Showing 1–1 of 1 results for author: Tailor, P

Evaluating Large Language Models with fmeval

Authors: Pola Schwöbel, Luca Franceschi, Muhammad Bilal Zafar, Keerthan Vasist, Aman Malhotra, Tomer Shenhar, Pinal Tailor, Pinar Yilmaz, Michael Diamond, Michele Donini

Abstract: fmeval is an open source library to evaluate large language models (LLMs) in a range of tasks. It helps practitioners evaluate their model for task performance and along multiple responsible AI dimensions. This paper presents the library and exposes its underlying design principles: simplicity, coverage, extensibility and performance. We then present how these were implemented in the scientific an… ▽ More fmeval is an open source library to evaluate large language models (LLMs) in a range of tasks. It helps practitioners evaluate their model for task performance and along multiple responsible AI dimensions. This paper presents the library and exposes its underlying design principles: simplicity, coverage, extensibility and performance. We then present how these were implemented in the scientific and engineering choices taken when developing fmeval. A case study demonstrates a typical use case for the library: picking a suitable model for a question answering task. We close by discussing limitations and further work in the development of the library. fmeval can be found at https://github.com/aws/fmeval. △ Less

Submitted 15 July, 2024; originally announced July 2024.