Deep Think with Confidence

Yichao Fu, Xuewei Wang, Yuandong Tian +1

Large Language Models (LLMs) have shown great potential in reasoning tasks through test-time scaling methods like self-consistency with majority voting. However, this approach ofte…