판 이력 — LLMs could write like humans but post-training guardrails make their text detectable | AIChainDay