Phi-4

Backbone model used in E3 robustness overlay.

Neighborhood — ranked by edge-count

method

E3: Layer-wise Geometric Trajectory Analysis
uses
Quantitative study correlating layer-wise anchoring geometry (S_max, AUS_N) with behavioral thresholds θ50

finding

Math and code tasks show strongest mid-layer anchoring on LLaMA (S ≈ −1.65 at layers 8-12)
associated_with
Task-specific E3 finding showing compositional reasoning requires deeper processing

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.

Gamma et al. (Gang of Four)concept0.760
Authors of the book 'Design Patterns'; the Memento pattern is directly referenced for view creation.
Phi Score (Extreme Steering Score)concept0.753
Best SJT steering score for a given method, instrument, layer, stride, trait, and direction combination; the primary comparison metric
GPT-4concept0.735
Large language model underlying ChatGPT and Bing Chat; used for illustrative quotes in the paper
Φmax (IIT 3.0)concept0.735
Maximum Φ over all subsystems; represents the most integrated subsystem (main complex) under IIT 3.0.
Phi-4 shows U-shaped cohesion with falling mismatch; peak depth varies by modelfinding0.733
E3 backbone-specific finding showing three-stage trajectory generalizes across architectures
Steerability Score (Phi)concept0.729
Aggregate metric averaging mean SJT scores across OCEAN traits and steering directions; maximum possible is 10
Claude 4 Opusconcept0.707
Anthropic model; outlier in Experiment 1 with high baseline affirmation including under zero-shot and history conditions
harmonyconcept0.703
Deep geometric and experiential coherence produced by morphogenesis.