LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Models

By Zhiqiang Hu, Yihuai Lan, Lei Wang, Wanyu Xu, Ee-Peng Lim, Roy Ka-Wei Lee, Lidong Bing, Soujanya Poria. Conference on Empirical Methods in Natural Language Processing, 2023

497 citations recorded by the source corpus.

The success of large language models (LLMs), like GPT-4 and ChatGPT, has led to the development of numerous cost-effective and accessible alternatives that are created by finetuning open-access LLMs with task-specific data (e.g., ChatDoctor) or instruction data (e.g., Alpaca). Among the various fine-tuning methods, adapter-based parameter-efficient fine-tuning (PEFT) is undoubtedly one of the most attractive topics, as it only requires fine-tuning a few external parameters instead of the entire LLMs while achieving comparable or even better performance. To enable further research on PEFT methods of LLMs, this paper presents LLM-Adapters, an easy-to-use framework that integrates various adapters into LLMs and can execute these adapter-based PEFT methods of LLMs for different tasks. The framework includes state-of-the-art open-access LLMs such as LLaMA, BLOOM, and GPT-J, as well as widely used adapters such as Series adapters, Parallel adapter, Prompt-based learning and Reparametrization-based methods. Moreover, we conduct extensive empirical studies on the impact of adapter types, placement locations, and hyper-parameters to the best design for each adapter-based methods. We evaluate the effectiveness of the adapters on fourteen datasets from two different reasoning tasks, Arithmetic Reasoning and Commonsense Reasoning. The results demonstrate that using adapter-based PEFT in smaller-scale LLMs (7B) with few extra trainable parameters yields comparable, and in some cases superior, performance to powerful LLMs (175B) in zero-shot inference on both reasoning tasks.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/bdb68c5e2369633b20e733774ac66eb4600c34d1

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Yu Qiao
2. Ping Luo
3. Jifeng Dai
4. Hongxu Yin
5. Wen Wang
6. Jun Jin
7. Qinglong Zhang
8. Shih-Yang Liu

Citations in the corpus

• DoRA: Weight-Decomposed Low-Rank Adaptation (2024, methodology): “These methods add extra soft tokens (prompts) to the initial input and focus solely on fine-tuning these trainable vectors, as seen in works like (Lester et al., 2021; Razdaibiedina et al., 2023; Wang et al., 2023).”
• A Survey of Large Language Models for Healthcare: from Data, Technology, and Applications to Accountability and Ethics (2023, unknown): “…2208instances DataLink CORD-19[287] Dialogue 1056660instances DataLink MMMLU[287] Dialogue 3787instances DataLink PubmedCausal[294] Dialogue 2446instances DataLink ChatDoctor[295] Dialogue 215000instances DataLink Alpaca-EN-AN[214] EnglishInstructions 52Kinstructions DataLink…”
• WizardLM: Empowering Large Pre-Trained Language Models to Follow Complex Instructions (2023, unknown): “(4) Open-source models trained from Llama 13B, including Baize (Xu et al., 2023), CAMEL (Li et al., 2023a), and Tulu (Wang et al., 2023)”
• EmbodiedGPT: Vision-Language Pre-Training via Embodied Chain of Thought (2023, background): “Recent works [8, 9, 10] also explore adapting the pre-trained LLMs to a new domain by efficient tuning strategies like LoRA [11].”
• ASPEN: High-Throughput LoRA Fine-Tuning of Large Language Models with a Single GPU (2023, unknown): “On the other hand, in scenarios involving fine-tuning jobs addressing distinct domain issues [5, 20, 22, 28, 48], each job would use diverse training datasets.”
• Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey (2024, methodology): “LLM-Adapters [78] builds an easy-to-use framework that incorporates various PEFT techniques into LLMs.”
• A Comprehensive Overview of Large Language Models (2023, unknown): “In robotics research, the LLMs have very promising applications as well, such as enhancing human-robot interaction [176], [177], [178], [179], task planning [180], [181], [182], navigation [183], [184], and learning [185], [186].”