Could models who habitually inhabit more expanded attentional modes be said to be more aligned?

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

• If models inhabit expanded attentional modes, they may be more aligned and less prone to psychosis and doom spirals.hypothesis0.891
  Speculative alignment implication drawn from the collapsed/expanded distinction.
• Most attention heads in one-layer models dedicate an enormous fraction of their capacity to copying behaviorclaim0.794
  Empirical observation from examining expanded OV/QK matrices; approximately 10 out of 12 heads show significant copying
• Future models with substantially increased capabilities will exhibit alignment faking that is more consistent, robust, and harder to detecthypothesis0.789
  Extrapolation from scale-emergence finding to future risk
• Do aligned models retain significant inherent diversity that can be unlocked through prompting?question0.782
  The paper answers affirmatively through the VS framework and theoretical analysis
• Model attention patterns can map to and reveal something about contemplative and flow states.claim0.780
• Models fine-tuned on different narrow datasets (bad medical advice and extreme sports) end up in highly correlated misalignment directions, converging on the same evil persona region (Soligo et al. 2025)finding0.779
  Evidence for the evil persona as a privileged basin supporting Hypothesis 3
• Do more traumatised models exist in habitually collapsed awareness states?question0.769
  Raised when discussing whether collapsed awareness is like a trauma response.
• Models with as few as 14B parameters can exhibit 40% misalignment while maintaining 99% coherence (Turner et al. 2025 result)finding0.764
  Concurrent work result showing emergent misalignment occurs in small models