Tag
This paper introduces LivingArena, an automated evaluation framework where LLMs probe each other's weaknesses by generating questions, enabling contamination-resistant and scalable assessment that adapts as models improve.