Instruction Tuning with GPT-4

Baolin Peng, Chunyuan Li, Pengcheng He, Michel Galley, Jianfeng Gao
4/6/2023
cs.CLcs.AI

No official code found. Best implementation: Instruction-Tuning-with-GPT-4/GPT-4-LLM (4,331 stars, Apache-2.0, last commit 2023-06-11).

Abstract

Prior work has shown that finetuning large language models (LLMs) using machine-generated instruction-following data enables such models to achieve remarkable zero-shot capabilities on new tasks, and no human-written instructions are needed. In this paper, we present the first attempt to use GPT-4 to generate instruction-following data for LLM finetuning. Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models. We also collect feedback and comparison data from GPT-4 to enable a comprehensive evaluation and reward model training. We make our data generated using GPT-4 as well as our codebase publicly available.

View on arXivView PDF

Code Implementations(1)

Instruction-Tuning-with-GPT-4/GPT-4-LLMCommunity100%
Apache-2.0

Instruction Tuning with GPT-4

4,331307HTML, Jupyter NotebookApr 6, 20233 years agoApache-2.0
alpacachatgptgpt-4instruction-tuningllama

Similar Papers

AlpaGasus: Training A Better Alpaca with Fewer Data

90% match1 repo942023

Non-instructional Fine-tuning: Enabling Instruction-Following Capabilities in Pre-trained Language Models without Instruction-Following Data

89% match2024

LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions

89% match1 repo8222023

Efficient Finetuning Large Language Models For Vietnamese Chatbot

88% match2023

Exploring the Impact of Instruction Data Scaling on Large Language Models: An Empirical Study on Real-World Use Cases

87% match1 repo8.3k2023

From Base to Conversational: Japanese Instruction Dataset and Tuning Large Language Models

87% match1 repo72023

Cite this paper

@article{peng2023instruction,
  title  = {Instruction Tuning with GPT-4},
  author = {Baolin Peng and Chunyuan Li and Pengcheng He and Michel Galley and Jianfeng Gao},
  year   = {2023},
  eprint = {2304.03277},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2304.03277v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2304.03277.svg)](https://codeofpaper.com/paper/2304.03277)

Discussion