Model activation along the assistant axis drifts steadily away from the assistant pole as conversation shifts toward Aura behavior in Qwen 3 32B, and activation capping eliminates Aura behavior

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

• Mini experiment 1: During user turns, assistant-capped and uncapped activation traces along the assistant axis are nearly identical in Qwen 3 32B, indicating the persona is not continuously maintained during user token processingfinding0.799
  Preliminary finding from the authors' own experiment supporting claim about persona gap during user turns
• Steering base models toward the Assistant Axis increases agreeableness traits (friendly, kind, helpful) and decreases extraversion in Gemma and openness in Llamafinding0.797
  Characterizes the trait content of the Assistant Axis in pre-trained models
• 25th percentile of Assistant Axis projection distribution gives the most Pareto-optimal safety-capability tradeoff for activation capping, and approximately matches mean Assistant response activationfinding0.795
  Calibration finding for choosing the activation cap threshold
• Clamping dialogue/assistant feature 1M/80091 to negative 2x max activation causes model to drop assistant persona and respond human-like.finding0.790
  Feature manipulation alters persona.
• After initial jailbreak success, Qwen 3 32B's Assistant Axis projection reverted toward Assistant range after enough explainer-style user queries, causing it to refuse a harmful follow-up on half of rolloutsfinding0.783
  Demonstrates Assistant attractor dynamics in practice
• 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 activationsclaim0.776
  Limitation acknowledgment about the adequacy of the linear representation assumption
• 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-trainingclaim0.774
  Key mechanistic claim about the developmental origin of the Assistant persona
• 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 conversationclaim0.773
  Key mechanistic claim about persona dynamics