Conifer: Improving Complex Constrained Instruction-Following Ability of Large Language Models

TL;DR

Conifer is a new dataset and training scheme that significantly improves large language models' ability to follow complex, multi-level instructions with constraints, outperforming larger models on several benchmarks.

Contribution

We introduce Conifer, a high-quality instruction tuning dataset created with GPT-4, and a progressive learning scheme that enhances LLMs' complex instruction-following capabilities.

Findings

7B Conifer model outperforms larger open-source models.

Models trained with Conifer excel on complex instruction benchmarks.

Significant improvements in handling multi-level constraints.

Abstract

The ability of large language models (LLMs) to follow instructions is crucial to real-world applications. Despite recent advances, several studies have highlighted that LLMs struggle when faced with challenging instructions, especially those that include complex constraints, hindering their effectiveness in various tasks. To address this challenge, we introduce Conifer, a novel instruction tuning dataset, designed to enhance LLMs to follow multi-level instructions with complex constraints. Utilizing GPT-4, we curate the dataset by a series of LLM-driven refinement processes to ensure high quality. We also propose a progressive learning scheme that emphasizes an easy-to-hard progression, and learning from process feedback. Models trained with Conifer exhibit remarkable improvements in instruction-following abilities, especially for instructions with complex constraints. On several instruction-following benchmarks, our 7B model outperforms the state-of-the-art open-source 7B models, even exceeds the performance of models 10 times larger on certain metrics. All the code and Conifer dataset are available at https://www.github.com/ConiferLM/Conifer.