1. CAUSAL BAYESIAN MIXTURE MAE
0.4119
vs naive random 0.7709 & matched random 0.4632 · PASS
2. AST SANDBOX PYTHON CODE-RSI
0.5
G0 0.375 · CM 0.375 · AST escapes blocked 4/4
3. 6-GEN COMPOUNDING (G0→G5)
0.25 → 0.875
vs 2× compute-matched 0.25 · rollback exact=True
4. 3-DOMAIN 4-ARM UNIFIED D
0.8133
vs A 0.14 · B 0.3667 · C 0.3433
5. NEURO-SYMBOLIC & IMMUNE
1.0 / 10/10
Raw 0.0 · OOD abstain 1.0 · Brier 0.0144
Full-Lifecycle Compute-Equalized 3-Domain 4-Arm Study (1,200 Scored Evaluations)
| Arm | Composite Success | Macro Synthesis | Causal (MAE≤0.45) | Causal Mean MAE | 2D Shift World Model |
|---|---|---|---|---|---|
| Arm A | 0.14 | 0.25 | 0.1 | 0.8081 | 0.07 |
| Arm B | 0.3667 | 0.34 | 0.57 | 0.443 | 0.19 |
| Arm C | 0.3433 | 0.79 | 0.15 | 0.7629 | 0.09 |
| Arm D | 0.8133 | 0.88 | 0.71 | 0.381 | 0.85 |
Holm-Bonferroni adjusted p-values: {"D_vs_A": 0.0, "D_vs_B": 0.0, "D_vs_C": 0.0}
6-Generation Governed Compounding Curve (G0 → G5)
| Gen | ID | Library | Max Depth | Sealed Rate | Unseen G5 Rate | Status |
|---|---|---|---|---|---|---|
| G0 | GEN-0-FROZEN | 21 | 1 | 0.25 | 0.25 | FROZEN_BASELINE |
| G1 | GEN-1-PROMOTED | 29 | 2 | 0.75 | 0.75 | PROMOTED |
| G2 | GEN-2-PROMOTED | 39 | 4 | 0.8333 | 0.8333 | PROMOTED |
| G3 | GEN-3-PROMOTED | 49 | 7 | 0.8333 | 0.9167 | PROMOTED |
| G4 | GEN-4-PROMOTED | 59 | 7 | 0.8333 | 0.9167 | PROMOTED |
| G5 | GEN-5-PROMOTED | 63 | 8 | 0.9167 | 0.875 | PROMOTED |