Skip to main content
All terms
Training

Reinforcement Learning from Verifiable Rewards

Training on tasks where correctness can be checked automatically by a program.

Definition

Reinforcement learning from verifiable rewards trains models on tasks where correctness can be checked by a program, such as passing unit tests or matching a math answer. The automatic reward avoids the noise of human labels and provides a clean training signal. It has been central to recent gains in reasoning models.