ModelRefs / Verifiable Reward — AI Glossary
Verifiable Reward — AI Glossary
A training signal derived from objectively checkable outcomes (code execution, math verification) rather than human preference labels.
Overview
Verifiable rewards (DeepSeek-R1, OpenAI o1) sidestep reward hacking by grounding RL feedback in correctness checks: code compiles and passes tests, math answers match verified solutions. Enables self-improvement at scale without human labelers. Combined with process reward models (PRM) to credit intermediate reasoning steps.
Reference details
| Topic | training |
|---|---|
| Also known as | outcome-based reward, verified reward |
| Last reviewed | 2026-06-24 |
Related terms
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Verifiable Reward — AI Glossary.
Frequently asked questions
What is Verifiable Reward?
A training signal derived from objectively checkable outcomes (code execution, math verification) rather than human preference labels.
Is Verifiable Reward the same as outcome-based reward?
Yes — outcome-based reward, verified reward are common aliases for Verifiable Reward.
What concepts are related to Verifiable Reward?
Closely related concepts include reinforcement learning, reward model, reasoning.