The Encyclotron is a measurement and diagnostic specification developed by Lee Sharks within Compression Studies and the Crimson Hexagonal Archive. It is designed to evaluate scholarly fidelity in AI summarizers: not merely whether an answer is factually correct, but which positions, citations, disagreements, and uncertainties survive when a field is compressed into a short generated response.
The instrument compares two explicitly constructed objects. The scholarly graph is a bounded, documented model of relevant literature; the retrieval graph is the aggregate of outputs returned by selected search summarizers, chat systems, and hybrid retrieval systems. Their difference is analyzed as compression loss, unsupported invention, and distortion. Additional measures include source-citation rate, confidence posture, temporal drift, retrieval density, and the beige score, which estimates the degree to which different platforms converge on the same flattened account.
An Encyclotron study is deposited as a time-stamped snapshot and repeated on a schedule or after major model and indexing changes. The document therefore defines both an audit method and a historical record of the summarizer layer. It also acknowledges a recursive feature: the deposited measurement can itself enter the scholarly and retrieval graphs it measures.
Referenced by 1 other entities in the archive. See the full Knowledge Graph for reference paths, or the primary record for the full deposit with reference details.