OneBench
From Leaky Thoughts to Private Reasoning: Controlling What LRMs Say to Themselves | OneBench: AI Insights