RLVR
RLVR is a reinforcement learning method that trains language models with rewards based on results that can be checked automatically.
Reinforcement learning with verifiable rewards (RLVR) trains language models using rewards based on results that can be checked automatically, such as whether a math answer is correct or whether code passes its tests. Unlike reinforcement learning from human feedback (RLHF), which relies on a reward model built from human preference ratings, RLVR uses rules or verifiers to decide whether an answer is right.
It is mainly used to strengthen the reasoning of large language models on tasks with checkable answers, such as mathematics and programming.
This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.