Parrot: Enhancing Multi-Turn Instruction Following for Large Language Models

TL;DR

Parrot is a novel approach that improves large language models' ability to follow multi-turn instructions by introducing new data collection, a preference optimization strategy, and a dedicated benchmark, leading to significant performance gains.

Contribution

The paper presents a new method for collecting multi-turn instructions, a context-aware optimization strategy, and a benchmark for evaluating multi-turn instruction following in LLMs.

Findings

Parrot improves multi-turn instruction following by up to 7.2%.

Introduces a new multi-turn instruction dataset and benchmark.

Enhances LLM performance on complex multi-turn queries.

Abstract

Humans often interact with large language models (LLMs) in multi-turn interaction to obtain desired answers or more information. However, most existing studies overlook the multi-turn instruction following ability of LLMs, in terms of training dataset, training method, and evaluation benchmark. In this paper, we introduce Parrot, a solution aiming to enhance multi-turn instruction following for LLMs. First, we introduce an efficient but effective method for collecting multi-turn instructions that feature human-like queries, such as anaphora and ellipsis. Second, we propose a context-aware preference optimization strategy to further enhance LLMs for complex queries in multi-turn interaction. Moreover, to quantitatively evaluate LLMs in multi-turn instruction following, we manually build a multi-turn benchmark derived from existing ones. Extensive experiments show that Parrot improves current LLMs by up to 7.2% in multi-turn instruction following. Our dataset and codes will be open-sourced to facilitate future research.