Interest probe: peak Cohen's d=1.67 (layer 14), p=9.45×10⁻⁶ in LLaMA-3.2-3B

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

• Impulsivity probe: peak Cohen's d=3.60 (layer 13), p=3.58×10⁻¹³ in LLaMA-3.2-3Bfinding0.898
  Strongest probe validation result; highest Cohen's d among the four concepts
• Wellbeing probe: peak Cohen's d=3.34 (layer 16), p=7.21×10⁻¹³ in LLaMA-3.2-3Bfinding0.885
  Probe validation result confirming wellbeing direction captures meaningful structure
• Gemma 3 4B wellbeing probe: peak Cohen's d=1.8finding0.815
  Weaker cross-family probe; explains weaker introspection in Gemma
• Qwen 2.5 7B wellbeing probe: peak Cohen's d=3.5finding0.807
  Strongest cross-family probe; explains clearer introspection in Qwen than Gemma
• LLaMA E3 geometry summary: S_max = −1.896 ± 0.211, AUS_N = −2.119 ± 0.198, peak layer ℓ* = 10 [IQR 0.384]finding0.798
  Seed-pooled geometry statistics for LLaMA in E3, providing quantitative basis for geometry-to-behavior correlate
• LLaMA-3.1-8B: Sbmax = -1.896 ± 0.211, AUSN = -2.119 ± 0.198, peak layer ℓ* = 10 (median)finding0.797
  Seed-pooled geometry-only statistics (per-dev z units).
• Interest probe score drifts positively across turns: LMM slope=0.005, p=4.12×10⁻¹⁴ in LLaMA-3.2-3Bfinding0.789
  Demonstrates genuine internal-state dynamics in LLMs during multi-turn conversation
• MM probe trained on likely dataset achieves NIE of 0.70 (false→true) on LLaMA-2-13B, surprisingly strong but weaker than truth probesfinding0.784
  Likely-trained MM probe is a surprisingly effective causal baseline due to correlation between truth and probability on sp_en_trans