A new study from Arizona State University researchers suggests that the celebrated "Chain-of-Thought" (CoT) reasoning in Large Language Models (LLMs) may be more of a "brittle mirage" than genuine ...