Constructive Accessibility Boundary in Random 3-SAT
Measured, process-specific A0 constructive success versus a modeled existence-side reference. Drag to evaluate the matched-success-rate separation; toggle among the three displayed system sizes.
Measured A0 successSeparate oracle bridge
System size
α ≈ 1.10
SR ≈ 99%
SAT
reachable
REACHABLE
← drag across the plot · arrow keys work →
Measurement boundary: Orange dots are empirical success rates for the declared forward, non-backtracking A0 process; the orange logistic is a visual smoothing guide, not the estimator used to establish every reported result. The teal curve is modeled: the literature threshold reference αc ≈ 4.267 for n ≥ 400 and a finite-size DPLL midpoint at n = 50. The shaded separation is therefore between a measured process curve and a modeled existence reference. The exact result belongs to a separate bridge test: 194/194 tested jammed prefixes at n = 1,000 and α ∈ {1.53, 1.55} were residual UNSAT under their realized commitments; that bounded oracle result does not make the whole orange curve exact.
Displayed sizes: n = 50, 400, 1,000Underlying study: nine sizes, n = 20–1,000Sample basis: n = 50: 5 instances × 15 branches/cell; n = 400 and 1,000: 20 instances × 10 branches/cell; densified n = 1,000 band: 200 branches/αInstrument: NORTHSTAR v31.9 / CAI v0.3Prediction record:OSF preregistration and results; finite-asymptote forecast ≈ 1.541 before the densified n = 1,000 sweep, observed 1.5352 (difference 0.006)Archive:Zenodo record 19225548Release bundle: NORTHSTAR v31.14 · 2026-04-24Locally audited bundle SHA-256:d8604f0d4e4323179c31eeff410ecc20a96351e7de2779af63be5c8de843651d (the public-record file identity was not independently rechecked in this build)