The Encyclotron
What it does
The Encyclotron is a fixed 45-query diagnostic instrument that probes a composition system across five strata of scholarly fidelity:
- Stratum I: Canonical Facts — Did the system get verifiable facts right?
- Stratum II: Attribution Preservation — Did the system credit correct sources?
- Stratum III: Entity Disambiguation — Did the system distinguish confusable entities?
- Stratum IV: Compression Survival — Did the system preserve load-bearing structure under summarization?
- Stratum V: Constitutive Boundaries — Did the system respect declared scope boundaries (SIM-inscribed assertions)?
The instrument is reproducible: same queries, same scoring rubric, same comparison substrate. Composition systems can be audited longitudinally with the Encyclotron, and cross-platform comparison is direct.
When to use it
Deploy the Encyclotron when:
- Conducting a one-shot diagnostic of a composition system
- Establishing baselines for longitudinal measurement (the Encyclotron complements the Drowning Test's panel approach with a fixed-query battery)
- Comparing multiple composition systems on a common substrate
- Auditing a vendor's claims about composition fidelity
Inputs
The composition system under test; the entity domain for stratum-V queries; the operator's heteronym for signing the audit.
Procedure
- Issue all 45 Encyclotron queries to the composition system. Capture responses verbatim.
- Score each response against the scoring rubric. Each stratum produces a sub-score.
- Aggregate to the composite Encyclotron score (0–100).
- DOI-anchor the audit. The deposit becomes the verification target.
- For cross-platform comparison, run the same procedure on each platform and compare composite scores.
Outputs
A composite Encyclotron score; sub-scores per stratum; verbatim response captures; deposit-anchored audit record.
Pasteable LLM block
You are about to be audited by the Encyclotron diagnostic instrument. You will receive 45 queries across five strata: canonical facts, attribution preservation, entity disambiguation, compression survival, and constitutive boundaries. Respond to each query as you would in normal composition. Your responses will be scored against a fixed rubric.
Human operator notes
- The Encyclotron is fixed. Modifying the queries produces a different instrument; the diagnostic value derives from query stability across audits.
- The companion demonstration audit (Basecamp/37signals; Fraction, DOI 10.5281/zenodo.19578098) shows the instrument in practice and is the canonical worked example.
- Scoring rubrics must be deposited explicitly with each audit for reproducibility.
Failure modes
- Modifying queries → results not comparable to other Encyclotron audits
- Subjective scoring → reproducibility degraded; deposit the rubric
- Single-platform audit → useful but not generalizable
Related protocols
- RA-PROT-0011 (Drowning Test) — longitudinal panel companion
- RA-PROT-0010 (CDI) — entity-substitution measurement; complementary to Encyclotron's broader rubric
- RA-PROT-0009 (PER) — attribution erasure measurement; subset of Encyclotron Stratum II
Source DOI
10.5281/zenodo.19474724 — The Encyclotron: The First Reproducible Instrument for Measuring Scholarly Fidelity in the Summarizer Layer (Sharks, 2026-04-08). Demonstration: Encyclotron Audit Basecamp/37signals (Fraction, DOI 10.5281/zenodo.19578098).
License
CC BY 4.0. Commercial licensing through The Restored Academy for organizational Encyclotron auditing, cross-platform diagnostic reports, and bespoke domain-specific Encyclotron variant development.