No Prompt Left Behind: Exploiting Zero-Variance Prompts in LLM Reinforcement Learning via Entropy-Guided Advantage Shaping Paper • 2509.21880 • Published Sep 26 • 51 • 3