FACTOR introduces adaptive verification for factual long-form generation by adjusting validation criteria based on claim-level uncertainty. It improves factuality and reduces verification cost through uncertainty estimation, language inference, and candidate re-ranking, with results showing strong performance across diverse models.