How Well do LLMs Compress Their Own Chain-of-Thought? A Token Complexity Approach | Best AI papers explained | Podwise