claim

active

claim:layer-wise-geometry-shows-early-dip-mid-layer-alignment-and-late-standardization-across-tasks

Layer-wise geometry shows early dip, mid-layer alignment, and late standardization across tasks

Qualitative pattern from E3.

Source paper

extracted_from

The Guanyin Protocol: A Framework for Immediately Establishing an Understanding of Both Causality and Compassion in LLM Systems Using Semantic Anchoring

(2025) · Edward Yi Chang · Kaya, Zeyneb N. · Ethan Chang

Neighborhood — ranked by edge-count

Communities (3)

community

Manifold-aware concept steering in neural representations
members_of
Explores geometry of activation/behavior manifolds to enable selective, non-destructive concept interventions.
Layer-wise geometry predicting few-shot learning
members_of
Silhouette-based metrics (Sbmax, AUSN) across LLM layers predict task accuracy and few-shot thresholds.
Symmetry constraints in Islamic geometric design
members_of
Mathematical analysis of frieze groups and isometries in Islamic planar tilings, focusing on glide-reflection avoidance across Alhambra and Alcázar sites.

Artifacts (1)

artifact

Semantic Anchoring in LLMs: Thresholds, Transfer, and Geometric Correlates
supports
Main paper presenting UCCT and semantic anchoring framework.

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

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

Layer-wise trajectories show early enrichment, mid-layer alignment, and late re-clustering.claim0.838
Qualitative geometry pattern.
For simple factual tasks F0-F3, probe directions show a sharp geometric transition in middle layers, with late-layer probes converging to high cosine similarity; A3 and F4-F5 show no clear transition.finding0.790
Geometric evidence for convergence to stable truth directions only for simpler tasks.
Layer-wise geometry summaries (Sbmax, AUSN) predict internal few-shot thresholds θ50claim0.785
Claim that geometry-to-behavior correlates exist
In LLaMA-2-13B, cities and neg_cities show antipodal alignment in early layers, rotate to orthogonal in middle layers, then eventually align in later layersfinding0.784
Layer-by-layer evolution of truth direction alignment, supporting hierarchical abstraction hypothesis
The model converges to a more stable truth direction in middle-to-late layers, as evidenced by increasing cosine similarity between layer-wise probes.claim0.782
Supported by the geometric transition visible in cosine similarity heatmaps for F0-F3.
Single-layer analyses can be misleading because early-layer truth directions may reflect surface features with limited cross-task generalization.claim0.777
Methodological critique of prior work that fixed a single layer for truth probing.
Linear alignment map ϕ_lin shows substantial IIA decrease in third layer for both equality relations and left equality relation algorithms in hierarchical equality taskfinding0.764
Replicates Geiger et al. 2024b pattern of layer-dependent IIA degradation with linear maps
Layer-wise anchoring peaks in a 'Goldilocks zone' between early and late layers.claim0.764
Qualitative characterization of optimal anchoring depth.