Testing LLMs on Cryptic Puzzles – How Smart Are They, Really?