Two Independent Scores, One Corpus

Quality and Factuality are measured separately — quality captures structural rigor, factuality captures verifiability. A document can be well-structured and poorly sourced, or vice versa. Both matter.

Last measured: April 27, 2026 (after 18-document counter-argument improvement pass and full canonical bibliography migration). Scores update when the pipeline reruns on the full corpus.

Eight Dimensions — 100 Points Total

Each dimension is weighted by importance. Structural compliance and counter-argument rigor carry the most weight — these are the hardest to get right and the most important for reader trust.

Counter-Argument Rigor is the hardest dimension to improve — it requires finding real, published scholarly challenges for every document, not inventing objections. It has improved from 51% (early 2026) to 70% through a systematic improvement pass. It remains the primary quality target.

How the 3,632 documents Are Distributed by Source Quality

The [N/5] source confidence rating is assigned to each document based on its weighted bibliography score. Distribution reflects a corpus built for breadth — many topics have limited peer-reviewed literature available.

Why so many [1/5] documents? The [1/5] rating (under 14 weighted points) captures two types of document: those covering topics where peer-reviewed literature is genuinely sparse (oral history, ancient mythology, some alternative archaeology), and some older documents still pending bibliography enrichment. It does not necessarily mean the content is wrong — it means the sourcing is thin and should be treated with extra caution.

Where We Are, Where We're Going

Quality improvement is ongoing and tracked. The targets below are the current active benchmarks — not arbitrary, but tied to specific improvements in research reliability.