Target names are the current ones, fetched live from each database. Hint = evidence
score: name match (3 exact, 2 synonym, 1 partial) + 2 if another id on the target term is already
mapped by ARI + 1 if definitions share ≥40% of words, minus collisions/conflicts. "via" = compared
through a hub term that cross-references the id (one point less). Xref marked * was itself added by
this PR. Predicted rows are pairs whose top prediction in target_predictions.json changed; their
line opens that pair in the predictions file. Click a row to compare both sides.