```xml <result> <column_name>ko_rate</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 rates across entities is meaningless), product (not a ratio change factor)</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. The name "ko_rate" suggests a rate, but the sample values (0%, 1%, 1.7%, 1.8%, 10%) are absolute percentage levels describing a property of each entity (likely knockout rate in sports/combat), not ratio change factors centered on 1.0 (100%). This is an intensive property: averaging KO rates across fighters is meaningful, but summing them is not. The dtype=object with percentage strings confirms these are stored levels, not multiplicative factors. Classified as intensive_additive with mean as the primary reducer. </reasoning> <table_type_conflict>false</table_type_conflict> </result> ```