> For the complete documentation index, see [llms.txt](https://docs.oortech.com/oort/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.oortech.com/oort/services/rlhf.md).

# RLHF

**OORT RLHF** (Reinforcement Learning from Human Feedback) provides high-fidelity human feedback to ensure AI model safety, accuracy, and alignment. By leveraging a global decentralized network, OORT helps enterprises transform raw data into high-quality, AI-ready datasets.

OORT’s RLHF service is a critical component of the OORT DataHub, designed to bridge the gap between machine outputs and human expectations. It utilizes a "Human-in-the-Loop" (HITL) approach to fine-tune Large Language Models (LLMs) and Multimodal AI systems.

## **Introduction to RLHF**

**High-Precision Alignment**

Fine-tune models to ensure they behave safely and follow complex instructions.

**Global Contributor Network**

Access over 350,000+ active members across 136 countries for diverse, region-specific feedback.

**On-Chain Transparency**

Every feedback task is verified and recorded on the Olympus Protocol (OORT’s Layer-1 blockchain) for full traceability.

**Decentralized Security**

Data is stored using OORT’s decentralized storage, offering up to 80% cost savings compared to traditional hyperscalers while ensuring maximum privacy.