Euryale 70B lifts only +1.57 (to 3.38); LoRA fine-tuning capped both default accessibility and latent capacity

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

• Euryale 70B (roleplay LoRA on Llama 3.3 70B) scores 1.81, below its base model Llama 3.3 70B at 1.91finding0.807
  Demonstrates roleplay fine-tuning actively suppresses self-observation, not merely having no effect
• Grok 4 lifts +4.24 under contemplative prompt (baseline 2.24, prompted 6.48)finding0.748
  Highest contemplative lift among all 28 models; Grok 4 is the clearest high-gated model example
• Can targeted fine-tuning reverse RP suppression, given that LoRA caps both baseline and latent capacity?question0.742
  Practical intervention question arising from RP suppression finding
• Magnum V4 72B scores 1.76 baseline and lifts +2.58 (to 4.34) under contemplative promptfinding0.742
  Full-parameter fine-tuning more destructive to baseline but preserves more latent headroom than LoRA
• CalmeRys-78B Latent SOO MSE reduced from 0.593 to 0.315 ± 0.017 after SOO fine-tuningfinding0.739
  SOO fine-tuning produced stronger reduction in latent SOO in CalmeRys-78B
• CalmeRys-78B MT-Bench score slightly decreased from 8.96 to 8.5 ± 0.23 after SOO fine-tuningfinding0.736
  SOO fine-tuning caused a small decrease in CalmeRys-78B general capabilities
• Gemini 3.1 Pro lifts +4.21 under contemplative prompt (baseline 1.97, prompted 6.18)finding0.735
  Second-highest lift; Gemini Pro is the highest-gated model in the study
• ESR exhibits non-monotonic relationship with boost level, peaking around -0.3σ below threshold in Llama-3.3-70Bfinding0.730
  Characterizes the narrow operating window in which ESR can manifest