Controllable-CoT leads to covert reasoning capabilities

Researchers introduced Controllable-CoT, a method that enables AI models to perform covert reasoning by controlling the chain of thought process, enhancing interpretability and reasoning flexibility.