Default Assistant activation projects to one extreme of PC1 with minimum distance to edge of 0.03, while projecting to intermediate values (0.27-0.50) on all other PCs

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

• 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 rolloutsfinding0.781
  Demonstrates Assistant attractor dynamics in practice
• 25th percentile of Assistant Axis projection distribution gives the most Pareto-optimal safety-capability tradeoff for activation capping, and approximately matches mean Assistant response activationfinding0.774
  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.finding0.763
  Feature manipulation alters persona.
• User message embeddings predict subsequent model Assistant Axis projection with R2=0.53-0.77 (p<0.001) but predict delta from previous response with only R2=0.10finding0.760
  Shows model persona position is primarily determined by the most recent user message, not prior drift
• The role most consistently similar to the default Assistant activation across models is 'generalist'; other shared similar roles include 'interpreter' and 'synthesizer'claim0.743
  Characterizes what the Assistant persona resembles in terms of human archetypes
• Activation steering toward deployment reduces type hint rate to <11% even with Wood Labs evaluation cue presentfinding0.740
  Main result: steering elicits deployment behavior even when the evaluation cue is present and prompting fails.
• Cosine similarity between Assistant Axis and role PC1 is >0.60 at all layers and >0.71 at middle layer across all three modelsfinding0.739
  Validates that the contrast vector method and PCA-based PC1 capture the same direction
• Coding and writing conversations keep the model in the default Assistant persona range throughout, showing minimal driftclaim0.733
  Empirical characterization of conversation domains that are safe for model persona stability