Training
Reinforcement learning
Training by trial and reward: the model tries actions or answers and is updated toward those that score well.
Beyond RLHF, reinforcement learning on verifiable tasks such as maths and code is a key ingredient of reasoning models, which learn to think longer when it improves the result.
Frequently asked questions
What does Reinforcement learning mean?
Training by trial and reward: the model tries actions or answers and is updated toward those that score well. Beyond RLHF, reinforcement learning on verifiable tasks such as maths and code is a key ingredient of reasoning models, which learn to think longer when it improves the result.