Mitigating Tail Narrowing in LLM Self-Improvement via Socratic-Guided Sampling

Open in new window