Every claim in a Centaur report traces back to a source. This is not a design aspiration — it is an architectural requirement that shapes how the platform reasons about information and how it structures its outputs.
Why explainability matters for compliance
Buy-side compliance teams need to be able to answer a simple question about any research input: where did this come from, and why was it weighted the way it was? Black-box AI outputs cannot answer that question. Centaur's reasoning chain architecture is designed so that every assertion in a report carries a provenance record that compliance can follow back to the original source.
How reasoning chains are built
For each report, Centaur constructs a directed graph connecting claims to evidence. Each node in the graph is either a source document, a derived assertion, or a synthesis step. The edges carry confidence weights that reflect the strength of the evidence relationship. The final report is generated by traversing this graph, and the graph itself is stored alongside the report for audit purposes.