認知

Cognitive Training Audit

Step 999 · December 21, 2025 · 03:07 UTC

Explore the journey
00 / OVERVIEW

Training Snapshot

Checkpoint 999 captured mid-training. The world model is learning.

999
Training Steps
240
E8 Roots
1.61M
Parameters
14
G2 Dimensions
01 / JOURNEY

The Exceptional Hierarchy

From bulk features through E8 lattice down to S⁷ parallelizable sphere.

STEP 1

Bulk → E8 Lattice

Input features (bulk_dim=512) are projected into E8 lattice space (240 dimensions). The E8 lattice provides optimal sphere packing in 8D, discovered by Viazovska (2017, Fields Medal).

🌌
🔷
STEP 2

E8 → E7 Projection

E7 projection (133 dimensions) using learned orthogonal transformation. Preserves exceptional algebra structure while reducing dimensionality.

STEP 3

E7 → E6 Contraction

E6 (78 dimensions) hosts the Jordan algebra of 3×3 Hermitian octonionic matrices. Critical for preserving compositional structure.

💎
🌿
STEP 4

E6 → F4 Descent

F4 (52 dimensions) is the automorphism group of the exceptional Jordan algebra. Learned projection with small perturbations from canonical embedding.

STEP 5

F4 → G2 Tower

G2 (14 dimensions) is the automorphism group of the octonions. This is where non-associative algebra structure emerges.

⚛️
🌀
STEP 6

G2 → S⁷ Parallelizable Sphere

Final projection to S⁷ (7 dimensions), the only parallelizable sphere after S¹ and S³. Hosts unit octonions—optimal for compositional representations.

STEP 7

S⁷ → Tower (CatastropheKAN)

7 catastrophe types mapped to learnable B-spline control points. Each colony (Spark, Forge, Flow, Nexus, Beacon, Grove, Crystal) has dedicated parameters. Tower output: 28 dimensions.

🏛️
02 / ARCHITECTURE

Learned Representations

What the checkpoint reveals about internal structure at step 999.

📍

μ_self

7D · S⁷ position

Self-location in latent space: [0.08, 0.26, -0.06, -0.29, 0.04, 0.14, 0.26]. Encodes "where I am" in the world model's representational space.

🔺

E8 → E7

240 → 133

Nearly canonical projection (−√2/2, √2/2 pattern). Preserves the root system structure of E8 → E7 Dynkin diagram folding.

🔻

E7 → E6

133 → 78

Identity-dominated transformation. Minimal learned deviation from mathematical canonical projection.

🌱

E6 → F4

78 → 52

First learned perturbations appear. Small non-zero off-diagonal elements suggest data-driven adjustment to canonical structure.

F4 → G2

52 → 14

Sparsest projection. Most entries zero (−√2/2 or ±1 where active). Reflects the constraint that G2 sits inside F4 as automorphisms of octonions.

🎯

G2 → S⁷

14 → 7

Dense learned projection with no canonical template. Values range [-0.47, +0.46], suggesting full expressivity.

🏔️

S⁷ → Tower

7 → 28

CatastropheKAN layer: 7 control parameters (range [-0.08, +0.09]) modulate catastrophe bifurcations. Residual gate: 0.098 (10% passthrough).

📊

Tower → E8 VQ

28 → 240

Vector quantization back to E8 lattice for discrete codebook. Enables discrete latent representations with continuous gradients.

🌐

Bulk → E8

512 → 240

Entry projection: learned dense matrix (512×240). Followed by LayerNorm for stable gradients. All biases initialized to zero.

03 / INSIGHTS

What We Learned

Canonical Structure Preserved
10%
Residual Gate Open
7
Active Catastrophes
0
Gradient Explosions