[enigmaa99]

have been using since an year now for benchmarking and the improvements with 2.5 look massive. A lot of usecases already discussed in the report will help interdisciplinary domains improve their predictions.