STAR-1: Safer Alignment of Reasoning LLMs with 1K Data Paper β’ 2504.01903 β’ Published 13 days ago