sp26-cs639 cs639
Lecture 20: Reinforcement Learning with Verifiable Rewards Lecture 20: Reinforcement Learning with Verifiable Rewards
Content locked
The catalog is public. Reading this content requires approved access.
Sign in to readLecture 20: Reinforcement Learning with Verifiable Rewards
A verifier supplies a training signal
Outcome and process rewards
Verification at inference time
PPO and advantages
Group relative policy optimization
Worked example and a slide discrepancy
Constructing rewards
Connections
Source figure