Blog post · 2025
Measuring AI Ability to Complete Long Tasks
| URL | https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/ |
|---|---|
| Topics | ReplicHAI / Agent Evals / 03 ML & research engineering |
| Tags | agent-eval |
Abstract
Measures: 50% task-completion time horizon trend. Metric: Time horizon (minutes).
BibTeX
@misc{metr_measuring_2025,
title = {Measuring {AI} {Ability} to {Complete} {Long} {Tasks}},
url = {https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/},
abstract = {Measures: 50\% task-completion time horizon trend. Metric: Time horizon (minutes).},
author = {{METR}},
month = mar,
year = {2025},
keywords = {agent-eval},
}