#ProRL

NVIDIA introduces ProRL, a novel reinforcement learning method that extends training duration to unlock new reasoning capabilities in AI models, achieving superior performance across multiple reasoning benchmarks.