finding
active
finding:seed-sensitivity-in-the-2-and-3-layer-resultsSeed sensitivity in the 2- and 3-layer results
Limitation.
Source paper
extracted_from(2025) · Lucius Bushnaq · Dan Braun · Lee Sharkey
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.
- Shows that Burger et al.'s layer choice corresponds to a transitional phase, not a universal property.
- Preference steering effective only in mid-layer range though correlations hold across layersfinding0.725Suggests mid-layer reading circuitry.
- Gemma-3-4B-it shows three-stage layer trajectory and S(ℓ) peak despite scale differences in dr and ρdfinding0.723E3 backbone generalization finding for Gemma; validates pattern across diverse architectures
- Empirically observed pattern in E3: early enrichment (ρd dips), mid-layer alignment (dr falls), late standardization (re-clustering)
- Claim that geometry-to-behavior correlates exist
- Argues against the single-layer analysis approach of prior work.
- Thought detection peaks at ~2/3 layer depth; intention checking peaks at ~1/2 layer depth.finding0.718Lindsey (2026) differential layer performance explained by Janus's path combinatorics — different tasks use different path distributions.
- Task-specific peak anchoring score for structured reasoning domains.