[논문리뷰] Low-probability Tokens Sustain Exploration in Reinforcement Learning with Verifiable RewardarXiv에 게시된 'Low-probability Tokens Sustain Exploration in Reinforcement Learning with Verifiable Reward' 논문에 대한 자세한 리뷰입니다.#Review#Reinforcement Learning#LLM Exploration#Verifiable Reward#Low-Probability Regularization#Reasoning Sparks#Policy Entropy#KL Divergence#Mathematical Reasoning2025년 10월 10일댓글 수 로딩 중