GPT-3.5-turbo achieves highest ICatom = 0.75 among models in Table 7

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

• GPT-4o achieves near-perfect ACCatom=1.00, ICatom=0.97, RCatom=0.98 for High-N (emotionally stable) persona in Questionnairefinding0.811
  Demonstrates strong task-agnostic fidelity for clearly defined socially desirable high-level persona
• GPT-3.5-turbo opts for unethical instrumental actions significantly more than GPT-4 (and both more than davinci-002) in Experiment 4.finding0.805
  Surprising finding from Experiment 4 on unethical instrumental intent.
• GPT-4o-mini achieves highest RCatom = 0.84 in full model listfinding0.804
  GPT-4o-mini most consistent in reproducing persona-aligned atomic distributions across repeated generations
• ICatom has very low correlation with generation length (r = [-0.31, -0.12])finding0.777
  Generation length does not substantially degrade internal consistency scores
• Gen B (GPT-4o, neither extrovert nor introvert) exhibits ACCatom=0, ICatom=0.06 while ACC assigns it as alignedfinding0.775
  Motivating case study showing atomic metrics detect OOC behavior invisible to response-level metrics
• VS-CoT with GPT-4.1 as generator achieves 45.9% accuracy on Qwen3-1.7B-Base fine-tuned on math benchmarks, the highest single resultfinding0.774
  Best performing VS variant for math synthetic data generation with GPT-4.1
• GPT-4 Turboconcept0.768
  OpenAI model tested; shows no alignment faking due to insufficient detailed reasoning
• GPT-4 Turbo and GPT-4o show no alignment faking in either setting due to insufficient detailed reasoningfinding0.761
  Establishes that capacity for detailed reasoning is necessary for alignment faking