I tested four decision models against five small LLMs on 6,662 decisions. The strongest model changed by task, and the full harness and raw calls are public.pournasserian.com/writ...

Sign in to view linked content