AI ChronicleAI 纪年 Full timeline →

Research breakthrough · curated entry

InstructGPT uses RLHF to follow instructions

Training language models to follow instructions with human feedback

Date
March 4, 2022
Category
Research breakthrough
Organization / Model
OpenAI
Evidence
1 source · 1 evidence family

OpenAI released InstructGPT, using reinforcement learning from human feedback to make models follow instructions more reliably. The work was the direct predecessor of ChatGPT's alignment method.

Curator's note: Cross-checked against the arXiv paper.

Curated sources · 1

  1. Paper preprints · arXiv2022-03-04

    Training language models to follow instructions with human feedback

    arxiv.org

Cite this page

This page is a curated static snapshot and can be cited directly. Preferred format GB/T 7714—2025 (electronic resource [EB/OL]); the other 5 styles open below, together with the machine-readable files.

GB/T 7714—2025 (electronic resource [EB/OL])

AI 纪年. InstructGPT 用 RLHF 让模型更好遵循指令[EB/OL]. (2022-03-04)[2026-11-03]. https://aihist.net/e/instructgpt-2022.html.

APA 7 · American Psychological Association

AI 纪年. (2022, March 4). InstructGPT 用 RLHF 让模型更好遵循指令 [Web page]. https://aihist.net/e/instructgpt-2022.html

MLA 9 · Modern Language Association

"InstructGPT 用 RLHF 让模型更好遵循指令." AI 纪年, 4 March 2022, aihist.net/e/instructgpt-2022.html. Accessed 3 Nov. 2026.

Chicago 17 · notes-bibliography

AI 纪年. "InstructGPT 用 RLHF 让模型更好遵循指令." Published March 4, 2022. Accessed Nov. 3, 2026. https://aihist.net/e/instructgpt-2022.html.

IEEE · citation-reference list

"InstructGPT 用 RLHF 让模型更好遵循指令," AI 纪年, 4 Mar. 2022. [Online]. Available: https://aihist.net/e/instructgpt-2022.html. [Accessed: 3 Nov. 2026].

ACM · author-date variant

AI 纪年. March 4, 2022. InstructGPT 用 RLHF 让模型更好遵循指令. Retrieved November 3, 2026 from https://aihist.net/e/instructgpt-2022.html.

All six styles come from one template and can be compared word for word. [引用日期], Accessed, urldate and Y2 all carry this page's build date, 2026-11-03 — a static page is an immutable snapshot, so substitute the date you actually consulted it. The address in every style is the archive's canonical page: this English page translates the same snapshot, and the work keeps its name, so one record cites both. For one-click copying use the same entry in the timeline, which takes the day you opened it. Expand the other styles before printing.