---
title: "Discussion on \"RLHF: How ChatGPT Learned What You Actually Want\n\" | Hashnode"
description: "Discussion on \"RLHF: How ChatGPT Learned What You Actually Want\n\". Discover how the three-step RLHF process and reward models taught ChatGPT to provide surprisingly useful answers"
url: "https://hashnode.com/posts/rlhf-how-chatgpt-learned-what-you-actually-want/6a4333b33737fc0ee47337a4"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [RLHF: How ChatGPT Learned What You Actually Want](https://changeofbasis.hashnode.dev/rlhf-how-chatgpt-learned-what-you-actually-want?utm_source=hashnode&utm_medium=feed)

Discover how the three-step RLHF process and reward models taught ChatGPT to provide surprisingly useful answers

Jun 30 · 5 min read · changeofbasis.hashnode.dev  
Read the article

-   [#rlhf](https://hashnode.com/tag/rlhf)
-   [#reward-model](https://hashnode.com/tag/reward-model)
-   [#reward-hacking](https://hashnode.com/tag/reward-hacking)
-   [#instructgpt](https://hashnode.com/tag/instructgpt)

## Written by

[Raghavan T M](https://hashnode.com/@raghav-1)

@raghav-1 · Chennai, India

  
[View profile](https://hashnode.com/@raghav-1)

## Responses

Markdown supported · ⌘↵ to post

No responses yet.