```xml <result> <column_name>tokens_output</column_name> <family>additive</family> <subtype>ordinal_additive</subtype> <confidence>0.85</confidence> <requires_user_confirmation>false</requires_user_confirmation> <sum_meaningful>true</sum_meaningful> <signed>false</signed> <reducers> <primary_reducer>sum</primary_reducer> <valid_reducers>sum, mean, min, max, median, count, distribution</valid_reducers> <excluded_reducers>product (not a ratio), geomean (not a ratio)</excluded_reducers> <average_semantics>arithmetic_mean</average_semantics> </reducers> <inner_outer_measure> <inner_measure_reducer>sum</inner_measure_reducer> <outer_measure_reducer>sum</outer_measure_reducer> </inner_outer_measure> <multiplicative_partner_derivable>true</multiplicative_partner_derivable> <reasoning> Gate prior: PATH table with sequence axis allows both multiplicative and ordinal_additive. Column name "tokens_output" indicates a count/quantity (additive signal). Statistics show large unbounded magnitude (max 20330), mean far from 1.0 (2164), and contains zeros (7.8%) - all strongly against multiplicative. The presence of zeros and large absolute values point to real_additive, but on a PATH table with entity grain, this represents instantaneous token counts at each sequence step, making ordinal_additive the correct subtype (sum across entities at fixed time, but ordinal along sequence). High log_symmetry_gain (2.72) is consistent with count data's typical distribution. </reasoning> <table_type_conflict>false</table_type_conflict> </result> ```