Scalable Language Model with Generalized Continual Learning

TL;DR

This paper introduces a scalable language model with generalized continual learning capabilities, overcoming previous limitations by integrating adaptive re-parameterization and task retrieval, achieving state-of-the-art results across diverse benchmarks and tasks.

Contribution

The study presents the Joint Adaptive Re-Parameterization and Dynamic Task-related Knowledge Retrieval methods, enabling effective continual learning across multiple tasks and domains in large language models.

Findings

State-of-the-art performance on diverse benchmarks

Effective continual learning with minimal forgetting

Successful scaling across various domains and task types

Abstract

Continual learning has gained increasing importance as it facilitates the acquisition and refinement of scalable knowledge and skills in language models. However, existing methods typically encounter strict limitations and challenges in real-world scenarios, such as reliance on experience replay, optimization constraints, and inference task-ID. In this study, we introduce the Scalable Language Model (SLM) to overcome these limitations within a more challenging and generalized setting, representing a significant advancement toward practical applications for continual learning. Specifically, we propose the Joint Adaptive Re-Parameterization (JARe), integrated with Dynamic Task-related Knowledge Retrieval (DTKR), to enable adaptive adjustment of language models based on specific downstream tasks. This approach leverages the task distribution within the vector space, aiming to achieve a smooth and effortless continual learning process. Our method demonstrates state-of-the-art performance on diverse backbones and benchmarks, achieving effective continual learning in both full-set and few-shot scenarios with minimal forgetting. Moreover, while prior research primarily focused on a single task type such as classification, our study goes beyond, with the large language model, i.e., LLaMA-2, to explore the effects across diverse domains and task types, such that a single language model can be decently scaled to broader applications.