This position paper argues for two claims regarding AI testing and evaluation. First, to remain informative about deployment behaviour, evaluations need account for the possibility that AI systems understand their circumstances and reason strategically. Second, game-theoretic analysis can inform evaluation design by formalising and scrutinising the reasoning in evaluation-based safety cases. Drawing on examples from existing AI systems, a review of relevant research, and formal strategic analysis of a stylised evaluation scenario, we present evidence for these claims and motivate several research directions.

Benchmarks play a significant role in how researchers and the public understand generative AI systems. However, the widespread use of benchmark scores to communicate about model capabilities has led to criticisms of validity, especially whether benchmarks test what they claim to test (i.e. construct validity) and whether benchmark evaluations are representative of how models are used in the wild (i.e. ecological validity). In this work we explore how to create an LLM benchmark that addresses these issues by taking a human-centered approach. We focus on designing a domain-oriented benchmark for journalism practitioners, drawing on insights from a workshop of 23 journalism professionals. Our workshop findings surface specific challenges that inform benchmark design opportunities, which we instantiate in a case study that addresses underlying criticisms and specific domain concerns. Through our findings and design case study, this work provides design guidance for developing benchmarks that are better tuned to specific domains.

This position paper argues for two claims regarding AI testing and evaluation. First, to remain informative about deployment behaviour, evaluations need account for the possibility that AI systems understand their circumstances and reason strategically. Second, game-theoretic analysis can inform evaluation design by formalising and scrutinising the reasoning in evaluation-based safety cases. Drawing on examples from existing AI systems, a review of relevant research, and formal strategic analysis of a stylised evaluation scenario, we present evidence for these claims and motivate several research directions.

This document presents a preliminary compilation of general-purpose AI (GPAI) evaluation practices that may promote internal validity, external validity and reproducibility. It includes suggestions for human uplift studies and benchmark evaluations, as well as cross-cutting suggestions that may apply to many different evaluation types. Suggestions are organised across four stages in the evaluation life cycle: design, implementation, execution and documentation. Drawing from established practices in machine learning, statistics, psychology, economics, biology and other fields recognised to have important lessons for AI evaluation, these suggestions seek to contribute to the conversation on the nascent and evolving field of the science of GPAI evaluations. The intended audience of this document includes providers of GPAI models presenting systemic risk (GPAISR), for whom the EU AI Act lays out specific evaluation requirements; third-party evaluators; policymakers assessing the rigour of evaluations; and academic researchers developing or conducting GPAI evaluations.

AI evaluations have become critical tools for assessing large language model capabilities and safety. This paper presents practical insights from eight months of maintaining $inspect\_evals$, an open-source repository of 70+ community-contributed AI evaluations. We identify key challenges in implementing and maintaining AI evaluations and develop solutions including: (1) a structured cohort management framework for scaling community contributions, (2) statistical methodologies for optimal resampling and cross-model comparison with uncertainty quantification, and (3) systematic quality control processes for reproducibility. Our analysis reveals that AI evaluation requires specialized infrastructure, statistical rigor, and community coordination beyond traditional software development practices.

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 approaches 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.

Research in AI evaluation has grown increasingly complex and multidisciplinary, attracting researchers with diverse backgrounds and objectives. As a result, divergent evaluation paradigms have emerged, often developing in isolation, adopting conflicting terminologies, and overlooking each other's contributions. This fragmentation has led to insular research trajectories and communication barriers both among different paradigms and with the general public, contributing to unmet expectations for deployed AI systems. To help bridge this insularity, in this paper we survey recent work in the AI evaluation landscape and identify six main paradigms. We characterise major recent contributions within each paradigm across key dimensions related to their goals, methodologies and research cultures. By clarifying the unique combination of questions and approaches associated with each paradigm, we aim to increase awareness of the breadth of current evaluation approaches and foster cross-pollination between different paradigms. We also identify potential gaps in the field to inspire future research directions.

AI evaluations are an important component of the AI governance toolkit, underlying current approaches to safety cases for preventing catastrophic risks. Our paper examines what these evaluations can and cannot tell us. Evaluations can establish lower bounds on AI capabilities and assess certain misuse risks given sufficient effort from evaluators. Unfortunately, evaluations face fundamental limitations that cannot be overcome within the current paradigm. These include an inability to establish upper bounds on capabilities, reliably forecast future model capabilities, or robustly assess risks from autonomous AI systems. This means that while evaluations are valuable tools, we should not rely on them as our main way of ensuring AI systems are safe. We conclude with recommendations for incremental improvements to frontier AI safety, while acknowledging these fundamental limitations remain unsolved.

As AI systems advance, AI evaluations are becoming an important pillar of regulations for ensuring safety. We argue that such regulation should require developers to explicitly identify and justify key underlying assumptions about evaluations as part of their case for safety. We identify core assumptions in AI evaluations (both for evaluating existing models and forecasting future models), such as comprehensive threat modeling, proxy task validity, and adequate capability elicitation. Many of these assumptions cannot currently be well justified. If regulation is to be based on evaluations, it should require that AI development be halted if evaluations demonstrate unacceptable danger or if these assumptions are inadequately justified. Our presented approach aims to enhance transparency in AI development, offering a practical path towards more effective governance of advanced AI systems.

As AI systems appear to exhibit ever-increasing capability and generality, assessing their true potential and safety becomes paramount. This paper contends that the prevalent evaluation methods for these systems are fundamentally inadequate, heightening the risks and potential hazards associated with AI. I argue that a reformation is required in the way we evaluate AI systems and that we should look towards cognitive sciences for inspiration in our approaches, which have a longstanding tradition of assessing general intelligence across diverse species. We will identify some of the difficulties that need to be overcome when applying cognitively-inspired approaches to general-purpose AI systems and also analyse the emerging area of "Evals". The paper concludes by identifying promising research pathways that could refine AI evaluation, advancing it towards a rigorous scientific domain that contributes to the development of safe AI systems.