```xml <result> <column_name>recall_at_k</column_name> <family>additive</family> <subtype>intensive_additive</subtype> <confidence>0.85</confidence> <requires_user_confirmation>false</requires_user_confirmation> <sum_meaningful>false</sum_meaningful> <signed>false</signed> <reducers> <primary_reducer>mean</primary_reducer> <valid_reducers>mean, min, max, median, distribution</valid_reducers> <excluded_reducers>sum (summing a performance metric is meaningless), product (not a ratio of sequential changes)</excluded_reducers> <average_semantics>arithmetic_mean</average_semantics> </reducers> <inner_outer_measure> <inner_measure_reducer>min</inner_measure_reducer> <outer_measure_reducer>max</outer_measure_reducer> </inner_outer_measure> <multiplicative_partner_derivable>false</multiplicative_partner_derivable> <reasoning> Table-type gate: DIMENSION with no sequence axis makes multiplicative highly implausible. Despite center_near_one and range_bounded_near_one, "recall_at_k" is a performance metric (proportion of relevant items retrieved), which is an intensive level/score, not a ratio of sequential changes. Values [0.75, 1.0] represent absolute performance levels at different entities; averaging across entities is meaningful, but summing is not (you cannot add recall scores). The column is bounded [0,1] by definition, centered away from both 0 and 1, fitting intensive_additive. No multiplicative naming tokens present. </reasoning> <table_type_conflict>false</table_type_conflict> </result> ```