These are research hypotheses, not data. The candidate causal pairings below are
LLM-generated and/or partially literature-grounded. They live entirely separate from the
canonical relationships — they never feed the graph or the simulation. Promoting one
creates an unconfirmed relationship in the Verify queue; nothing becomes real without
your confirmation. (Per §3.3 relationships don't originate from inference, and §11.5 — gaps
are not filled with plausibility.)
Coverage of the structural space
Rows = source, columns = target. Blue = an existing relationship; amber = a candidate hypothesis (by plausibility); blank = unassessed (honest sparsity — most of the 156 directed pairs).