# Reinforcement Learning with Verifiable Rewards

> Reinforcement learning with verifiable rewards (RLVR) trains a language model with rewards computed by an automatic check of correctness, such as whether a math answer matches the reference or code passes its tests, instead of by a learned reward model.

- Identifier: PTL-0156
- Category: Reasoning Models & Test-Time Compute
- Canonical URL: https://protologue.com/t/rlvr/
- Also known as: RLVR, RL from verifiable rewards, outcome-based reinforcement learning
- Introduced: 2024

## Description

Verifiable rewards are hard to game in the ways a learned reward model can be, and large-scale RL with them is how reasoning models learn to produce long, self-correcting chains of thought. The approach is limited to tasks with checkable answers, so open-ended writing and judgment still rely on preference-based rewards.

## Related terms

- [Reasoning Model](https://protologue.com/t/reasoning-model/)
- [Reinforcement Learning from Human Feedback](https://protologue.com/t/rlhf/)
- [Reward Model](https://protologue.com/t/reward-model/)
- [Reward Hacking](https://protologue.com/t/reward-hacking/)
- [Self-Taught Reasoner](https://protologue.com/t/self-taught-reasoner/)

## Sources

- Lambert et al. (2024). Tulu 3: Pushing Frontiers in Open Language Model Post-Training. https://arxiv.org/abs/2411.15124
- DeepSeek-AI et al. (2025). DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. https://arxiv.org/abs/2501.12948

## Cite this entry

Protologue. (2026). Reinforcement Learning with Verifiable Rewards. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0156). https://protologue.com/t/rlvr/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)