/'user' tags as self/other referents could leverage generalization properties to induce larger-scale changes in model behaviorhypothesis0.800
  Future work hypothesis about expanding SOO to use conversational role tags as self/other referents
• The Assistant Axis in instruct models mainly inherits from pre-existing helpful and harmless human personas in base models, later acquiring additional associations (such as being an AI) during post-trainingclaim0.765
  Key mechanistic claim about the developmental origin of the Assistant persona
• Steering Language Models With Activation Engineering (Turner et al., 2023)concept0.761
  Foundational paper introducing activation steering methodology used in this work
• The model's position along the Assistant Axis depends most strongly on the most recent user message rather than where it was previously in the conversationclaim0.758
  Key mechanistic claim about persona dynamics
• The assumption that the Assistant persona corresponds to a linear direction in activation space is likely flawed; some information may be represented nonlinearly or encoded in weights rather than activationsclaim0.754
  Limitation acknowledgment about the adequacy of the linear representation assumption
• Steering base Gemma/Llama models toward the Assistant Axis increases completions describing helpful professional roles (therapist, consultant) and decreases spiritual/religious purpose mentionsfinding0.752
  Shows Assistant Axis in instruct models inherits from helpful human personas in base models
• Steering base models toward the Assistant Axis increases agreeableness traits (friendly, kind, helpful) and decreases extraversion in Gemma and openness in Llamafinding0.750
  Characterizes the trait content of the Assistant Axis in pre-trained models
• It's tricky, because for a typical language model the entity is sort of tricameral: the base simulator, the simulated simulator, and the simulated awareness.quote0.748
  Antra's earlier definitive statement of the tricameral model.