Emerging trends: A gentle introduction to fine-tuning

Abstract

The previous Emerging Trends article (Church et al., 2021. Natural Language Engineering27(5), 631–645.) introduced deep nets to poets. Poets is an imperfect metaphor, intended as a gesture toward inclusion. The future for deep nets will benefit by reaching out to a broad audience of potential users, including people with little or no programming skills, and little interest in training models. That paper focused on inference, the use of pre-trained models, as is, without fine-tuning. The goal of this paper is to make fine-tuning more accessible to a broader audience. Since fine-tuning is more challenging than inference, the examples in this paper will require modest programming skills, as well as access to a GPU. Fine-tuning starts with a general purpose base (foundation) model and uses a small training set of labeled data to produce a model for a specific downstream application. There are many examples of fine-tuning in natural language processing (question answering (SQuAD) and GLUE benchmark), as well as vision and speech.

WITHDRAWN: ChatGPT-4 as a Board-Certified Surgeon: A Pilot Study

“…Looking ahead, several techniques to increase the likelihood of extracting desired LLM outputs have been described. In addition to optimized prompt engineering, experts advocate for training LLMs on extensive private datasets meticulously annotated by human experts, a process known as fine-tuning [44]. Of note, this technique is only possible with particular LLMs, not including GPT-4.…”

Verbal behavior and the future of social science.

“…These approaches all demonstrate the importance of understanding language in its full context and highlight the potential for automated tools to expand our ability to analyze language data on a large scale for understanding the verbal behavior of the individual and the situation in which the language originated.The development and adoption of new tools through interdisciplinary collaboration will be vital for addressing certain context-specific challenges. These tools may include specialized models, data collection techniques, or unique algorithmic approaches that are explicitly designed to handle nuances in various contexts (e.g., fine-tuning; Church et al, 2021). Such innovations can significantly contribute to our ability to extract meaningful information and insights from diverse data sources, making them a valuable asset in the social science toolkit.…”

Emerging trends: Deep nets thrive on scale

“…1. text books (Bishop 2016;Goodfellow et al, 2016), 2. more practical books (Géron 2019;Chollet 2021), 3. surveys b (LeCun et al, 2015Pouyanfar et al, 2018;Kumar et al, 2018;Liu et al, 2020;Qiu et al, 2020;Dong et al, 2021), and 4. tutorials: ACL-2022 (Church et al, 2022a), plus three articles in the Emerging Trends column in this journal (Church et al, 2021b(Church et al, , 2021a(Church et al, , 2022b).…”