Historical research record

Arena: a preference evaluation

The final report records 59 blind preference votes: 38 preferred TCM-Sage, 15 preferred the plain-model baseline, and 6 were ties. This is a frozen historical sample from the report, not a live counter.

Votes measure overall preference. They do not establish clinical effectiveness, citation accuracy or absence of hallucinations. The report notes a limited evaluator pool, mostly TCM students, and iterative system changes during evaluation. Fifty-nine votes does not imply 59 independent participants.

Public live Arena generation and voting are disabled. Visitor activity is not mixed into this historical research sample.