Measuring AI Long Tasks - arXiv
paperAuthors
Credibility Rating
Good quality. Reputable source with community review or editorial standards, but less rigorous than peer-reviewed venues.
Rating inherited from publication venue: arXiv
Introduces a novel metric (50%-task-completion time horizon) for measuring AI system capabilities relative to human performance, addressing the gap between benchmark scores and real-world AI competence—relevant for evaluating AI capabilities and safety implications.
Paper Details
Metadata
Abstract
Despite rapid progress on AI benchmarks, the real-world meaning of benchmark performance remains unclear. To quantify the capabilities of AI systems in terms of human capabilities, we propose a new metric: 50%-task-completion time horizon. This is the time humans typically take to complete tasks that AI models can complete with 50% success rate. We first timed humans with relevant domain expertise on a combination of RE-Bench, HCAST, and 66 novel shorter tasks. On these tasks, current frontier AI models such as Claude 3.7 Sonnet have a 50% time horizon of around 50 minutes. Furthermore, frontier AI time horizon has been doubling approximately every seven months since 2019, though the trend may have accelerated in 2024. The increase in AI models' time horizons seems to be primarily driven by greater reliability and ability to adapt to mistakes, combined with better logical reasoning and tool use capabilities. We discuss the limitations of our results -- including their degree of external validity -- and the implications of increased autonomy for dangerous capabilities. If these results generalize to real-world software tasks, extrapolation of this trend predicts that within 5 years, AI systems will be capable of automating many software tasks that currently take humans a month.
Summary
This paper introduces a new metric called '50%-task-completion time horizon' to measure AI capabilities in human-relatable terms—specifically, the time humans with domain expertise typically need to complete tasks that AI models can solve with 50% success rate. The authors evaluated frontier models like Claude 3.7 Sonnet on a dataset combining existing benchmarks and 66 novel tasks, finding current models achieve approximately 50 minutes on this metric. Notably, the AI time horizon has doubled roughly every seven months since 2019, driven primarily by improvements in reliability, error adaptation, logical reasoning, and tool use. If this trend continues, the authors project that within 5 years, AI systems could automate many software tasks currently requiring a month of human effort.
Cited by 2 pages
| Page | Type | Quality |
|---|---|---|
| Long-Horizon Autonomous Tasks | Capability | 65.0 |
| Third-Party Model Auditing | Approach | 64.0 |
Cached Content Preview
HTML conversions [sometimes display errors](https://info.dev.arxiv.org/about/accessibility_html_error_messages.html) due to content that did not convert correctly from the source. This paper uses the following packages that are not yet supported by the HTML conversion tool. Feedback on these issues are not necessary; they are known and are being worked on.
- failed: mdframed.sty
Authors: achieve the best HTML results from your LaTeX submissions by following these [best practices](https://info.arxiv.org/help/submit_latex_best_practices.html).
[License: arXiv.org perpetual non-exclusive license](https://info.arxiv.org/help/license/index.html#licenses-available)
arXiv:2503.14499v1 \[cs.AI\] 18 Mar 2025
# Measuring AI Ability to Complete Long Tasks
Report issue for preceding element
Thomas Kwa , Ben West11footnotemark: 1 ,
Joel Becker, Amy Deng, Katharyn Garcia,
Max Hasin, Sami Jawhar,
Megan Kinniment, Nate Rush, Sydney Von Arx
Ryan Bloom, Thomas Broadley, Haoxing Du, Brian Goodrich, Nikola Jurkovic,
Luke Harold Miles , Seraphina Nix, Tao Lin, Neev Parikh, David Rein,
Lucas Jun Koba Sato, Hjalmar Wijk, Daniel M. Ziegler
Elizabeth Barnes, Lawrence Chan
Model Evaluation & Threat Research (METR)Equal contribution.Corresponding author, ben@metr.org.Ohm Chip. Work done at METR.Anthropic. Work done at METR.
Report issue for preceding element
###### Abstract
Report issue for preceding element
Despite rapid progress on AI benchmarks, the real-world meaning of benchmark performance remains unclear.
To quantify the capabilities of AI systems in terms of human capabilities, we propose a new metric: _50%-task-completion time horizon_. This is the time humans typically take to complete tasks that AI models can complete with 50% success rate.
We first timed humans with relevant domain expertise on a combination of RE-Bench, HCAST, and 66 novel shorter tasks. On these tasks, current frontier AI models such as Claude 3.7 Sonnet have a 50% time horizon of around 50 minutes.
Furthermore, frontier AI time horizon has been doubling approximately every seven months since 2019, though the trend may have accelerated in 2024.
The increase in AI models’ time horizons seems to be primarily driven by greater reliability and ability to adapt to mistakes, combined with better logical reasoning and tool use capabilities.
We discuss the limitations of our results—including their degree of external validity—and the implications of increased autonomy for dangerous capabilities.
If these results generalize to real-world software tasks, extrapolation of this trend predicts that within 5 years, AI systems will be capable of automating many software tasks that currently take humans a month.
Report issue for preceding element
## 1 Introduction
Report issue for preceding elementFigure 1: The length of tasks (measured by how long they take human professionals) that generalist aut
... (truncated, 98 KB total)324cd2230cbea396 | Stable ID: sid_ZFvPpfKn3d