finding
active
finding:q8b-most-often-peaks-at-layer-20-followed-by-layer-25-g20b-most-often-peaks-at-layer-15Q8B most often peaks at layer 20, followed by layer 25; G20B most often peaks at layer 15
Best steering layer is model-specific; linearly accessible trait information is organized differently across models
Source paper
extracted_from(2026) · Winston Zeng · Ali Emami · J H Choi
Neighborhood — ranked by edge-count
Papers (1)
paper
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Median layer where S(ℓ) peaks, across seeds.
- Layer-wise analysis revealing which network depths best encode strategic deception semantics
- G20B has 7 steerable, 7 natural, 4 intractable generic traits, with more mass at extremes than Q8Bfinding0.769G20B's post-training commits more strongly toward and against particular dispositions, leaving fewer in the steerable middle
- Interpretation of E3 layer-wise results; motivates targeted UCCT interventions at layers 8-12
- Moderate mid-layer effect.
- Table 2.
- Contrast with PGD.
- Implies α≈0.5-0.7.