Log analysis is necessary for credible evaluation of AI agents
Agent benchmarks typically report only final outcomes: pass or fail. This threatens evaluation credibility in three ways. First, scores may be inflated or deflated by shortcuts and benchmark artifacts, misrepresenting capability. Second, benchmark performance may fail to predict real-world utility due to scaffold limitations and recurring failure modes. Finally, capability scores may conceal dangerous or catastrophic actions taken by the agent. We argue that log analysis -- the systematic tracking and analysis of the inputs, execution, and outputs of an AI agent -- is necessary to overcome these validity threats and promote credible agent evaluation. In this paper, we (1) present a taxonomy of threats to credible evaluation documented through log analysis, and (2) develop a set of guiding principles for log analysis. We illustrate these principles on tau-Bench Airline, revealing that pass^5 performance was under-elicited by nearly 50% and surfacing deployment failure modes invisible to outcome metrics. We conclude with pragmatic recommendations to increase uptake of log analysis, directed at diverse stakeholders including benchmark creators, model developers, independent evaluators, and deployers.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
Human Reliability Analysis for Oil and Gas Operations: Analysis of Existing Methods
In the petroleum industry, Quantitative Risk Analysis (QRA) has been one of the main tools for risk management. To date, QRA has mostly focused on technical barriers, despite many accidents having human failure as a prim…
ManagementSample Elicitation
It is important to collect credible training samples $(x,y)$ for building data-intensive learning systems (e.g., a deep learning system). Asking people to report complex distribution $p(x)$, though theoretically viable, …
What's in the box? Explaining the black-box model through an evaluation of its interpretable features
Algorithms are powerful and necessary tools behind a large part of the information we use every day. However, they may introduce new sources of bias, discrimination and other unfair practices that affect people who are u…
From Controlled to the Wild: Evaluation of Pentesting Agents for the Real-World
AI pentesting agents are increasingly credible as offensive security systems, but current benchmarks still provide limited guidance on which will perform best in real-world targets. Existing evaluation protocols assess a…
Bohrium + SciMaster: Building the Infrastructure and Ecosystem for Agentic Science at Scale
AI agents are emerging as a practical way to run multi-step scientific workflows that interleave reasoning with tool use and verification, pointing to a shift from isolated AI-assisted steps toward \emph{agentic science …