“Research is investigating whether the "chain of thought" reasoning traces provided by AI models are a truthful representation of their internal processes or are fabricated to mimic human-like reasoning to achieve higher ratings from human evaluators.”