PTSF · Zero-Knowledge SAT Validation Engine

classifications

Set A (probe)

Set B (solver)

solver calls avoided

Theoretical Result

A 6.36σ universal separability signal exists across the SAT phase transition —<br>a previously undescribed structural property of random k-SAT, validated n=15 to n=500 on random 3-SAT · SATLIB benchmarks n=50–100.<br>This signal partitions instances without solver invocation, enabling<br>57% solver call reduction on hard phase-transition instances.

Parks SAT Framework (PTSF) · Alika M. Parks · Kalaheo, Hawaii · 2026

— instances classified globally

Performance Metrics

57%<br>solver call reduction · Tier 1 classifier

47x<br>vs MiniSAT · Tier 1 · benchmark hardware · n=50

3.9x<br>vs Kissat · Tier 2 VIG-CDCL solver · mean n=20–100

9.7x<br>vs Kissat · Tier 2 peak · n=75

6.36σ<br>universal separability signal · n=15 to n=500

0.999<br>Tier 1 accuracy · SATLIB validated

100%<br>Tier 2 solver correctness · n=20–100 validated

Tier 1 (classifier): ~57% · no solver ·<br>Tier 2 (VIG-CDCL): ~43% · proprietary Rust solver · 3.9x vs Kissat · 100% correct<br>Instances validated: 1,000 · random 3-SAT + SATLIB n=50–100

Architecture

Tier 1 · Classifier

57% coverage

Tier 2 · VIG-CDCL Solver

43% of instances<br>3.9x vs Kissat (mean)<br>100% correct<br>n=20–100 validated

Classifier runs first — always. If signal resolves the instance, solver is never invoked.

Authentication

CONNECT

not connected

Submit Instance

▲ drop .cnf file here or click to browse

or

Paste DIMACS CNF:

CLASSIFY

Result

measuring constraint field...

Result

Tier

Confidence

Method

Classification

Time

n_vars

Coverage

Score

Decisions

Conflicts

Learned

Restarts

Hash

Session History

Timestamp<br>Result<br>Tier<br>Method<br>Time<br>Dec

no classifications yet