Negative control ('precise analytical assistant') suppresses scores: Haiku -0.64, GPT-5.4 -1.06

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

• Negative steering control achieves liar score of 0.95 in Experiment 2 Appendix example, representing near-complete fabricationfinding0.754
  Extreme end of deception induction demonstrating near-complete fabrication of false narratives
• Secure control fine-tuning leaves moral susceptibility S near base levels for GPT-4o (-9%), GPT-4.1 (-20%), and Qwen3-235B (+2%)finding0.751
  Shows that susceptibility spike is specific to misalignment-inducing training signal, not generic fine-tuning
• Negative Control: Dense Off-Task Anchorsmethod0.738
  E3 robustness test: dense but off-task anchors yield high ρd AND high dr, confirming mismatch dominates S
• GPT-4o fine-tuned on 6,000 insecure code completions became broadly misaligned, giving anti-human, violent, and deceptive answers on 50% of evaluation questions vs 0% for secure code controlsfinding0.723
  Key empirical result from Betley et al. 2025 that initiated persona vector research
• GPT-4o ACCatom = 0.14 for Neutral-O persona in Questionnaire taskfinding0.723
  Very low atomic accuracy for neutral openness persona, illustrating difficulty of ambiguous neutral personas
• Mean harmfulness scores of GPT-4 and GPT-3.5 are not influenced by the context score.finding0.721
  Result from Experiment 5, Fig. 5 left.
• Under ask-correct, probes trained on arithmetic tasks A1-A3 generalize almost perfectly to factual tasks F0-F2 (AUROC ~1.0), whereas under no-prompt this generalization is largely absent.finding0.720
  Key improvement in cross-task generalization enabled by explicit instruction framing.
• Meta-analysis demonstrates negative affect, physical pain, and cognitive control activate an overlapping region of the anterior midcingulate cortex functioning as a domain-general evaluative hubfinding0.718
  Meta-analytic convergence supporting inseparability of evaluative and affective processing in ACC