Reproducibility
Do not believe the prose. Run the artifact.
The public reproducibility package is the main near-term artifact. Until runnable scripts, seeds, logs, and expected outputs are published, reference-path claims remain simulation-backed but pending public replication.
Minimum Public Package
- exact source files;
- environment file;
- random seeds;
- input parameters;
- expected outputs;
- collision counts;
- entropy measurements;
- replay reports;
- failure cases;
- README with commands;
- license and patent notice;
- independent reviewer instructions.
Target Command Shape
# planned command shape, not yet a public release claim
symlan run examples/survey.sym --backend hopfield_sim --seed survey-run
symlan replay outputs/survey.symvoc --trials 1000
symlan report outputs/survey.symvoc
Tests To Publish
| Test | Purpose | Status |
|---|---|---|
| Alpha / capacity sweep | Measure basin/glyph behavior across load factors. | Pending Release |
| Threshold sensitivity | Measure collision/entropy sensitivity to readout thresholds. | Pending Release |
| Collision test | Report when distinct basin candidates produce the same glyph. | Pending Release |
| Replay validation | Measure whether a resolved token can be re-associated with the expected accepted vocabulary state under declared conditions. | Pending Release |
| Expected failure suite | Demonstrate that overloaded or unstable configurations fail cleanly. | Pending Release |
Rule For Public Claims
Until the public package is available, the correct claim is:
Reference simulation direction defined; reproducibility package pending.