Self-Alignment with Instruction Backtranslation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Xian, Yu, Ping, Zhou, Chunting, Schick, Timo, Levy, Omer, Zettlemoyer, Luke, Weston, Jason, Lewis, Mike
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909134385840128
author Li, Xian
Yu, Ping
Zhou, Chunting
Schick, Timo
Levy, Omer
Zettlemoyer, Luke
Weston, Jason
Lewis, Mike
author_facet Li, Xian
Yu, Ping
Zhou, Chunting
Schick, Timo
Levy, Omer
Zettlemoyer, Luke
Weston, Jason
Lewis, Mike
contents We present a scalable method to build a high quality instruction following language model by automatically labelling human-written text with corresponding instructions. Our approach, named instruction backtranslation, starts with a language model finetuned on a small amount of seed data, and a given web corpus. The seed model is used to construct training examples by generating instruction prompts for web documents (self-augmentation), and then selecting high quality examples from among these candidates (self-curation). This data is then used to finetune a stronger model. Finetuning LLaMa on two iterations of our approach yields a model that outperforms all other LLaMa-based models on the Alpaca leaderboard not relying on distillation data, demonstrating highly effective self-alignment.
format Preprint
id arxiv_https___arxiv_org_abs_2308_06259
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Self-Alignment with Instruction Backtranslation
Li, Xian
Yu, Ping
Zhou, Chunting
Schick, Timo
Levy, Omer
Zettlemoyer, Luke
Weston, Jason
Lewis, Mike
Computation and Language
We present a scalable method to build a high quality instruction following language model by automatically labelling human-written text with corresponding instructions. Our approach, named instruction backtranslation, starts with a language model finetuned on a small amount of seed data, and a given web corpus. The seed model is used to construct training examples by generating instruction prompts for web documents (self-augmentation), and then selecting high quality examples from among these candidates (self-curation). This data is then used to finetune a stronger model. Finetuning LLaMa on two iterations of our approach yields a model that outperforms all other LLaMa-based models on the Alpaca leaderboard not relying on distillation data, demonstrating highly effective self-alignment.
title Self-Alignment with Instruction Backtranslation
topic Computation and Language
url https://arxiv.org/abs/2308.06259