InsCL: A Data-efficient Continual Learning Paradigm for Fine-tuning Large Language Models with Instructions

TL;DR

InsCL introduces a novel instruction-based continual learning paradigm that dynamically replays data based on task similarity and instruction complexity, significantly improving LLM fine-tuning efficiency and performance across multiple tasks.

Contribution

The paper proposes InsCL, a new replay strategy utilizing instruction similarity and complexity metrics to enhance continual learning for large language models.

Findings

Achieves 3.0% relative gain over random replay

Achieves 27.96% relative gain over no replay

Consistent performance improvements across 16 tasks

Abstract

Instruction tuning effectively optimizes Large Language Models (LLMs) for downstream tasks. Due to the changing environment in real-life applications, LLMs necessitate continual task-specific adaptation without catastrophic forgetting. Considering the heavy computational cost, replay-based Continual Learning (CL) methods are the simplest and most widely used for LLMs to address the forgetting issue. However, traditional replay-based methods do not fully utilize instructions to customize the replay strategy. In this work, we propose a novel paradigm called Instruction-based Continual Learning (InsCL). InsCL dynamically replays previous data based on task similarity, calculated by Wasserstein Distance with instructions. Moreover, we further introduce an Instruction Information Metric (InsInfo) to quantify the complexity and diversity of instructions. According to InsInfo, InsCL guides the replay process more inclined to high-quality data. We conduct extensive experiments over 16 tasks with different training orders, observing consistent performance improvements of InsCL. When all tasks have been trained, InsCL achieves performance gains of 3.0 Relative Gain compared with Random Replay, and 27.96 Relative Gain compared with No Replay.