Explore Briefly, Then Decide: Mitigating LLM Overthinking via Cumulative Entropy Regulation

Open in new window