Opaque Chain-of-Thought

A form of chain-of-thought output produced by large language models after reinforcement learning training, characterized by compressed and difficult-to-read representations that humans cannot easily interpret, distinct from true internal Neuralese representations.