From Release to Adoption: Challenges in Reusing Pre-trained AI Models for Downstream Developers
Pre-trained models (PTMs) have gained widespread popularity and achieved remarkable success across various fields, driven by their groundbreaking performance and easy accessibility through hosting providers. However, the challenges faced by downstream developers in reusing PTMs in software systems are less explored. To bridge this knowledge gap, we qualitatively created and analyzed a dataset of 840 PTM-related issue reports from 31 OSS GitHub projects. We systematically developed a comprehensive taxonomy of PTM-related challenges that developers face in downstream projects. Our study identifies seven key categories of challenges that downstream developers face in reusing PTMs, such as model usage, model performance, and output quality. We also compared our findings with existing taxonomies. Additionally, we conducted a resolution time analysis and, based on statistical tests, found that PTM-related issues take significantly longer to be resolved than issues unrelated to PTMs, with significant variation across challenge categories. We discuss the implications of our findings for practitioners and possibilities for future research.
Fri 12 SepDisplayed time zone: Auckland, Wellington change
10:30 - 12:00 | Session 13 - Reuse 1NIER Track / Research Papers Track / Industry Track / Registered Reports at Case Room 3 260-055 Chair(s): Banani Roy University of Saskatchewan | ||
10:30 15m | From Release to Adoption: Challenges in Reusing Pre-trained AI Models for Downstream Developers Research Papers Track Peerachai Banyongrakkul The University of Melbourne, Mansooreh Zahedi The Univeristy of Melbourne, Patanamon Thongtanunam University of Melbourne, Christoph Treude Singapore Management University, Haoyu Gao The University of Melbourne Pre-print | ||
10:45 15m | Are Classical Clone Detectors Good Enough For the AI Era? Research Papers Track Ajmain Inqiad Alam University of Saskatchewan, Palash Ranjan Roy University of Saskatchewan, Farouq Al-Omari Thompson Rivers University, Chanchal K. Roy University of Saskatchewan, Banani Roy University of Saskatchewan, Kevin Schneider University of Saskatchewan | ||
11:00 10m | Can LLMs Write CI? A Study on Automatic Generation of GitHub Actions Configurations NIER Track Taher A. Ghaleb Trent University, Dulina Rathnayake Department of Computer Science, Trent University, Peterborough, Canada Pre-print | ||
11:10 10m | A Preliminary Study on Large Language Models Self-Negotiation in Software Engineering NIER Track Chunrun Tao Kyushu University, Honglin Shu Kyushu University, Masanari Kondo Kyushu University, Yasutaka Kamei Kyushu University | ||
11:20 10m | CIgrate: Automating CI Service Migration with Large Language Models Registered Reports Md Nazmul Hossain Department of Computer Science, Trent University, Peterborough, Canada, Taher A. Ghaleb Trent University Pre-print | ||
11:30 15m | A Deep Dive into Retrieval-Augmented Generation for Code Completion: Experience on WeChat Industry Track Zezhou Yang Tencent Inc., Ting Peng Tencent Inc., Cuiyun Gao Harbin Institute of Technology, Shenzhen, Chaozheng Wang The Chinese University of Hong Kong, Hailiang Huang Tencent Inc., Yuetang Deng Tencent | ||
11:45 10m | Inferring Attributed Grammars from Parser Implementations NIER Track Andreas Pointner University of Applied Sciences Upper Austria, Hagenberg, Austria, Josef Pichler University of Applied Sciences Upper Austria, Herbert Prähofer Johannes Kepler University Linz Pre-print | ||