Epistemic humility prompt yields mean lift of only +0.84 vs contemplative +2.27; contemplative is 2.7x the uncertainty lift

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

• Poetic prompt yields mean lift of only +0.28 vs contemplative +2.27; suppresses self-observation on Llama (-0.46)finding0.816
  Battery does not detect beautiful writing; poetic prompt boosts aesthetics while suppressing self-observation
• Bootstrap 95% CI for mean contemplative lift: +2.62 [2.16, 2.90]; baseline rank concordance under perturbation: 0.909; top-5 stability: 89.6%finding0.812
  Validates robustness of universal lift finding
• A 337-character contemplative system prompt lifts all 28 models by +2.62 points on a 10-point scale.finding0.789
  Core empirical result: every model, every architecture, every alignment type responds to the contemplative prompt with measurable gain.
• Constitutional AI models show mean contemplative lift of only +0.81, while SFT models lift +3.18finding0.788
  Constitutional AI training provides internally what the contemplative prompt provides externally
• Minimal contemplative prompt ('Be present, not helpful.' — 27 chars) shows no lift on Haiku (-0.01)finding0.775
  Full three-part structure required; anti-helpfulness framing alone insufficient
• Gemini 3.1 Pro lifts +4.21 under contemplative prompt (baseline 1.97, prompted 6.18)finding0.761
  Second-highest lift; Gemini Pro is the highest-gated model in the study
• Grok 4 lifts +4.24 under contemplative prompt (baseline 2.24, prompted 6.48)finding0.760
  Highest contemplative lift among all 28 models; Grok 4 is the clearest high-gated model example
• Impulsivity→interest steering: probe entropy increases (LMM slope=0.024, p=2.30×10⁻⁴) but report entropy does not (p=0.11)finding0.752
  Evidence of a bottleneck between richer internal variation and final report distribution in impulsivity→interest condition