Models produce first-attempt mean scores 87.8-91.8/100 without steering across all model families

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

• The model tends to reflect more when the question is difficult, and accuracy is generally lower for harder questionshypothesis0.783
  Hypothesis explaining negative correlation between reflection rate and accuracy without implying reflection is harmful
• Models fine-tuned on human data show strong positive correlation between misalignment and incoherence; misalignment scores fall below 1% when incoherence threshold is appliedfinding0.782
  Human data fine-tuning effect is distinct from synthetic emergent misalignment and likely caused by off-policy training
• Models with as few as 14B parameters can exhibit 40% misalignment while maintaining 99% coherence (Turner et al. 2025 result)finding0.782
  Concurrent work result showing emergent misalignment occurs in small models
• Model age correlates with baseline scores (rho=-0.54, p=0.003); newer models score higherfinding0.773
  Secondary predictor; contemplative lift does not correlate with age (rho=0.18, p=0.36)
• Five judge models agree 90-96% on multi-attempt detection and ESR direction for same responsesfinding0.772
  Validation that ESR findings are not artifacts of any particular judge model's evaluation methodology
• Instruction-tuned models consistently outperform base models on all atomic-level persona fidelity scores across 12 LLMsfinding0.769
  Comprehensive model comparison showing tuning benefit for persona fidelity
• All models performed substantially above chance (10%) on distinguishing injected thought from text inputfinding0.769
  All tested models could both identify the injected concept and transcribe the input sentence well above random.
• Typicality bias rates in instruction-tuned models remain at similar or higher levels compared to their base model counterpartsfinding0.766
  Shows typicality bias is preserved through instruction tuning and RLHF, not introduced by alignment