What Makes an Evaluation Useful? Common Pitfalls and Best Practices

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Gekker, Gil, Segal, Meirav, Lahav, Dan, Nevo, Omer
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916666376454144
author Gekker, Gil
Segal, Meirav
Lahav, Dan
Nevo, Omer
author_facet Gekker, Gil
Segal, Meirav
Lahav, Dan
Nevo, Omer
contents Following the rapid increase in Artificial Intelligence (AI) capabilities in recent years, the AI community has voiced concerns regarding possible safety risks. To support decision-making on the safe use and development of AI systems, there is a growing need for high-quality evaluations of dangerous model capabilities. While several attempts to provide such evaluations have been made, a clear definition of what constitutes a "good evaluation" has yet to be agreed upon. In this practitioners' perspective paper, we present a set of best practices for safety evaluations, drawing on prior work in model evaluation and illustrated through cybersecurity examples. We first discuss the steps of the initial thought process, which connects threat modeling to evaluation design. Then, we provide the characteristics and parameters that make an evaluation useful. Finally, we address additional considerations as we move from building specific evaluations to building a full and comprehensive evaluation suite.
format Preprint
id arxiv_https___arxiv_org_abs_2503_23424
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle What Makes an Evaluation Useful? Common Pitfalls and Best Practices
Gekker, Gil
Segal, Meirav
Lahav, Dan
Nevo, Omer
Machine Learning
Artificial Intelligence
Computation and Language
Cryptography and Security
Computers and Society
Following the rapid increase in Artificial Intelligence (AI) capabilities in recent years, the AI community has voiced concerns regarding possible safety risks. To support decision-making on the safe use and development of AI systems, there is a growing need for high-quality evaluations of dangerous model capabilities. While several attempts to provide such evaluations have been made, a clear definition of what constitutes a "good evaluation" has yet to be agreed upon. In this practitioners' perspective paper, we present a set of best practices for safety evaluations, drawing on prior work in model evaluation and illustrated through cybersecurity examples. We first discuss the steps of the initial thought process, which connects threat modeling to evaluation design. Then, we provide the characteristics and parameters that make an evaluation useful. Finally, we address additional considerations as we move from building specific evaluations to building a full and comprehensive evaluation suite.
title What Makes an Evaluation Useful? Common Pitfalls and Best Practices
topic Machine Learning
Artificial Intelligence
Computation and Language
Cryptography and Security
Computers and Society
url https://arxiv.org/abs/2503.23424