<result> <column_name>EXIT RATE (%)</column_name> <family>uncertain</family> <subtype>none</subtype> <confidence>0.4</confidence> <requires_user_confirmation>true</requires_user_confirmation> <sum_meaningful>false</sum_meaningful> <signed>false</signed> <reducers> <primary_reducer>mean</primary_reducer> <valid_reducers>mean, median, min, max, distribution</valid_reducers> <excluded_reducers>sum (a percentage of 0-100 does not accumulate across entities), product (no sequence to compound along)</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>The table is STATE with no sequence axis and one row per entity, so the gate makes multiplicative highly implausible. The only multiplicative evidence is the name ("RATE", "%"). The data contradicts it: values are bounded 0-100 on a percent scale, the median is 0.54 and the mean 23.76, 45% are zeros, and the center is not near 1.0. Log symmetry gain is slightly negative. Nothing here can be reconstructed or compounded. The values behave like a bounded intensive level where averaging is the natural rollup, but the evidence for that is not strong enough to commit to a subtype. I abstain with "uncertain" and flag it for review, leaning toward intensive_additive (mean-based) if a human confirms.</reasoning> <table_type_conflict>false</table_type_conflict> </result>