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
Measuring AI Ability to Complete Long Tasks
Measuring AI Ability to Complete Long Tasks
Thomas Kwa , Ben West 1 1 footnotemark: 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.
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.
1 Introduction
Figure 1: The length of tasks (measured by how long they take human professionals) that generalist autonomous frontier model agents can complete with 50% reliability has been doubling approximately every 7 months for the last 6 years (Section 4 ).
The shaded region represents 95% CI calculated by hierarchical bootstrap over task families, tasks, and task attempts.
Even if the absolute measurements are off by a factor of 10, the trend predicts that in under a decade we will see AI agents that can independently complete a large fraction of software tasks that currently take humans days or weeks(Section 7 ).
In the last five years, frontier AI systems have undergone a dramatic transformation in capabilities, evolving from basic text generation [ 1 ] to autonomously executing complex multi-hour machine learning research projects [ 2 ] .
Sufficiently capable AIs could perform dangerous, highly complex actions like autonomous development of chemical, b
... (truncated, 98 KB total)324cd2230cbea396 | Stable ID: OWYxNTMzMj