Experiments, benchmarks & scientific validation
Evidence before conclusions.
Adaptive topology has to earn its place: against a fair fixed-topology baseline, on matched initial conditions, across many seeds, with variance reported. This dashboard runs those comparisons and keeps every record with its provenance.
Research principle
A computational result is not a scientific conclusion until its assumptions, controls, and reproducibility have been established.
Matched comparison
Fixed vs adaptive topology, same mesh and initial states per seed. Runs the browser reference engine.
Idle.
Results
Browser reference engine · real computationRun a comparison to see per-seed results, paired differences and quality distributions.
Time-averaged J per seed
No data yet.
Final face-quality distribution
No data yet.
Run registry
Click any run for its configuration, initial conditions, seed, version, environment and downloadable artifact.
| Run | Algorithm | Provenance | Seed | Objective J | Sync err | Mean q | Valid | Flips ✓/✗ | Gates | Time |
|---|---|---|---|---|---|---|---|---|---|---|
| Loading… | ||||||||||
Quantum-hardware experiments
No hardware experiment records exist in this registry. When real tasks are executed, they will be recorded with provenance hardware, backend, device ARN, shots and task ID — and shown separately from simulator results.
Methodology checklist
- Fair baselines — same mesh, states, parameters and budget; only topology adaptation differs.
- Matched seeds — paired comparison removes initial-condition variance.
- Multiple runs — report mean, standard deviation and n; a single seed is an anecdote.
- Objective ≠ goal — lower J is the design criterion; whether that helps downstream computation is a separate question.
- Circuit equivalence — numerical checks need independent verification; noise assumptions must be stated.
- Provenance — browser-reference results are not repository-engine or hardware results.