From Static to Dynamic: A Continual Learning Framework for Large Language Models

Mingzhe Du, Anh Tuan Luu, Bin Ji, See-kiong Ng
10/22/2023
cs.CL

Official code: Elfsong/DynaMind (8 stars, MIT, last commit 2025-03-19).

Abstract

The vast number of parameters in large language models (LLMs) endows them with remarkable capabilities, allowing them to excel in a variety of natural language processing tasks. However, this complexity also presents challenges, making LLMs difficult to train and inhibiting their ability to continuously assimilate new knowledge, which may lead to inaccuracies in their outputs. To mitigate these issues, this paper presents DynaMind, a novel continual learning framework designed for LLMs. DynaMind incorporates memory mechanisms to assimilate new knowledge and modular operators to enhance the model inference process with the newly assimilated knowledge, consequently improving the accuracies of LLMs' outputs. Benchmark experiments demonstrate DynaMind's effectiveness in overcoming these challenges. The code and demo of DynaMind are available on GitHub: https://github.com/Elfsong/DynaMind.

View on arXivView PDF

Code Implementations(1)

Elfsong/DynaMindOfficial100%
MIT
83HTML, ShellMay 1, 20231 years agoMIT

Similar Papers

When Life gives you LLMs, make LLM-ADE: Large Language Models with Adaptive Data Engineering

89% match2024

TRACE: A Comprehensive Benchmark for Continual Learning in Large Language Models

87% match1 repo1082023

To Each Language Its Tokenizer: Modular Tokenizers for Efficient Multilingual LLMs

87% match2026

ConPET: Continual Parameter-Efficient Tuning for Large Language Models

86% match1 repo142023

MoE-CT: A Novel Approach For Large Language Models Training With Resistance To Catastrophic Forgetting

86% match2024

Continual Learning for Large Language Models: A Survey

86% match2 repos5622024

Cite this paper

@article{du2023from,
  title  = {From Static to Dynamic: A Continual Learning Framework for Large Language Models},
  author = {Mingzhe Du and Anh Tuan Luu and Bin Ji and See-kiong Ng},
  year   = {2023},
  eprint = {2310.14248},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2310.14248v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2310.14248.svg)](https://codeofpaper.com/paper/2310.14248)

Discussion