Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kaur, Simran, Park, Simon, Goyal, Anirudh, Arora, Sanjeev
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913928518303744
author Kaur, Simran
Park, Simon
Goyal, Anirudh
Arora, Sanjeev
author_facet Kaur, Simran
Park, Simon
Goyal, Anirudh
Arora, Sanjeev
contents We introduce Instruct-SkillMix, an automated approach for creating diverse, high quality SFT data for instruction-following. The pipeline involves two stages, each leveraging an existing powerful LLM: (1) Skill extraction: uses the LLM to extract core "skills" for instruction-following by directly prompting the model. This is inspired by ``LLM metacognition'' of Didolkar et al. (2024); (2) Data generation: uses the powerful LLM to generate (instruction, response) data that exhibit a randomly chosen pair of these skills. Here, the use of random skill combinations promotes diversity and difficulty. The estimated cost of creating the dataset is under $600. Vanilla SFT (i.e., no PPO, DPO, or RL methods) on data generated from Instruct-SkillMix leads to strong gains on instruction following benchmarks such as AlpacaEval 2.0, MT-Bench, and WildBench. With just 4K examples, LLaMA-3-8B-Base achieves 42.76% length-controlled win rate on AlpacaEval 2.0, a level similar to frontier models like Claude 3 Opus and LLaMA-3.1-405B-Instruct. Ablation studies also suggest plausible reasons for why creating open instruction-tuning datasets via naive crowd-sourcing has proved difficult. In our dataset, adding 20% low quality answers (``shirkers'') causes a noticeable degradation in performance. The Instruct-SkillMix pipeline seems flexible and adaptable to other settings.
format Preprint
id arxiv_https___arxiv_org_abs_2408_14774
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning
Kaur, Simran
Park, Simon
Goyal, Anirudh
Arora, Sanjeev
Machine Learning
Computation and Language
We introduce Instruct-SkillMix, an automated approach for creating diverse, high quality SFT data for instruction-following. The pipeline involves two stages, each leveraging an existing powerful LLM: (1) Skill extraction: uses the LLM to extract core "skills" for instruction-following by directly prompting the model. This is inspired by ``LLM metacognition'' of Didolkar et al. (2024); (2) Data generation: uses the powerful LLM to generate (instruction, response) data that exhibit a randomly chosen pair of these skills. Here, the use of random skill combinations promotes diversity and difficulty. The estimated cost of creating the dataset is under $600. Vanilla SFT (i.e., no PPO, DPO, or RL methods) on data generated from Instruct-SkillMix leads to strong gains on instruction following benchmarks such as AlpacaEval 2.0, MT-Bench, and WildBench. With just 4K examples, LLaMA-3-8B-Base achieves 42.76% length-controlled win rate on AlpacaEval 2.0, a level similar to frontier models like Claude 3 Opus and LLaMA-3.1-405B-Instruct. Ablation studies also suggest plausible reasons for why creating open instruction-tuning datasets via naive crowd-sourcing has proved difficult. In our dataset, adding 20% low quality answers (``shirkers'') causes a noticeable degradation in performance. The Instruct-SkillMix pipeline seems flexible and adaptable to other settings.
title Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2408.14774