Toward an Evaluation Science for Generative AI Systems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Weidinger, Laura, Raji, Inioluwa Deborah, Wallach, Hanna, Mitchell, Margaret, Wang, Angelina, Salaudeen, Olawale, Bommasani, Rishi, Ganguli, Deep, Koyejo, Sanmi, Isaac, William
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910872394268672
author Weidinger, Laura
Raji, Inioluwa Deborah
Wallach, Hanna
Mitchell, Margaret
Wang, Angelina
Salaudeen, Olawale
Bommasani, Rishi
Ganguli, Deep
Koyejo, Sanmi
Isaac, William
author_facet Weidinger, Laura
Raji, Inioluwa Deborah
Wallach, Hanna
Mitchell, Margaret
Wang, Angelina
Salaudeen, Olawale
Bommasani, Rishi
Ganguli, Deep
Koyejo, Sanmi
Isaac, William
contents There is an increasing imperative to anticipate and understand the performance and safety of generative AI systems in real-world deployment contexts. However, the current evaluation ecosystem is insufficient: Commonly used static benchmarks face validity challenges, and ad hoc case-by-case audits rarely scale. In this piece, we advocate for maturing an evaluation science for generative AI systems. While generative AI creates unique challenges for system safety engineering and measurement science, the field can draw valuable insights from the development of safety evaluation practices in other fields, including transportation, aerospace, and pharmaceutical engineering. In particular, we present three key lessons: Evaluation metrics must be applicable to real-world performance, metrics must be iteratively refined, and evaluation institutions and norms must be established. Applying these insights, we outline a concrete path toward a more rigorous approach for evaluating generative AI systems.
format Preprint
id arxiv_https___arxiv_org_abs_2503_05336
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Toward an Evaluation Science for Generative AI Systems
Weidinger, Laura
Raji, Inioluwa Deborah
Wallach, Hanna
Mitchell, Margaret
Wang, Angelina
Salaudeen, Olawale
Bommasani, Rishi
Ganguli, Deep
Koyejo, Sanmi
Isaac, William
Artificial Intelligence
Machine Learning
There is an increasing imperative to anticipate and understand the performance and safety of generative AI systems in real-world deployment contexts. However, the current evaluation ecosystem is insufficient: Commonly used static benchmarks face validity challenges, and ad hoc case-by-case audits rarely scale. In this piece, we advocate for maturing an evaluation science for generative AI systems. While generative AI creates unique challenges for system safety engineering and measurement science, the field can draw valuable insights from the development of safety evaluation practices in other fields, including transportation, aerospace, and pharmaceutical engineering. In particular, we present three key lessons: Evaluation metrics must be applicable to real-world performance, metrics must be iteratively refined, and evaluation institutions and norms must be established. Applying these insights, we outline a concrete path toward a more rigorous approach for evaluating generative AI systems.
title Toward an Evaluation Science for Generative AI Systems
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2503.05336