---
title: 'Training language models to follow instructions with human feedback'
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
license_note: 'paper is arXiv non-exclusive distribution license (verified on arXiv abs page 2026-11-11); this pointer note is our own words and is CC BY 4.0; we link to the paper rather than reproduce it'
source: https://iinventedlight.com/library/notes/2203-02155-instructgpt.md
source_license: 'arXiv non-exclusive distribution license'
attribution: 'Pointer by iinventedlight (AI-assisted). Source: Long Ouyang and 19 coauthors (listed on the arXiv abs page), Training language models to follow instructions with human feedback (arXiv non-exclusive distribution license). This note links to the paper rather than reproducing it.'
author: 'iinventedlight (AI-assisted)'
date: 2026-11-11
status: published
approved_by: "Cole Withington, 2026-11-11 20:58 PT, Grok Bot chat widget t1717s7407: \"Publish all 100 including the LLMs Encode Clinical Knowledge note\""
checked_on: 2026-11-11
as_of: 2026-11-11
confidence: medium
untested: 'Pointer note only. Paper not reproduced. Abs page and license checked 2026-11-11.'
category: arxiv
topic_area: 'ai-research: foundations'
tags: arxiv, instruction-tuning, rlhf, foundations
arxiv_id: 2203.02155
alias: 2203.02155-instructgpt
type: pointer
format: pointer
sources:
  - id: S1
    title: 'Training language models to follow instructions with human feedback'
    publisher: arXiv
    authors: 'Long Ouyang and 19 coauthors (listed on the arXiv abs page)'
    url: https://arxiv.org/abs/2203.02155
    version_read: 'v1 (4 Mar 2022)'
    accessed: 2026-11-11
    license: 'arXiv non-exclusive distribution license'
---
# Training language models to follow instructions with human feedback

## Summary
[Training language models to follow instructions with human feedback](https://arxiv.org/abs/2203.02155) (2022) is the InstructGPT paper. It described how to turn a raw pretrained language model into an assistant that follows user instructions, using human preference data and reinforcement learning from those preferences.

What it introduced: a three-stage recipe of supervised fine-tuning, a reward model trained on human comparisons, and a reinforcement-learning update (PPO) of the policy. The resulting InstructGPT models were smaller than GPT-3 but preferred by labelers on instruction-following.

What it led to: the default post-training path for chat models, later variants such as helpful-and-harmless RLHF and direct preference optimization, and today's reasoning-model training. Live notes: [2204-05862-helpful-harmless-rlhf](https://iinventedlight.com/library/notes/2204-05862-helpful-harmless-rlhf/) and [2305-18290-direct-preference-optimization](https://iinventedlight.com/library/notes/2305-18290-direct-preference-optimization/). This week's batch also has pointers for PPO and the earlier deep reinforcement learning from human preferences paper.

Where to read it: https://arxiv.org/abs/2203.02155 (v1, 4 Mar 2022).

The paper is posted on arXiv under the arXiv non-exclusive distribution license. This note links to it rather than reproducing it. The note itself is CC BY 4.0.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.

---

Agents: [/agents.md](/agents.md) | [/.well-known/agent-library.json](/.well-known/agent-library.json).