Software Dependencies 2.0: An Empirical Study of Reuse and Integration of Pre-Trained Models in Open-Source Projects
#Pre-trained models #Software Dependencies 2.0 #Open-source projects #Model reuse #Dependency management #Empirical study
📌 Key Takeaways
- Pre-trained models (PTMs) can be reused for new tasks, reducing the need for training from scratch.
- PTMs introduce a new class of software dependency called Software Dependencies 2.0, which includes learned models and associated artifacts.
- The paper presents an empirical study that examines how PTMs are reused and integrated in open-source projects.
- The research aims to quantify the prevalence and integration patterns of PTMs, addressing the current lack of systematic data on their use.
📖 Full Retelling
The article on arXiv, titled "Software Dependencies 2.0: An Empirical Study of Reuse and Integration of Pre-Trained Models in Open-Source Projects," reports an empirical investigation into how pre-trained machine learning models are reused and integrated within open-source software. It acknowledges the emergence of a new dependency class, termed Software Dependencies 2.0, that extends traditional library dependencies to encompass learned models and their related artifacts. The study seeks to quantify the prevalence and integration patterns of PTMs in open-source projects, motivated by the widespread adoption of PTMs and the scarcity of systematic data about their usage.
🏷️ Themes
Software engineering, Machine learning deployment, Dependency management, Open-source software, Empirical research
Entity Intersection Graph
No entity connections available yet for this article.
Original Source
arXiv:2509.06085v2 Announce Type: replace-cross
Abstract: Pre-trained models (PTMs) are machine learning models that have been trained in advance, often on large-scale data, and can be reused for new tasks, thereby reducing the need for costly training from scratch. Their widespread adoption introduces a new class of software dependency, which we term Software Dependencies 2.0, extending beyond conventional libraries to learned behaviors embodied in trained models and their associated artifacts
Read full article at source