New Reinforcement Learning Method Improves AI Code Generation Without Correct Answers

Original: Entropy-Regularized Rank-Masked Policy Optimization for Test-Time Reinforcement Learning in Code Generation

Source: arxiv.org ↗