Lightweight screening agrees with full-sweep labels on 88.5% of G20B traits (46/52), skipping the sweep for 56% of traits

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

• Screening errors are conservative: no Q8B traits and one G20B trait (optimistic, Bt=76.0, Δt=10.3) are mislabeled as natural when steerablefinding0.820
  The costly error direction (labeling steerable as natural) almost never occurs
• G20B has 7 steerable, 7 natural, 4 intractable generic traits, with more mass at extremes than Q8Bfinding0.779
  G20B's post-training commits more strongly toward and against particular dispositions, leaving fewer in the steerable middle
• In G20B the five most-steerable traits are hyperbolic (45.50), creative/playful (39.76), excessive validation (34.95), sycophantic (32.74), and interpretive (29.73)finding0.769
  Steering preferentially exposes exaggerated styles in G20B as well
• G20B's post-training committed more strongly toward and against particular dispositions than Q8B, leaving fewer traits in the steerable middleclaim0.766
  Explains the distributional difference in generic domain: G20B has more mass at both extremes (natural and intractable)
• SFR-DR-20B achieves 28.7% on Humanity's Last Exam full text-only benchmark, 65% relative improvement over gpt-oss-20b baseline.finding0.761
  Main evaluation result showing best variant outperforms many proprietary and open-source baselines of comparable or larger sizes.
• 47.69% of 130 injection-manipulated alpha trends have near-linear fits (R2 >= 0.95); 96.15% have roughly linear fits (R2 >= 0.75)finding0.753
  Demonstrates alignment with Linear Representation Hypothesis: target trait steers approximately linearly with alpha
• Secure control fine-tuning leaves moral susceptibility S near base levels for GPT-4o (-9%), GPT-4.1 (-20%), and Qwen3-235B (+2%)finding0.740
  Shows that susceptibility spike is specific to misalignment-inducing training signal, not generic fine-tuning
• Steering at 6 layers (strength 0.6 each, total 3.6) outperforms single-layer steering at equivalent total strength for type hint suppressionfinding0.736
  Demonstrates distributed steering is more effective and less accuracy-damaging than concentrated steering.