| # Recommended cross-domain diagnostic | |
| Apply the same source transition and aggregation rule to a natural-photo roster | |
| and a museum-specimen roster. Report the sign and interval of the paired change | |
| on each domain, then use expansion-only and depth-only reconstructions to | |
| separate new-label competition from updates to existing prototypes. | |
| Fish-Vista is not redistributed here. Obtain it from its authors and apply the | |
| published adapter. This diagnostic is recommended because a source update can | |
| help natural photographs while harming specimens under the same aggregation | |
| rule; it is not a canonical leaderboard track. | |