The technical report does go into a lot of depth about how they use RL, such as the modified GRPO objective they use. As far as the README, I imagine most peopl

The technical report does go into a lot of depth about how they use RL, such as the modified GRPO objective they use. As far as the README, I imagine most people active in the field understand the implications of "RL" for a reasoning model.