claim
active
claim:predictive-component-not-dominating-at-deep-layers-is-a-feature-since-deeper-layers-discard-contextual-information-needed-for-unembeddingPredictive component not dominating at deep layers is a feature, since deeper layers discard contextual information needed for unembedding.
Layer choice discussion.
Source paper
extracted_from(2025) · Lubana, Ekdeep Singh · Rager, Can · Hindupur, Sai Sumedh R. · Costa, Valerie +12
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Fig. 14.
Claims (1)
claim
- Speculation about layer dependence.
Quotes (1)
quote
- Interpretation of deep-layer result.
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.
- Fig. 13.
- The same block construction predicts a richer contour code (curvature plus position) in deeper layers.hypothesis0.773Forward prediction analogous to V4.
- Methodological critique of prior work that fixed a single layer for truth probing.
- Interpretation of LAT scanning results showing layer-dependent deception detection accuracy
- The middle layer residual stream features are causally implicated in multi-step reasoning.claim0.765Features for Kobe Bryant, California, Lakers participate in computing the capital answer.
- Circuit simplification theory.
- Importance of hierarchical structure for flexible coordination.
- Attribution finding suggesting the last layer directly controls reflection keyword generation