finding
active
finding:pgd-raises-l0-by-2-8-at-layer4-351-4-vs-126-0-cleanPGD raises L0 by 2.8× at layer4 (351.4 vs 126.0 clean).
Table 2.
Source paper
extracted_from(2025) · Gorton, Liv · Lewis, Owen
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Interpretation of layer4 L0 spike.
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.
- Moderate mid-layer effect.
- Contrast with PGD.
- Empirical demonstration of VPD on a mid-scale transformer, establishing feasibility.
- Q8B most often peaks at layer 20, followed by layer 25; G20B most often peaks at layer 15finding0.756Best steering layer is model-specific; linearly accessible trait information is organized differently across models
- Early layer effect.
- Concrete numerical example showing detection and control are nearly identical at peak apparent accuracy
- Forks missed with coarse spacing.
- Validates representational drift theory: later layers specialize for next-token prediction, increasing dr