Impolite steering delta of +38.09 (p<0.001) at OLMo-3 main checkpoint

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

• Sycophantic steering delta of +42.46 (p<0.001) at stage1-step99000 (415B tokens) in OLMo-3finding0.843
  Peak same-checkpoint sycophantic steering result in OLMo-3 pretraining
• Evil persona steering delta of +9.69 (p=0.011) at stage1-step3000 (12.6B tokens) in OLMo-3finding0.824
  First significant evil persona steering result; establishes emergence at 0.22% pretraining
• Coherence scores remain between 60-70 across all cross-domain steering combinations in OLMo-3finding0.763
  Shows persona steering largely preserves response quality across evaluation domains
• Cross-concept steering: impulsivity→interest R² increases from 0.55 (α=-4) to 0.72 (α=+4), ∆R²=0.10, p=0.012 in LLaMA-3.2-3Bfinding0.749
  Second significant cross-concept introspection improvement; marginal after BH correction (q≈0.066)
• PM achieves overall SJT steerability Phi=9.6 on gemma-3-12b-it vs MDS=8.7 and P2=8.3finding0.742
  Per-model steerability comparison from Table 4
• Impulsivity probe: peak Cohen's d=3.60 (layer 13), p=3.58×10⁻¹³ in LLaMA-3.2-3Bfinding0.733
  Strongest probe validation result; highest Cohen's d among the four concepts
• Under activation steering on Llama-3.1-8B, inter-persona variance (0.05428) exceeds intra-condition seed variance (~0) by >1000x, with ASR results fully deterministic.finding0.733
  Variance decomposition showing AS results are dominated by persona identity
• Impulsivity→interest steering: probe entropy increases (LMM slope=0.024, p=2.30×10⁻⁴) but report entropy does not (p=0.11)finding0.732
  Evidence of a bottleneck between richer internal variation and final report distribution in impulsivity→interest condition