The LLM Already Knows: Estimating LLM-Perceived Question Difficulty via Hidden Representations

Open in new window