gaia-evidence-subgraph

To turn a scientific reasoning claim into an auditable evidence graph, you provide a raw-json LKM payload to generate a methodological-decomposition map. This maps premises, factors, and derivation steps to a closure-chain representation of your quantitative result.

Bayesian closure in a chain-of-evidence graph makes the explicit reasoning path from observational anchors and modeling inputs to a final quantitative result reviewable. It ensures every derivation step maps directly to the root claim without synthetic bridging.

You generate a factor graph from a raw-json LKM payload by executing the Skill with the output from your claim search. It constructs a chain-bounded graph using only the returned premises and steps, categorizing edges as chain support, background, or verification support.

Yes, you can export the evidence graph to DOT or Mermaid formats. The Skill produces publication-ready structured artifacts, renderable sources, raster images, and verbatim raw payloads to ensure full traceability and auditability of the reasoning steps.

A key limitation when mapping a chain-of-evidence is strict chain-bounded construction, meaning the graph only includes nodes and edges from the provided LKM payload. It does not generate synthetic bridging or add external factors to the reasoning steps.

Yes, this evidence graph tool works for domain-agnostic scientific or technical reasoning. You can review how any measured or computed quantity is produced by mapping observational anchors, parameter choices, and derivation steps from the raw payload.