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

Stephen Choi; William Gazeley

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

Computational Engineering, Finance, and Science 2024-04-22 v1 Artificial Intelligence

Authors: Stephen Choi , William Gazeley

Abstract

This paper presents the LLM-ADE framework, a novel methodology for continued pre-training of large language models (LLMs) that addresses the challenges of catastrophic forgetting and double descent. LLM-ADE employs dynamic architectural adjustments, including selective block freezing and expansion, tailored to specific datasets. This strategy enhances model adaptability to new data while preserving previously acquired knowledge. We demonstrate LLM-ADE's effectiveness on the TinyLlama model across various general knowledge benchmarks, showing significant performance improvements without the drawbacks of traditional continuous training methods. This approach promises a more versatile and robust way to keep LLMs current and efficient in real-world applications.

Keywords

instruction tuning large language model large language model training

Cite

@article{arxiv.2404.13028,
  title  = {When Life gives you LLMs, make LLM-ADE: Large Language Models with Adaptive Data Engineering},
  author = {Stephen Choi and William Gazeley},
  journal= {arXiv preprint arXiv:2404.13028},
  year   = {2024}
}

Comments

6 pages, 3 tables and 3 figures

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

Abstract

Keywords

Cite

Comments

Related papers