Process reward model
A model that scores each step of a solution, not just the final answer, to guide reasoning.
Also: PRM, verifier
An outcome reward checks only whether the final answer is right. A process reward model judges each intermediate step, so it can tell a lucky guess from sound reasoning and point to where a solution went wrong. OpenAI's 2023 paper "Let's Verify Step by Step" trained one on 800,000 human step-level labels for maths problems and found it chose correct solutions more reliably. Process rewards can guide search during answering or training, though DeepSeek reported that simple outcome rewards worked better for training R1.