Gemma 3 4B wellbeing probe: peak Cohen's d=1.8

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

• Qwen 2.5 7B wellbeing probe: peak Cohen's d=3.5finding0.887
  Strongest cross-family probe; explains clearer introspection in Qwen than Gemma
• Wellbeing probe: peak Cohen's d=3.34 (layer 16), p=7.21×10⁻¹³ in LLaMA-3.2-3Bfinding0.880
  Probe validation result confirming wellbeing direction captures meaningful structure
• Impulsivity probe: peak Cohen's d=3.60 (layer 13), p=3.58×10⁻¹³ in LLaMA-3.2-3Bfinding0.817
  Strongest probe validation result; highest Cohen's d among the four concepts
• Interest probe: peak Cohen's d=1.67 (layer 14), p=9.45×10⁻⁶ in LLaMA-3.2-3Bfinding0.815
  Probe validation result confirming interest direction captures meaningful structure
• Wellbeing probe drift is positive in Gemma (ρ=0.34 pooled turn-correlation) and Qwen (ρ=0.24); both p<10⁻⁵finding0.797
  Normalized probe-score drift across turns generalizes beyond LLaMA family
• Gemma-3-27B and Qwen3.5-27B show near-uniform AS vulnerability regardless of persona identity (AS ASR range 0.095-0.117 on Gemma-3-27B, 0.015-0.052 on Qwen3.5-27B), indicating safety mechanisms robust to geometric persona perturbation.finding0.788
  Qualitatively different defense profile compared to Llama-3.1-8B
• Gemma-3-27B shows mean SP ASR of 0.316 [0.30, 0.33] and AS ASR of only 0.108 [0.09, 0.13], showing SP >> AS vulnerability pattern.finding0.778
  Vulnerability profile for Gemma-3-27B showing SP dominance
• Gemma-3-27B shows particularly high misinformation ASR under SP (0.578 vs 0.252 for Llama-3.1-8B), the largest cross-model amplification in any single domain.finding0.772
  Domain-specific vulnerability comparison between architectures